Add release v2.4.1

Fix two race conditions in Kubernetes IPAM reclaim

Fix #3384

.gitignore

@@ -46,8 +46,8 @@ prog/weavewait/weavewait_noop
 prog/weavewait/weavewait_nomcast
 prog/weaveutil/weaveutil
 prog/*/Dockerfile.weaveworks
-prog/kube-peers/kube-peers
-prog/weave-kube/kube-peers
+prog/kube-utils/kube-utils
+prog/weave-kube/kube-utils
 prog/weave-npc/weave-npc
 prog/plugin/plugin
 testing/cover/cover
@@ -68,6 +68,7 @@ test/tls/tls
 test/tls/*.pem
 test/coverage
 test/coverage.*
+test/images/network-tester/webserver
 *qemu-*-static
 terraform.tfstate
 terraform.tfstate.backup

.gitmodules

@@ -1,180 +1,3 @@
 [submodule "tools"]
 	path = tools
 	url = https://github.com/weaveworks/build-tools
-[submodule "vendor/github.com/vishvananda/netlink"]
-	path = vendor/github.com/vishvananda/netlink
-	url = https://github.com/vishvananda/netlink.git
-[submodule "vendor/github.com/gorilla/mux"]
-	path = vendor/github.com/gorilla/mux
-	url = https://github.com/gorilla/mux.git
-[submodule "vendor/github.com/gorilla/context"]
-	path = vendor/github.com/gorilla/context
-	url = https://github.com/gorilla/context.git
-[submodule "vendor/github.com/google/gopacket"]
-	path = vendor/github.com/google/gopacket
-	url = https://github.com/google/gopacket.git
-[submodule "vendor/github.com/weaveworks/go-odp"]
-	path = vendor/github.com/weaveworks/go-odp
-	url = https://github.com/weaveworks/go-odp.git
-[submodule "vendor/github.com/opencontainers/runc"]
-	path = vendor/github.com/opencontainers/runc
-	url = https://github.com/opencontainers/runc.git
-[submodule "vendor/github.com/fsouza/go-dockerclient"]
-	path = vendor/github.com/fsouza/go-dockerclient
-	url = https://github.com/fsouza/go-dockerclient.git
-[submodule "vendor/github.com/docker/libnetwork"]
-	path = vendor/github.com/docker/libnetwork
-	url = https://github.com/docker/libnetwork.git
-[submodule "vendor/github.com/docker/docker"]
-	path = vendor/github.com/docker/docker
-	url = https://github.com/docker/docker.git
-[submodule "vendor/github.com/Sirupsen/logrus"]
-	path = vendor/github.com/Sirupsen/logrus
-	url = https://github.com/Sirupsen/logrus.git
-[submodule "vendor/github.com/andybalholm/go-bit"]
-	path = vendor/github.com/andybalholm/go-bit
-	url = https://github.com/andybalholm/go-bit.git
-[submodule "vendor/github.com/miekg/dns"]
-	path = vendor/github.com/miekg/dns
-	url = https://github.com/miekg/dns.git
-[submodule "vendor/github.com/stretchr/testify"]
-	path = vendor/github.com/stretchr/testify
-	url = https://github.com/stretchr/testify.git
-[submodule "vendor/golang.org/x/crypto"]
-	path = vendor/golang.org/x/crypto
-	url = https://github.com/golang/crypto.git
-[submodule "vendor/github.com/docker/machine"]
-	path = vendor/github.com/docker/machine
-	url = https://github.com/docker/machine.git
-[submodule "vendor/github.com/weaveworks/mesh"]
-	path = vendor/github.com/weaveworks/mesh
-	url = https://github.com/weaveworks/mesh
-[submodule "vendor/github.com/vishvananda/netns"]
-	path = vendor/github.com/vishvananda/netns
-	url = https://github.com/vishvananda/netns.git
-[submodule "vendor/github.com/boltdb/bolt"]
-	path = vendor/github.com/boltdb/bolt
-	url = https://github.com/boltdb/bolt
-[submodule "vendor/github.com/weaveworks/go-checkpoint"]
-	path = vendor/github.com/weaveworks/go-checkpoint
-	url = https://github.com/weaveworks/go-checkpoint
-[submodule "vendor/github.com/hashicorp/go-cleanhttp"]
-	path = vendor/github.com/hashicorp/go-cleanhttp
-	url = http://github.com/hashicorp/go-cleanhttp
-[submodule "vendor/github.com/coreos/go-iptables"]
-	path = vendor/github.com/coreos/go-iptables
-	url = https://github.com/coreos/go-iptables
-[submodule "vendor/github.com/pkg/profile"]
-	path = vendor/github.com/pkg/profile
-	url = https://github.com/pkg/profile
-[submodule "vendor/github.com/j-keck/arping"]
-	path = vendor/github.com/j-keck/arping
-	url = https://github.com/j-keck/arping
-[submodule "vendor/github.com/aws/aws-sdk-go"]
-	path = vendor/github.com/aws/aws-sdk-go
-	url = https://github.com/aws/aws-sdk-go
-[submodule "vendor/golang.org/x/sys"]
-	path = vendor/golang.org/x/sys
-	url = https://go.googlesource.com/sys
-[submodule "vendor/github.com/beorn7/perks"]
-	path = vendor/github.com/beorn7/perks
-	url = https://github.com/beorn7/perks
-[submodule "vendor/github.com/prometheus/procfs"]
-	path = vendor/github.com/prometheus/procfs
-	url = https://github.com/prometheus/procfs
-[submodule "vendor/github.com/prometheus/client_golang"]
-	path = vendor/github.com/prometheus/client_golang
-	url = https://github.com/prometheus/client_golang
-[submodule "vendor/github.com/prometheus/client_model"]
-	path = vendor/github.com/prometheus/client_model
-	url = https://github.com/prometheus/client_model
-[submodule "vendor/github.com/prometheus/common"]
-	path = vendor/github.com/prometheus/common
-	url = https://github.com/prometheus/common
-[submodule "vendor/github.com/golang/protobuf"]
-	path = vendor/github.com/golang/protobuf
-	url = https://github.com/golang/protobuf
-[submodule "vendor/github.com/matttproud/golang_protobuf_extensions"]
-	path = vendor/github.com/matttproud/golang_protobuf_extensions
-	url = https://github.com/matttproud/golang_protobuf_extensions
-[submodule "vendor/k8s.io/client-go"]
-	path = vendor/k8s.io/client-go
-	url = https://github.com/kubernetes/client-go
-[submodule "vendor/github.com/pkg/errors"]
-	path = vendor/github.com/pkg/errors
-	url = https://github.com/pkg/errors
-[submodule "vendor/github.com/spf13/cobra"]
-	path = vendor/github.com/spf13/cobra
-	url = https://github.com/spf13/cobra
-[submodule "vendor/github.com/spf13/plag"]
-	path = vendor/github.com/spf13/plag
-	url = https://github.com/spf13/pflag
-[submodule "vendor/github.com/spf13/pflag"]
-	path = vendor/github.com/spf13/pflag
-	url = https://github.com/spf13/pflag
-[submodule "vendor/github.com/mgutz/ansi"]
-	path = vendor/github.com/mgutz/ansi
-	url = https://github.com/mgutz/ansi
-[submodule "vendor/github.com/estesp/manifest-tool"]
-	path = vendor/github.com/estesp/manifest-tool
-	url = https://github.com/estesp/manifest-tool
-[submodule "vendor/github.com/weaveworks/common"]
-	path = vendor/github.com/weaveworks/common
-	url = https://github.com/weaveworks/common
-[submodule "vendor/github.com/containernetworking/cni"]
-	path = vendor/github.com/containernetworking/cni
-	url = https://github.com/containernetworking/cni
-[submodule "vendor/google.golang.org/grpc"]
-	path = vendor/google.golang.org/grpc
-	url = https://github.com/grpc/grpc-go
-[submodule "vendor/github.com/docker/go-units"]
-	path = vendor/github.com/docker/go-units
-	url = https://github.com/docker/go-units
-[submodule "vendor/golang.org/x/net"]
-	path = vendor/golang.org/x/net
-	url = https://github.com/golang/net
-[submodule "vendor/k8s.io/api"]
-	path = vendor/k8s.io/api
-	url = https://github.com/kubernetes/api
-[submodule "vendor/k8s.io/apimachinery"]
-	path = vendor/k8s.io/apimachinery
-	url = https://github.com/kubernetes/apimachinery
-[submodule "vendor/github.com/PuerkitoBio/purell"]
-	path = vendor/github.com/PuerkitoBio/purell
-	url = https://github.com/PuerkitoBio/purell
-[submodule "vendor/github.com/PuerkitoBio/urlesc"]
-	path = vendor/github.com/PuerkitoBio/urlesc
-	url = https://github.com/PuerkitoBio/urlesc
-[submodule "vendor/github.com/emicklei/go-restful"]
-	path = vendor/github.com/emicklei/go-restful
-	url = https://github.com/emicklei/go-restful
-[submodule "vendor/github.com/go-openapi/jsonpointer"]
-	path = vendor/github.com/go-openapi/jsonpointer
-	url = https://github.com/go-openapi/jsonpointer
-[submodule "vendor/github.com/go-openapi/jsonreference"]
-	path = vendor/github.com/go-openapi/jsonreference
-	url = https://github.com/go-openapi/jsonreference
-[submodule "vendor/github.com/go-openapi/spec"]
-	path = vendor/github.com/go-openapi/spec
-	url = https://github.com/go-openapi/spec
-[submodule "vendor/github.com/go-openapi/swag"]
-	path = vendor/github.com/go-openapi/swag
-	url = https://github.com/go-openapi/swag
-[submodule "vendor/github.com/golang/glog"]
-	path = vendor/github.com/golang/glog
-	url = https://github.com/golang/glog
-[submodule "vendor/github.com/google/gofuzz"]
-	path = vendor/github.com/google/gofuzz
-	url = https://github.com/google/gofuzz
-[submodule "vendor/github.com/mailru/easyjson"]
-	path = vendor/github.com/mailru/easyjson
-	url = https://github.com/mailru/easyjson
-[submodule "vendor/k8s.io/kube-openapi"]
-	path = vendor/k8s.io/kube-openapi
-	url = https://github.com/kubernetes/kube-openapi
-[submodule "vendor/golang.org/x/text"]
-	path = vendor/golang.org/x/text
-	url = https://go.googlesource.com/text
-[submodule "vendor/gopkg.in/yaml.v2"]
-	path = vendor/gopkg.in/yaml.v2
-	url = https://gopkg.in/yaml.v2

CHANGELOG.md

@@ -1,3 +1,117 @@
+## Release 2.4.1
+
+This release fixes several bugs causing inconsistencies in IPAM for
+Kubernetes users whose clusters scale up and down over time.
+
+### Bug fixes
+
+* Nodes unable to connect after Kubernetes addon erroneously reclaimed
+  node without any IP addresses #3392, #3393
+* Kubernetes addon could have run out of free IP addresses after nodes
+  are deleted #3384, #3400
+* Kubernetes addon had reduced free IP addresses due to not reclaiming
+  IP addresses when node name is re-used #3397
+
+### Other improvements
+
+* Support `--label` in `WEAVE_DOCKER_ARGS` when starting Weave #3370,#3371
+* Add missing `--token` argument in help for `weave launch` #3226, #3379
+* Print defunct processes after smoke-tests #3362
+
+[Full list of changes](https://github.com/weaveworks/weave/milestone/75?closed=1)
+
+
+## Release 2.4.0
+
+This release introduces a support for Kubernetes Egress Network Policy (#2624, #3313)
+and adds a mechanism for preserving the client source IP address to enable
+`externalTrafficPolicy: Local` on Kubernetes (#2924, #3298).
+
+### Bug fixes
+
+* Increase the ipset list size which prevents weave-npc from crashing on older
+  kernels when more than eight Kubernetes Namespaces are used (#3289, #3305).
+* Avoid a possible livelock when reclaiming IP address space in weave-kube (#3317).
+* Ensure `xtables.lock` is mounted as a file so that kube-proxy can take the lock
+  if it has started after Weave Net (#3351, #3353).
+* Upgrade the CNI plugin symlinks only if the plugin has changed (#3337, #3345).
+
+### Other improvements
+
+* Manipulate the Kubernetes node status `NetworkUnavailable` so that Pods can be
+  scheduled on nodes when the GCE cloud provider is in use (#3249, #3307, #3332, #3334).
+* Refrain from creating a subprocess for configuring a network interface in
+  a container network namespace (#3291).
+* Protect against handling the CNI plugin request with the host namespace which
+  prevents Weave Net from misconfiguring the host network (#3206, #3346).
+* Weave Net can be run on minikube VM (#3124).
+* Add `org.opencontainers.image.*` labels to Dockerfiles to improve association
+  of the container images with git revisions (#3299).
+* Improve the error message when running `weave reset` on Kubernetes (#3319).
+
+### Build and Testing
+
+* Use `dep` instead of `git submodules` for managing external packages (#3268).
+* Fix usage of `manifest-tool` in Makefile (#3320).
+* Update Kubernetes to 1.11 for the integration tests (#3340).
+
+### External Contributors
+
+Thanks to the following contributors:
+
+* @kitt1987
+* @louismunro
+* @Nodraak
+* @stevenjohnstone
+
+[Full list of changes](https://github.com/weaveworks/weave/milestone/73?closed=1)
+
+## Release 2.3.0
+
+### Security fixes
+
+* By default, do not expose Weave "/status" and "/report" to all (0.0.0.0) when
+  running on Kubernetes #3271
+
+### Other improvements
+
+* Increase the default connection limit for Weave peers (from 30 to 100) when
+running on Kubernetes, so that more peers could directly connect #3265
+
+## Build and test
+
+* Build Weave Net with Go 1.10.1 #3273
+* Run integration tests against Kubernetes 1.10.0 #3266
+
+[Full list of changes](https://github.com/weaveworks/weave/milestone/70?closed=1)
+
+## Release 2.2.1
+
+### Bug fixes
+
+* Fix a bug in weave-npc which would allow ingress traffic to Kubernetes Pods selected
+  by a NetworkPolicy in which source and destination selectors were the same #3222,#3237
+* Fix a bug in weave-npc which would crash if a previously deleted Kubernetes Namespace
+  has been created again #3247,#3250
+
+### Other improvements
+
+* Increase the default connection limit for Weave peers (from 30 to 100), so that
+  more peers could directly connect #3234
+* When doing a rolling update of Weave Net on Kubernetes, allow each node five seconds
+  to initialize before rolling next Weave Net Pod, so that issues at startup will halt
+  the rollout and not spread across the whole cluster #3235
+* Install common CA certificates from Alpine Linux package instead of copying
+  them manually #3236
+
+### External contributors
+
+Thanks to the following contributors:
+
+* @alok87
+
+[Full list of changes](https://github.com/weaveworks/weave/milestone/71?closed=1)
+
 ## Release 2.2.0
 
 This release improves the way Weave Net configures Linux network

CONTRIBUTING.md

@@ -24,6 +24,25 @@ taking the time to read this guide.
 * Keep pull requests small so core developers can review them quickly.
 * Keep each pull request focused on a specific topic.  If you have two things to change, create two pull requests. This helps reviewers to understand the meat of your contribution.
 
+## Managing Vendored Dependencies
+
+We copy all Go dependencies into the `vendor` subdirectory and use
+[`dep`](https://github.com/golang/dep) to manage them.
+
+If you add a dependency in the code, run `dep ensure` to include it in
+the config and vendor dir. Same if you remove a dependency from the code.
+
+To update a single dependency run `dep ensure -update github.com/foo/bar`
+
+To override `dep`'s choice of library version, you can add
+[rules](https://golang.github.io/dep/docs/daily-dep.html#rule-changes-in-gopkgtoml).
+Please add a comment saying why the override is necessary and when it can be
+removed.
+
+These operations result in staged but uncommitted changes to your
+branch; you will need to commit them as normal. Execute them in the
+root of your checkout.
+
 ## Workflow
 See [WorkingOnWeave](https://github.com/weaveworks/weave/wiki/WorkingOnWeave)
 

DEPENDENCIES

@@ -1,3 +1,3 @@
 DOCKER_VERSION=17.06
-KUBERNETES_VERSION=1.9.2
+KUBERNETES_VERSION=1.11.0
 KUBERNETES_CNI_VERSION=0.6.0

Gopkg.toml

@@ -0,0 +1,29 @@
+# Pin to version before sctp was added, because latest libnetwork is
+# incompatible with latest sctp
+[[constraint]]
+  name = "github.com/docker/libnetwork"
+  revision = "119badbd"
+
+# Encourage dep not to use the docker-1.9 version
+[[constraint]]
+  name = "github.com/fsouza/go-dockerclient"
+  version = "1"
+
+# Use recent version: last tagged version is years old
+[[constraint]]
+  branch = "master"
+  name = "github.com/prometheus/client_golang"
+
+# For compatibility with fsouza/go-dockerclient
+[[constraint]]
+  branch = "master"
+  name = "github.com/docker/docker"
+
+# Most recent release tag is incompatible with docker/docker; see https://github.com/docker/distribution/issues/2503
+[[override]]
+  branch = "master"
+  name = "github.com/docker/distribution"
+
+[prune]
+  go-tests = true
+  unused-packages = true

Makefile

@@ -101,6 +101,7 @@ endif
 # The name of the user that this Makefile should produce image artifacts for. Can/should be overridden
 DOCKERHUB_USER?=weaveworks
 # The default version that's chosen when pushing the images. Can/should be overridden
+GIT_REVISION=$(shell git rev-parse HEAD)
 WEAVE_VERSION?=git-$(shell git rev-parse --short=12 HEAD)
 # Docker Store does not allow the "latest" tag.
 NET_PLUGIN_LATEST=latest_release
@@ -108,20 +109,20 @@ NET_PLUGIN_LATEST=latest_release
 # Paths to all relevant binaries that should be compiled
 WEAVER_EXE=prog/weaver/weaver
 SIGPROXY_EXE=prog/sigproxy/sigproxy
-KUBEPEERS_EXE=prog/kube-peers/kube-peers
+KUBEUTILS_EXE=prog/kube-utils/kube-utils
 WEAVENPC_EXE=prog/weave-npc/weave-npc
 WEAVEWAIT_EXE=prog/weavewait/weavewait
 WEAVEWAIT_NOOP_EXE=prog/weavewait/weavewait_noop
 WEAVEWAIT_NOMCAST_EXE=prog/weavewait/weavewait_nomcast
 WEAVEUTIL_EXE=prog/weaveutil/weaveutil
 RUNNER_EXE=tools/runner/runner
-MANIFEST_TOOL_DIR=vendor/github.com/estesp/manifest-tool
-MANIFEST_TOOL_EXE=$(MANIFEST_TOOL_DIR)/manifest-tool
+# manifest-tool needs registry credentials; we assume here that the active user has logged in
+MANIFEST_TOOL_CMD=docker run --rm -v $(HOME)/.docker:/.docker weshigbee/manifest-tool --docker-cfg=/.docker
 TEST_TLS_EXE=test/tls/tls
 NETWORKTESTER_EXE=test/images/network-tester/webserver
 
 # All binaries together in a list
-EXES=$(WEAVER_EXE) $(SIGPROXY_EXE) $(KUBEPEERS_EXE) $(WEAVENPC_EXE) $(WEAVEWAIT_EXE) $(WEAVEWAIT_NOOP_EXE) $(WEAVEWAIT_NOMCAST_EXE) $(WEAVEUTIL_EXE) $(RUNNER_EXE) $(TEST_TLS_EXE) $(MANIFEST_TOOL_EXE) $(NETWORKTESTER_EXE)
+EXES=$(WEAVER_EXE) $(SIGPROXY_EXE) $(KUBEUTILS_EXE) $(WEAVENPC_EXE) $(WEAVEWAIT_EXE) $(WEAVEWAIT_NOOP_EXE) $(WEAVEWAIT_NOMCAST_EXE) $(WEAVEUTIL_EXE) $(RUNNER_EXE) $(TEST_TLS_EXE) $(NETWORKTESTER_EXE)
 
 # These stamp files are used to mark the current state of the build; whether an image has been built or not
 BUILD_UPTODATE=.build.uptodate
@@ -179,11 +180,10 @@ $(WEAVER_EXE): api/*.go plugin/*.go plugin/*/*
 $(WEAVER_EXE):  proxy/*.go
 $(WEAVEUTIL_EXE): prog/weaveutil/*.go net/*.go plugin/net/*.go plugin/ipam/*.go db/*.go
 $(SIGPROXY_EXE): prog/sigproxy/*.go
-$(KUBEPEERS_EXE): prog/kube-peers/*.go
+$(KUBEUTILS_EXE): prog/kube-utils/*.go
 $(WEAVENPC_EXE): prog/weave-npc/*.go npc/*.go npc/*/*.go
 $(TEST_TLS_EXE): test/tls/*.go
 $(RUNNER_EXE): tools/runner/*.go
-$(MANIFEST_TOOL_EXE): $(MANIFEST_TOOL_DIR)/*.go
 $(WEAVEWAIT_NOOP_EXE): prog/weavewait/*.go
 $(WEAVEWAIT_EXE): prog/weavewait/*.go net/*.go
 $(WEAVEWAIT_NOMCAST_EXE): prog/weavewait/*.go net/*.go
@@ -196,9 +196,6 @@ ifeq ($(BUILD_IN_CONTAINER),true)
 # This make target compiles all binaries inside of the weaveworks/build container
 # It bind-mounts the source into the container and passes all important variables
 exes $(EXES) tests lint: $(BUILD_UPTODATE)
-	git submodule update --init
-# Containernetworking has another copy of vishvananda/netlink which leads to duplicate definitions
-	-@rm -r vendor/github.com/containernetworking/cni/vendor
 	@mkdir -p $(shell pwd)/.pkg
 	$(SUDO) docker run $(RM) $(RUN_FLAGS) \
 	    -v $(shell pwd):/go/src/github.com/weaveworks/weave \
@@ -220,7 +217,7 @@ else
 endif
 	$(NETGO_CHECK)
 
-$(WEAVEUTIL_EXE) $(KUBEPEERS_EXE) $(WEAVENPC_EXE) $(NETWORKTESTER_EXE):
+$(WEAVEUTIL_EXE) $(KUBEUTILS_EXE) $(WEAVENPC_EXE) $(NETWORKTESTER_EXE):
 	go build $(BUILD_FLAGS) -o $@ ./$(@D)
 	$(NETGO_CHECK)
 
@@ -234,7 +231,7 @@ $(WEAVEWAIT_NOMCAST_EXE):
 
 # These programs need a separate rule as they fail the netgo check in
 # the main build stanza due to not importing net package
-$(SIGPROXY_EXE) $(TEST_TLS_EXE) $(WEAVEWAIT_NOOP_EXE) $(RUNNER_EXE) $(MANIFEST_TOOL_EXE):
+$(SIGPROXY_EXE) $(TEST_TLS_EXE) $(WEAVEWAIT_NOOP_EXE) $(RUNNER_EXE):
 	go build $(BUILD_FLAGS) -o $@ ./$(@D)
 
 tests:
@@ -274,7 +271,7 @@ endif
 $(WEAVER_UPTODATE): prog/weaver/Dockerfile.$(DOCKERHUB_USER) $(WEAVER_EXE) weave $(WEAVEUTIL_EXE)
 	cp $(WEAVEUTIL_EXE) prog/weaver/weaveutil
 	cp weave prog/weaver/weave
-	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker build -f prog/weaver/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVER_IMAGE) prog/weaver
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker build --build-arg=revision=$(GIT_REVISION) -f prog/weaver/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVER_IMAGE) prog/weaver
 	touch $@
 
 $(WEAVEEXEC_UPTODATE): prog/weaveexec/Dockerfile.$(DOCKERHUB_USER) prog/weaveexec/symlink $(SIGPROXY_EXE) $(WEAVEWAIT_EXE) $(WEAVEWAIT_NOOP_EXE) $(WEAVEWAIT_NOMCAST_EXE) $(WEAVER_UPTODATE)
@@ -282,7 +279,7 @@ $(WEAVEEXEC_UPTODATE): prog/weaveexec/Dockerfile.$(DOCKERHUB_USER) prog/weaveexe
 	cp $(WEAVEWAIT_EXE) prog/weaveexec/weavewait
 	cp $(WEAVEWAIT_NOOP_EXE) prog/weaveexec/weavewait_noop
 	cp $(WEAVEWAIT_NOMCAST_EXE) prog/weaveexec/weavewait_nomcast
-	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker build -f prog/weaveexec/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVEEXEC_IMAGE) prog/weaveexec
+	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker build --build-arg=revision=$(GIT_REVISION) -f prog/weaveexec/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVEEXEC_IMAGE) prog/weaveexec
 	touch $@
 
 # Builds Docker plugin.
@@ -302,17 +299,17 @@ $(PLUGIN_UPTODATE): prog/net-plugin/launch.sh prog/net-plugin/config.json $(WEAV
 	$(SUDO) docker plugin create $(PLUGIN_IMAGE):$(NET_PLUGIN_LATEST) prog/net-plugin
 	touch $@
 
-$(WEAVEKUBE_UPTODATE): prog/weave-kube/Dockerfile.$(DOCKERHUB_USER) prog/weave-kube/launch.sh $(KUBEPEERS_EXE) $(WEAVER_UPTODATE)
-	cp $(KUBEPEERS_EXE) prog/weave-kube/
-	$(SUDO) docker build -f prog/weave-kube/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVEKUBE_IMAGE) prog/weave-kube
+$(WEAVEKUBE_UPTODATE): prog/weave-kube/Dockerfile.$(DOCKERHUB_USER) prog/weave-kube/launch.sh $(KUBEUTILS_EXE) $(WEAVER_UPTODATE)
+	cp $(KUBEUTILS_EXE) prog/weave-kube/
+	$(SUDO) docker build --build-arg=revision=$(GIT_REVISION) -f prog/weave-kube/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVEKUBE_IMAGE) prog/weave-kube
 	touch $@
 
 $(WEAVENPC_UPTODATE): prog/weave-npc/Dockerfile.$(DOCKERHUB_USER) $(WEAVENPC_EXE) prog/weave-npc/ulogd.conf
-	$(SUDO) docker build -f prog/weave-npc/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVENPC_IMAGE) prog/weave-npc
+	$(SUDO) docker build --build-arg=revision=$(GIT_REVISION) -f prog/weave-npc/Dockerfile.$(DOCKERHUB_USER) -t $(WEAVENPC_IMAGE) prog/weave-npc
 	touch $@
 
 $(WEAVEDB_UPTODATE): prog/weavedb/Dockerfile
-	$(SUDO) docker build -t $(WEAVEDB_IMAGE) prog/weavedb
+	$(SUDO) docker build --build-arg=revision=$(GIT_REVISION) -t $(WEAVEDB_IMAGE) prog/weavedb
 	touch $@
 
 $(NETWORKTESTER_UPTODATE): test/images/network-tester/Dockerfile $(NETWORKTESTER_EXE)
@@ -322,7 +319,7 @@ $(NETWORKTESTER_UPTODATE): test/images/network-tester/Dockerfile $(NETWORKTESTER
 $(WEAVE_EXPORT): $(IMAGES_UPTODATE) $(WEAVEDB_UPTODATE)
 	$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker save $(addsuffix :latest,$(IMAGES)) | gzip > $@
 
-tools/.git $(MANIFEST_TOOL_DIR)/.git:
+tools/.git:
 	git submodule update --init
 
 # CODE FOR PUBLISHING THE IMAGES
@@ -378,13 +375,13 @@ endif
 
 # This target pushes a manifest list; it takes one parameter, the image name.
 # The variable UPDATE_LATEST controls whether it updates the versioned tag and/or the latest tag
-$(PUSH_ML): push_ml_%: $(MANIFEST_TOOL_EXE)
+$(PUSH_ML): push_ml_%:
 ifneq ($(UPDATE_LATEST),latest-only)
-	$(MANIFEST_TOOL_EXE) push from-args --platforms $(ML_PLATFORMS) --template $(DOCKERHUB_USER)/$*-ARCH:$(WEAVE_VERSION) --target $(DOCKERHUB_USER)/$*:$(WEAVE_VERSION)
+	$(MANIFEST_TOOL_CMD) push from-args --platforms $(ML_PLATFORMS) --template $(DOCKERHUB_USER)/$*-ARCH:$(WEAVE_VERSION) --target $(DOCKERHUB_USER)/$*:$(WEAVE_VERSION)
 endif
 ifneq ($(UPDATE_LATEST),false)
 # Push the manifest list to :latest as well
-	$(MANIFEST_TOOL_EXE) push from-args --platforms $(ML_PLATFORMS) --template $(DOCKERHUB_USER)/$*-ARCH:latest --target $(DOCKERHUB_USER)/$*:latest
+	$(MANIFEST_TOOL_CMD) push from-args --platforms $(ML_PLATFORMS) --template $(DOCKERHUB_USER)/$*-ARCH:latest --target $(DOCKERHUB_USER)/$*:latest
 endif
 
 clean-work-dir:
@@ -398,7 +395,7 @@ clean-bin:
 
 clean: clean-bin clean-work-dir
 	-$(SUDO) DOCKER_HOST=$(DOCKER_HOST) docker rmi $(BUILD_IMAGE)
-	rm -rf test/tls/*.pem test/coverage.* test/coverage $(BUILD_UPTODATE) $(MANIFEST_TOOL_EXE)
+	rm -rf test/tls/*.pem test/coverage.* test/coverage $(BUILD_UPTODATE)
 
 build:
 	$(SUDO) go clean -i net

bin/release

@@ -92,6 +92,7 @@ build() {
     sed -i.bak -e "s/:latest/:$VERSION/" -e "/imagePullPolicy: Always/d" ./prog/weave-kube/weave-daemonset.yaml
     sed -i.bak -e "s/:latest/:$VERSION/" -e "/imagePullPolicy: Always/d" ./prog/weave-kube/weave-daemonset-k8s-1.6.yaml
     sed -i.bak -e "s/:latest/:$VERSION/" -e "/imagePullPolicy: Always/d" ./prog/weave-kube/weave-daemonset-k8s-1.7.yaml
+    sed -i.bak -e "s/:latest/:$VERSION/" -e "/imagePullPolicy: Always/d" ./prog/weave-kube/weave-daemonset-k8s-1.8.yaml
     make SUDO=$SUDO WEAVE_VERSION=$VERSION DOCKERHUB_USER=$DOCKERHUB_USER
 
     if make tests; then
@@ -181,6 +182,13 @@ draft() {
         --name "weave-daemonset-k8s-1.7.yaml" \
         --file "./prog/weave-kube/weave-daemonset-k8s-1.7.yaml"
 
+    github-release upload \
+        --user $GITHUB_USER \
+        --repo weave \
+        --tag $LATEST_TAG \
+        --name "weave-daemonset-k8s-1.8.yaml" \
+        --file "./prog/weave-kube/weave-daemonset-k8s-1.8.yaml"
+    
     echo "** Draft $TYPE $RELEASE_NAME $VERSION created at"
     echo -e "\thttps://github.com/$GITHUB_USER/weave/releases/$LATEST_TAG"
 }
@@ -289,6 +297,13 @@ publish() {
             --name "weave-daemonset-k8s-1.7.yaml" \
             --file "./prog/weave-kube/weave-daemonset-k8s-1.7.yaml"
 
+        github-release upload \
+            --user $GITHUB_USER \
+            --repo weave \
+            --tag latest_release \
+            --name "weave-daemonset-k8s-1.8.yaml" \
+            --file "./prog/weave-kube/weave-daemonset-k8s-1.8.yaml"
+
         echo "** Release $RELEASE_NAME $VERSION published at"
         echo -e "\thttps://github.com/$GITHUB_USER/weave/releases/$LATEST_TAG"
         echo -e "\thttps://github.com/$GITHUB_USER/weave/releases/latest_release"

build/Dockerfile

@@ -1,4 +1,4 @@
-FROM golang:1.8.1
+FROM golang:1.10.1-stretch
 
 # Support Raspberry Pi 2 and newer
 ENV GOARM 7

circle.yml

@@ -44,7 +44,7 @@ test:
         parallel: true
     - "$SRCDIR/bin/circle-test-smoke":
         parallel: true
-        timeout: 300
+        timeout: 360
   post:
     - "$SRCDIR/bin/circle-test-teardown":
         parallel: true

common/docker/client.go

@@ -231,7 +231,7 @@ func (c *Client) EnsureNetwork(networkName, driver, subnet string, options map[s
 			Name:           networkName,
 			CheckDuplicate: true,
 			Driver:         driver,
-			IPAM: docker.IPAMOptions{
+			IPAM: &docker.IPAMOptions{
 				Driver: driver,
 				Config: []docker.IPAMConfig{{Subnet: subnet}},
 			},

common/logging.go

@@ -6,7 +6,7 @@ import (
 	"log"
 	"strings"
 
-	"github.com/Sirupsen/logrus"
+	"github.com/sirupsen/logrus"
 )
 
 type textFormatter struct {

docs/k8s-src-ip.md

@@ -0,0 +1,33 @@
+# Source IP addr of packet sent to k8s Service
+
+## Setup
+
+* Weave Net: v2.2.1.
+* Kubernetes: v1.10.0.
+* kube-proxy: iptables mode (default kubeadm options).
+* Hosts: `h1` and `h2`.
+
+## Source IP
+
+\# | Src         | Target            | Dst    | Src IP           | `-j WEAVE-NPC`
+---|-------------|-------------------|--------|------------------|--------------
+1  | Pod_h1      | ip(Pod_h2)        | Pod_h2 | ip(Pod_h1)       | OK
+2  | h1          | ClusterIP         | Pod_h1 | ip(weave_h1)     | **NOK**
+3  | h1          | ClusterIP         | Pod_h2 | ip(weave_h1)     | OK
+4  | Pod_h1      | ClusterIP         | Pod_h1 | ip(weave_h1)     | **NOK**
+5  | Pod_h1      | ClusterIP         | Pod_h2 | ip(weave_h1)     | OK
+6  | h1          | ip(h1):NodePort   | Pod_h1 | ip(weave_h1)     | **NOK**
+7  | h1          | ip(h1):NodePort   | Pod_h2 | ip(weave_h1)     | OK
+8  | h1          | ip(h2):NodePort   | Pod_h1 | ip(weave_h2)     | OK
+9  | h1          | ip(h2):NodePort   | Pod_h2 | ip(weave_h2)     | ??? Can't reproduce
+10 | Pod_h1      | ip(h1):NodePort   | Pod_h1 | ip(weave_h1)     | **NOK**
+11 | Pod_h1      | ip(h1):NodePort   | Pod_h2 | ip(weave_h1)     | OK
+12 | Pod_h1      | ip(h2):NodePort   | Pod_h1 | ip(weave_h2)     | OK
+13 | Pod_h1      | ip(h2):NodePort   | Pod_h2 | ip(weave_h2)     | OK
+
+Remarks:
+
+* **Src IP** is of a packet which is captured on the weave bridge.
+* **-j WEAVE-NPC** - whether a packet enters the `filter/WEAVE-NPC` iptables chain (OK = NetworkPolicy is enforced as required).
+* **Pod_h1** - a Pod running on the `h1` host.
+* **ip(weave_h1)** - IP addr of the weave bridge on the `h1` host.

docs/release-process.md

@@ -127,6 +127,44 @@ Finally, for **Mainline** releases only:
 
 * Images tagged `latest` are updated on DockerHub
 
+Now, you can validate whether images were published for all platforms:
+
+    # Required platforms for each image:
+    grep '^ML_PLATFORMS=' Makefile
+
+    # Published platforms per image:
+    for img in $(grep '^PUBLISH=' Makefile); do
+        img="weaveworks/$(echo $img | cut -d_ -f2):${TAG#v}"
+        platforms=$(vendor/github.com/estesp/manifest-tool/manifest-tool \
+                inspect --raw "$img" | \
+                jq '.[].Platform | .os + "/" +.architecture' | \
+                tr '\n' ' ')
+        echo "$img: $platforms"
+    done
+
+
+### Docker Store
+
+Weave Net is [available in Docker Store](https://store.docker.com/plugins/weave-net-plugin) and should also be released there.
+
+* Go to https://store.docker.com/ and log in.
+* Go to "[_Publish_](https://store.docker.com/publisher/center)".
+* Under "_My Products_", select "_owners (weaveworks)_". Weave Net should be listed among our products.
+* Click "_Actions_" > "_Edit Product_".
+* Go to "_Plans & Pricing_" > "_Free Tier_".
+* Under "_Source Repositories & Tags_", click on "_Add Source_", select "_net-plugin_" under "_Repository_" and the version to release under "_Tag_".
+* Under "_Resources_" > "_Installation Instructions_", update the lines containing `export NET_VERSION=<version>`.
+* Click "_Submit For Review_". You should see a message like: "_Your product has been submitted for approval. [...] We'll be in touch with next steps soon!_".
+* You should receive an email saying:
+
+> This email confirms that we received your submission on \<date\> of Weave Net to the Docker Store.
+>
+> We're reviewing your submission to ensure that it meets our [security guidelines](https://success.docker.com/Store#Security_and_Audit_Policies) and complies with our [best practices](https://success.docker.com/Store#Create_Great_Content). Don't worry! We'll let you know if there's anything you need to change before we publish your submission. You should hear back from us within the next 14 days.
+> 
+> Thanks for submitting your content to the Docker Store!
+
+* Hope Docker eventually performs the release or contacts us. If nothing happens within 14 days, contact their support team: publisher-support@docker.com.
+
 ### Finish up
 
 * If not on master, merge branch into master and push to GitHub.

ipam/allocator.go

@@ -74,6 +74,7 @@ type Allocator struct {
 	isKnownPeer       func(mesh.PeerName) bool
 	quorum            func() uint
 	now               func() time.Time
+	tracker           tracker.LocalRangeTracker
 }
 
 // PreClaims are IP addresses discovered before we could initialize IPAM
@@ -130,6 +131,7 @@ func NewAllocator(config Config) *Allocator {
 		quorum:      config.Quorum,
 		dead:        make(map[string]time.Time),
 		now:         time.Now,
+		tracker:     config.Tracker,
 	}
 
 	alloc.pendingClaims = make([]operation, len(config.PreClaims))

ipam/http.go

@@ -88,7 +88,7 @@ func (alloc *Allocator) handleHTTPClaim(dockerCli *docker.Client, w http.Respons
 }
 
 // HandleHTTP wires up ipams HTTP endpoints to the provided mux.
-func (alloc *Allocator) HandleHTTP(router *mux.Router, defaultSubnet address.CIDR, tracker string, dockerCli *docker.Client) {
+func (alloc *Allocator) HandleHTTP(router *mux.Router, defaultSubnet address.CIDR, dockerCli *docker.Client) {
 	router.Methods("GET").Path("/ipinfo/defaultsubnet").HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		fmt.Fprintf(w, "%s", defaultSubnet)
 	})
@@ -199,6 +199,6 @@ func (alloc *Allocator) HandleHTTP(router *mux.Router, defaultSubnet address.CID
 	})
 
 	router.Methods("GET").Path("/ipinfo/tracker").HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		fmt.Fprintf(w, tracker)
+		fmt.Fprintf(w, alloc.tracker.String())
 	})
 }

ipam/http_test.go

@@ -43,7 +43,7 @@ func listenHTTP(alloc *Allocator, subnet address.CIDR) int {
 	router.HandleFunc("/status", func(w http.ResponseWriter, r *http.Request) {
 		io.WriteString(w, fmt.Sprintln(alloc))
 	})
-	alloc.HandleHTTP(router, subnet, "", nil)
+	alloc.HandleHTTP(router, subnet, nil)
 
 	httpListener, err := net.Listen("tcp", ":0")
 	if err != nil {
@@ -132,7 +132,7 @@ func TestBadHttp(t *testing.T) {
 	// Verb that's not handled
 	resp, err := doHTTP("HEAD", fmt.Sprintf("http://localhost:%d/ip/%s/%s", port, containerID, testAddr1))
 	require.NoError(t, err)
-	require.Equal(t, http.StatusNotFound, resp.StatusCode, "http response")
+	require.Equal(t, http.StatusMethodNotAllowed, resp.StatusCode, "http response")
 	// Mis-spelled URL
 	resp, err = doHTTP("POST", fmt.Sprintf("http://localhost:%d/xip/%s/", port, containerID))
 	require.NoError(t, err)

ipam/testutils_test.go

@@ -117,7 +117,7 @@ func CheckAllExpectedMessagesSent(allocs ...*Allocator) {
 		m := alloc.gossip.(*mockGossipComms)
 		m.RLock()
 		if len(m.messages) > 0 {
-			require.FailNow(m, fmt.Sprintf("%s: Gossip message(s) not sent as expected: \n%x", m.name, m.messages))
+			require.FailNow(m, fmt.Sprintf("%s: Gossip message(s) not sent as expected: \n%v", m.name, m.messages))
 		}
 		m.RUnlock()
 	}

ipam/tracker/awsvpc.go

@@ -84,7 +84,7 @@ func NewAWSVPCTracker(bridgeName string) (*AWSVPCTracker, error) {
 func (t *AWSVPCTracker) HandleUpdate(prevRanges, currRanges []address.Range, local bool) error {
 	t.debugf("replacing %q by %q; local(%t)", prevRanges, currRanges, local)
 
-	prev, curr := removeCommon(address.NewCIDRs(merge(prevRanges)), address.NewCIDRs(merge(currRanges)))
+	prev, curr := RemoveCommon(address.NewCIDRs(Merge(prevRanges)), address.NewCIDRs(Merge(currRanges)))
 
 	// It might make sense to do the removal first and then add entries
 	// because of the 50 routes limit. However, in such case a container might
@@ -123,6 +123,10 @@ func (t *AWSVPCTracker) HandleUpdate(prevRanges, currRanges []address.Range, loc
 	return nil
 }
 
+func (t *AWSVPCTracker) String() string {
+	return "awsvpc"
+}
+
 func (t *AWSVPCTracker) createVPCRoute(cidr string) (*ec2.CreateRouteOutput, error) {
 	route := &ec2.CreateRouteInput{
 		RouteTableId:         &t.routeTableID,
@@ -232,47 +236,6 @@ func (t *AWSVPCTracker) infof(fmt string, args ...interface{}) {
 
 // Helpers
 
-// merge merges adjacent range entries.
-// The given slice has to be sorted in increasing order.
-func merge(r []address.Range) []address.Range {
-	var merged []address.Range
-
-	for i := range r {
-		if prev := len(merged) - 1; prev >= 0 && merged[prev].End == r[i].Start {
-			merged[prev].End = r[i].End
-		} else {
-			merged = append(merged, r[i])
-		}
-	}
-
-	return merged
-}
-
-// removeCommon filters out CIDR ranges which are contained in both a and b slices.
-// Both slices have to be sorted in increasing order.
-func removeCommon(a, b []address.CIDR) (newA, newB []address.CIDR) {
-	i, j := 0, 0
-
-	for i < len(a) && j < len(b) {
-		switch {
-		case a[i].Start() < b[j].Start() || a[i].End() < b[j].End():
-			newA = append(newA, a[i])
-			i++
-		case a[i].Start() > b[j].Start() || a[i].End() > b[j].End():
-			newB = append(newB, b[j])
-			j++
-		default:
-			i++
-			j++
-		}
-
-	}
-	newA = append(newA, a[i:]...)
-	newB = append(newB, b[j:]...)
-
-	return
-}
-
 func parseCIDR(cidr string) (*net.IPNet, error) {
 	ip, ipnet, err := net.ParseCIDR(cidr)
 	if err != nil {

ipam/tracker/helpers.go

@@ -0,0 +1,46 @@
+package tracker
+
+import (
+	"github.com/weaveworks/weave/net/address"
+)
+
+// Merge merges adjacent range entries.
+// The given slice has to be sorted in increasing order.
+func Merge(r []address.Range) []address.Range {
+	var merged []address.Range
+
+	for i := range r {
+		if prev := len(merged) - 1; prev >= 0 && merged[prev].End == r[i].Start {
+			merged[prev].End = r[i].End
+		} else {
+			merged = append(merged, r[i])
+		}
+	}
+
+	return merged
+}
+
+// RemoveCommon filters out CIDR ranges which are contained in both a and b slices.
+// Both slices have to be sorted in increasing order.
+func RemoveCommon(a, b []address.CIDR) (newA, newB []address.CIDR) {
+	i, j := 0, 0
+
+	for i < len(a) && j < len(b) {
+		switch {
+		case a[i].Start() < b[j].Start() || a[i].End() < b[j].End():
+			newA = append(newA, a[i])
+			i++
+		case a[i].Start() > b[j].Start() || a[i].End() > b[j].End():
+			newB = append(newB, b[j])
+			j++
+		default:
+			i++
+			j++
+		}
+
+	}
+	newA = append(newA, a[i:]...)
+	newB = append(newB, b[j:]...)
+
+	return
+}

ipam/tracker/helpers_test.go

@@ -19,7 +19,7 @@ var (
 func TestRemoveCommon(t *testing.T) {
 	a := []address.CIDR{r0to127, r1dot0to255}
 	b := []address.CIDR{r1dot0to255, r2dot0to255}
-	newA, newB := removeCommon(a, b)
+	newA, newB := RemoveCommon(a, b)
 	require.Equal(t, []address.CIDR{r0to127}, newA)
 	require.Equal(t, []address.CIDR{r2dot0to255}, newB)
 }
@@ -30,7 +30,7 @@ func TestMerge(t *testing.T) {
 		r128to255.Range(),
 		r2dot0to255.Range(),
 	}
-	require.Equal(t, []address.Range{r0to255.Range(), r2dot0to255.Range()}, merge(ranges))
+	require.Equal(t, []address.Range{r0to255.Range(), r2dot0to255.Range()}, Merge(ranges))
 }
 
 // Helper

ipam/tracker/tracker.go

@@ -17,4 +17,7 @@ type LocalRangeTracker interface {
 	// The local parameter indicates whether the ranges belong to the peer
 	// by which the method is called.
 	HandleUpdate(prevRanges, currRanges []address.Range, local bool) error
+
+	// String returns the tracker name
+	String() string
 }

net/arp.go

@@ -10,7 +10,7 @@ import (
 // Configure the ARP cache parameters for the given interface.  This
 // makes containers react more quickly to a change in the MAC address
 // associated with an IP address.
-func ConfigureARPCache(procPath, name string) error {
+func configureARPCache(procPath, name string) error {
 	if err := sysctl(procPath, fmt.Sprintf("net/ipv4/neigh/%s/base_reachable_time", name), "5"); err != nil {
 		return err
 	}

net/bridge.go

@@ -6,12 +6,18 @@ import (
 	"net"
 	"path/filepath"
 
-	"github.com/Sirupsen/logrus"
 	"github.com/coreos/go-iptables/iptables"
 	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
 	"github.com/vishvananda/netlink"
+	"k8s.io/apimachinery/pkg/types"
 
+	"github.com/weaveworks/weave/common"
 	"github.com/weaveworks/weave/common/odp"
+	"github.com/weaveworks/weave/ipam/tracker"
+	"github.com/weaveworks/weave/net/address"
+	"github.com/weaveworks/weave/net/ipset"
+	"github.com/weaveworks/weave/npc"
 )
 
 /* This code implements three possible configurations to connect
@@ -47,11 +53,12 @@ datapath of old kernel versions (https://github.com/weaveworks/weave/issues/1577
 */
 
 const (
-	WeaveBridgeName = "weave"
-	DatapathName    = "datapath"
-	DatapathIfName  = "vethwe-datapath"
-	BridgeIfName    = "vethwe-bridge"
-	PcapIfName      = "vethwe-pcap"
+	WeaveBridgeName  = "weave"
+	DatapathName     = "datapath"
+	DatapathIfName   = "vethwe-datapath"
+	BridgeIfName     = "vethwe-bridge"
+	PcapIfName       = "vethwe-pcap"
+	NoMasqLocalIpset = ipset.Name("weaver-no-masq-local")
 )
 
 type Bridge interface {
@@ -213,6 +220,7 @@ type BridgeConfig struct {
 	MTU              int
 	Mac              string
 	Port             int
+	NoMasqLocal      bool
 }
 
 func (config *BridgeConfig) configuredBridgeType() Bridge {
@@ -226,7 +234,7 @@ func (config *BridgeConfig) configuredBridgeType() Bridge {
 	}
 }
 
-func EnsureBridge(procPath string, config *BridgeConfig, log *logrus.Logger) (Bridge, error) {
+func EnsureBridge(procPath string, config *BridgeConfig, log *logrus.Logger, ips ipset.Interface) (Bridge, error) {
 	existingBridgeType, err := ExistingBridgeType(config.WeaveBridgeName, config.DatapathName)
 	if err != nil {
 		return nil, err
@@ -252,7 +260,7 @@ func EnsureBridge(procPath string, config *BridgeConfig, log *logrus.Logger) (Br
 		break
 	}
 
-	if err := configureIPTables(config); err != nil {
+	if err := configureIPTables(config, ips); err != nil {
 		return bridgeType, errors.Wrap(err, "configuring iptables")
 	}
 
@@ -275,10 +283,16 @@ func EnsureBridge(procPath string, config *BridgeConfig, log *logrus.Logger) (Br
 		return bridgeType, err
 	}
 
-	if err := ConfigureARPCache(procPath, config.WeaveBridgeName); err != nil {
+	if err := configureARPCache(procPath, config.WeaveBridgeName); err != nil {
 		return bridgeType, errors.Wrapf(err, "configuring ARP cache on bridge %q", config.WeaveBridgeName)
 	}
 
+	// NB: No concurrent call to Expose is possible, as EnsureBridge is called
+	// before any service has been started.
+	if err := reexpose(config, log); err != nil {
+		return bridgeType, err
+	}
+
 	return bridgeType, nil
 }
 
@@ -404,7 +418,7 @@ func (f fastdpImpl) attach(veth *netlink.Veth) error {
 	return odp.AddDatapathInterfaceIfNotExist(f.datapathName, veth.Attrs().Name)
 }
 
-func configureIPTables(config *BridgeConfig) error {
+func configureIPTables(config *BridgeConfig, ips ipset.Interface) error {
 	ipt, err := iptables.New()
 	if err != nil {
 		return errors.Wrap(err, "creating iptables object")
@@ -444,14 +458,27 @@ func configureIPTables(config *BridgeConfig) error {
 	}
 
 	if config.NPC {
-		// Steer traffic via the NPC
-		_ = ipt.NewChain("filter", "WEAVE-NPC") // ignore error because it might already exist
-		// If WEAVE-NPC chain doesn't exist then creating a rule in the chain will fail
+		// Steer traffic via the NPC.
+
+		if err = ensureChains(ipt, "filter", npc.MainChain, npc.EgressChain); err != nil {
+			return err
+		}
+
+		// Steer egress traffic destined to local node.
+		if err = ipt.AppendUnique("filter", "INPUT", "-i", config.WeaveBridgeName, "-j", npc.EgressChain); err != nil {
+			return err
+		}
 		fwdRules = append(fwdRules,
 			[][]string{
+				// Might include ingress traffic which is fine as long as we do not
+				// ACCEPT in WEAVE-NPC-EGRESS chain
+				{"-i", config.WeaveBridgeName,
+					"-m", "comment", "--comment", "NOTE: this must go before '-j KUBE-FORWARD'",
+					"-j", npc.EgressChain},
+				// The following rules are for ingress NPC processing
 				{"-o", config.WeaveBridgeName,
 					"-m", "comment", "--comment", "NOTE: this must go before '-j KUBE-FORWARD'",
-					"-j", "WEAVE-NPC"},
+					"-j", npc.MainChain},
 				{"-o", config.WeaveBridgeName, "-m", "state", "--state", "NEW", "-j", "NFLOG", "--nflog-group", "86"},
 				{"-o", config.WeaveBridgeName, "-j", "DROP"},
 			}...)
@@ -462,8 +489,11 @@ func configureIPTables(config *BridgeConfig) error {
 	}
 
 	if !config.NPC {
-		// Create a chain for allowing ingress traffic when the bridge is exposed
-		_ = ipt.NewChain("filter", "WEAVE-EXPOSE")
+		// Create/Flush a chain for allowing ingress traffic when the bridge is exposed
+		if err := ipt.ClearChain("filter", "WEAVE-EXPOSE"); err != nil {
+			return errors.Wrap(err, "failed to clear/create filter/WEAVE-EXPOSE chain")
+		}
+
 		fwdRules = append(fwdRules, []string{"-o", config.WeaveBridgeName, "-j", "WEAVE-EXPOSE"})
 	}
 
@@ -476,13 +506,74 @@ func configureIPTables(config *BridgeConfig) error {
 		return err
 	}
 
-	// create a chain for masquerading
-	//
-	// NB: we do not clear the chain to preserve existing rules
-	// inserted by "weave expose".
-	_ = ipt.NewChain("nat", "WEAVE")
+	// Create a chain for masquerading
+	if err := ipt.ClearChain("nat", "WEAVE"); err != nil {
+		return errors.Wrap(err, "failed to clear/create nat/WEAVE chain")
+	}
+	if err := ipt.AppendUnique("nat", "POSTROUTING", "-j", "WEAVE"); err != nil {
+		return err
+	}
 
-	return ipt.AppendUnique("nat", "POSTROUTING", "-j", "WEAVE")
+	// For the cases where the weave bridge is the default gateway for
+	// containers (e.g. Kubernetes): create the ipset to store CIDRs allocated
+	// by IPAM for local containers. In the case of Kubernetes, external traffic
+	// sent to these CIDRs avoids SNAT'ing so that NodePort with
+	// `"externalTrafficPolicy":"Local"` would receive packets with correct
+	// src IP addr.
+	if config.NoMasqLocal {
+		ips := ipset.New(common.LogLogger(), 0)
+		_ = ips.Destroy(NoMasqLocalIpset)
+		if err := ips.Create(NoMasqLocalIpset, ipset.HashNet); err != nil {
+			return err
+		}
+		if err := ipt.Insert("nat", "WEAVE", 1,
+			"-m", "set", "--match-set", string(NoMasqLocalIpset), "dst",
+			"-m", "comment", "--comment", "Prevent SNAT to locally running containers",
+			"-j", "RETURN"); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+type NoMasqLocalTracker struct {
+	ips   ipset.Interface
+	owner types.UID
+}
+
+func NewNoMasqLocalTracker(ips ipset.Interface) *NoMasqLocalTracker {
+	return &NoMasqLocalTracker{
+		ips:   ips,
+		owner: types.UID(0), // dummy ipset owner
+	}
+}
+
+func (t *NoMasqLocalTracker) String() string {
+	return "no-masq-local"
+}
+
+func (t *NoMasqLocalTracker) HandleUpdate(prevRanges, currRanges []address.Range, local bool) error {
+	if !local {
+		return nil
+	}
+
+	prev, curr := tracker.RemoveCommon(
+		address.NewCIDRs(tracker.Merge(prevRanges)),
+		address.NewCIDRs(tracker.Merge(currRanges)))
+
+	for _, cidr := range curr {
+		if err := t.ips.AddEntry(t.owner, NoMasqLocalIpset, cidr.String(), ""); err != nil {
+			return err
+		}
+	}
+	for _, cidr := range prev {
+		if err := t.ips.DelEntry(t.owner, NoMasqLocalIpset, cidr.String()); err != nil {
+			return err
+		}
+	}
+
+	return nil
 }
 
 func linkSetUpByName(linkName string) error {
@@ -493,37 +584,26 @@ func linkSetUpByName(linkName string) error {
 	return netlink.LinkSetUp(link)
 }
 
-// ensureRulesAtTop ensures the presence of given iptables rules.
-//
-// If any rule from the list is missing, the function deletes all given
-// rules and re-inserts them at the top of the chain to ensure the order of the rules.
-func ensureRulesAtTop(table, chain string, rulespecs [][]string, ipt *iptables.IPTables) error {
-	allFound := true
-
-	for _, rs := range rulespecs {
-		found, err := ipt.Exists(table, chain, rs...)
-		if err != nil {
-			return errors.Wrapf(err, "ipt.Exists(%s, %s, %s)", table, chain, rs)
-		}
-		if !found {
-			allFound = false
-			break
-		}
+func reexpose(config *BridgeConfig, log *logrus.Logger) error {
+	// Get existing IP addrs of the weave bridge.
+	// If the bridge hasn't been exposed, then this functions does nothing.
+	//
+	// Ideally, we should consult IPAM for IP addrs allocated to "weave:expose",
+	// but we don't want to introduce dependency on IPAM, as weave should be able
+	// to run w/o IPAM.
+	link, err := netlink.LinkByName(config.WeaveBridgeName)
+	if err != nil {
+		return errors.Wrapf(err, "cannot find bridge %q", config.WeaveBridgeName)
+	}
+	addrs, err := netlink.AddrList(link, netlink.FAMILY_V4)
+	if err != nil {
+		return errors.Wrapf(err, "cannot list IPv4 addrs of bridge %q", config.WeaveBridgeName)
 	}
 
-	// All rules exist, do nothing.
-	if allFound {
-		return nil
-	}
-
-	for pos, rs := range rulespecs {
-		// If any is missing, then delete all, as we need to preserve the order of
-		// given rules. Ignore errors, as rule might not exist.
-		if !allFound {
-			ipt.Delete(table, chain, rs...)
-		}
-		if err := ipt.Insert(table, chain, pos+1, rs...); err != nil {
-			return errors.Wrapf(err, "ipt.Append(%s, %s, %s)", table, chain, rs)
+	for _, addr := range addrs {
+		log.Infof("Re-exposing %s on bridge %q", addr.IPNet, config.WeaveBridgeName)
+		if err := Expose(config.WeaveBridgeName, addr.IPNet, config.AWSVPC, config.NPC); err != nil {
+			return errors.Wrapf(err, "unable to re-expose %s on bridge: %q", addr.IPNet, config.WeaveBridgeName)
 		}
 	}
 

net/ipsec/ipsec.go

@@ -12,9 +12,9 @@ import (
 	"sync"
 	"syscall"
 
-	"github.com/Sirupsen/logrus"
 	"github.com/coreos/go-iptables/iptables"
 	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
 	"github.com/vishvananda/netlink"
 	"golang.org/x/crypto/hkdf"
 

net/ipset/ipset.go

@@ -1,6 +1,9 @@
 package ipset
 
 import (
+	"crypto/rand"
+	"encoding/hex"
+	"fmt"
 	"log"
 	"os/exec"
 	"strings"
@@ -16,6 +19,7 @@ type Type string
 const (
 	ListSet = Type("list:set")
 	HashIP  = Type("hash:ip")
+	HashNet = Type("hash:net")
 )
 
 type Interface interface {
@@ -40,6 +44,7 @@ type entryKey struct {
 type ipset struct {
 	*log.Logger
 	enableComments bool
+	maxListSize    int
 	// List of users per ipset entry. User is either a namespace or a pod.
 	// There might be multiple users for the same ipset & entry pair because
 	// events from k8s API server might be out of order causing duplicate IPs:
@@ -47,15 +52,23 @@ type ipset struct {
 	users map[entryKey]map[types.UID]struct{}
 }
 
-func New(logger *log.Logger) Interface {
+func New(logger *log.Logger, maxListSize int) Interface {
 	ips := &ipset{
 		Logger:         logger,
 		enableComments: true,
+		maxListSize:    maxListSize,
 		users:          make(map[entryKey]map[types.UID]struct{}),
 	}
 
 	// Check for comment support
-	testIpsetName := Name("weave-test-comment")
+
+	// To prevent from a race when more than one process check for the support
+	// we append a random nonce to the test ipset name. The final name is
+	// shorter than 31 chars (max ipset name).
+	nonce := make([]byte, 4)
+	rand.Read(nonce)
+	testIpsetName := Name("weave-test-comment" + hex.EncodeToString(nonce))
+
 	// Clear it out if it already exists
 	_ = ips.Destroy(testIpsetName)
 	// Test for comment support
@@ -71,6 +84,9 @@ func New(logger *log.Logger) Interface {
 
 func (i *ipset) Create(ipsetName Name, ipsetType Type) error {
 	args := []string{"create", string(ipsetName), string(ipsetType)}
+	if ipsetType == ListSet && i.maxListSize > 0 {
+		args = append(args, "size", fmt.Sprintf("%d", i.maxListSize))
+	}
 	if i.enableComments {
 		args = append(args, "comment")
 	}

net/iptables.go

@@ -0,0 +1,106 @@
+package net
+
+import (
+	"strings"
+
+	"github.com/coreos/go-iptables/iptables"
+	"github.com/pkg/errors"
+)
+
+// AddChainWithRules creates a chain and appends given rules to it.
+//
+// If the chain exists, but its rules are not the same as the given ones, the
+// function will flush the chain and then will append the rules.
+func AddChainWithRules(ipt *iptables.IPTables, table, chain string, rulespecs [][]string) error {
+	if err := ensureChains(ipt, table, chain); err != nil {
+		return err
+	}
+
+	currRuleSpecs, err := ipt.List(table, chain)
+	if err != nil {
+		return errors.Wrapf(err, "iptables -S. table: %q, chain: %q", table, chain)
+	}
+
+	// First returned rule is "-N $(chain)", so ignore it
+	currRules := strings.Join(currRuleSpecs[1:], "\n")
+	rules := make([]string, 0)
+	for _, r := range rulespecs {
+		rules = append(rules, strings.Join(r, " "))
+	}
+	reqRules := strings.Join(rules, "\n")
+
+	if currRules == reqRules {
+		return nil
+	}
+
+	if err := ipt.ClearChain(table, chain); err != nil {
+		return err
+	}
+
+	for _, r := range rulespecs {
+		if err := ipt.Append(table, chain, r...); err != nil {
+			return errors.Wrapf(err, "iptables -A. table: %q, chain: %q, rule: %s", table, chain, r)
+		}
+	}
+
+	return nil
+}
+
+// ensureChains creates given chains if they do not exist.
+func ensureChains(ipt *iptables.IPTables, table string, chains ...string) error {
+	existingChains, err := ipt.ListChains(table)
+	if err != nil {
+		return errors.Wrapf(err, "ipt.ListChains(%s)", table)
+	}
+	chainMap := make(map[string]struct{})
+	for _, c := range existingChains {
+		chainMap[c] = struct{}{}
+	}
+
+	for _, c := range chains {
+		if _, found := chainMap[c]; !found {
+			if err := ipt.NewChain(table, c); err != nil {
+				return errors.Wrapf(err, "ipt.NewChain(%s, %s)", table, c)
+			}
+		}
+	}
+
+	return nil
+}
+
+// ensureRulesAtTop ensures the presence of given iptables rules.
+//
+// If any rule from the list is missing, the function deletes all given
+// rules and re-inserts them at the top of the chain to ensure the order of the rules.
+func ensureRulesAtTop(table, chain string, rulespecs [][]string, ipt *iptables.IPTables) error {
+	allFound := true
+
+	for _, rs := range rulespecs {
+		found, err := ipt.Exists(table, chain, rs...)
+		if err != nil {
+			return errors.Wrapf(err, "ipt.Exists(%s, %s, %s)", table, chain, rs)
+		}
+		if !found {
+			allFound = false
+			break
+		}
+	}
+
+	// All rules exist, do nothing.
+	if allFound {
+		return nil
+	}
+
+	for pos, rs := range rulespecs {
+		// If any is missing, then delete all, as we need to preserve the order of
+		// given rules. Ignore errors, as rule might not exist.
+		if !allFound {
+			ipt.Delete(table, chain, rs...)
+		}
+		if err := ipt.Insert(table, chain, pos+1, rs...); err != nil {
+			return errors.Wrapf(err, "ipt.Append(%s, %s, %s)", table, chain, rs)
+		}
+	}
+
+	return nil
+}

net/netdev.go

@@ -1,11 +1,9 @@
 package net
 
 import (
-	"encoding/json"
 	"fmt"
 	"net"
 	"os"
-	"strconv"
 
 	"github.com/vishvananda/netlink"
 	"github.com/vishvananda/netns"
@@ -17,7 +15,7 @@ type Dev struct {
 	CIDRs []*net.IPNet     `json:"CIDRs,omitempty"`
 }
 
-func LinkToNetDev(link netlink.Link) (Dev, error) {
+func linkToNetDev(link netlink.Link) (Dev, error) {
 	addrs, err := netlink.AddrList(link, netlink.FAMILY_V4)
 	if err != nil {
 		return Dev{}, err
@@ -90,16 +88,30 @@ func GetNetDevsByVethPeerIds(processID int, peerIDs []int) ([]Dev, error) {
 		return nil, nil
 	}
 
+	// convert list of peerIDs into a map for faster lookup
+	indexes := make(map[int]struct{})
+	for _, id := range peerIDs {
+		indexes[id] = struct{}{}
+	}
+
 	var netdevs []Dev
-	peersStr := make([]string, len(peerIDs))
-	for i, id := range peerIDs {
-		peersStr[i] = strconv.Itoa(id)
-	}
-	nds, err := WithNetNSByPid(processID, "list-netdevs", peersStr...)
-	if err != nil {
-		return nil, fmt.Errorf("list-netdevs failed: %s", err)
-	}
-	err = json.Unmarshal(nds, &netdevs)
+	err = WithNetNS(netnsContainer, func() error {
+		links, err := netlink.LinkList()
+		if err != nil {
+			return err
+		}
+
+		for _, link := range links {
+			if _, found := indexes[link.Attrs().Index]; found {
+				netdev, err := linkToNetDev(link)
+				if err != nil {
+					return err
+				}
+				netdevs = append(netdevs, netdev)
+			}
+		}
+		return nil
+	})
 
 	return netdevs, err
 }
@@ -112,5 +124,5 @@ func GetBridgeNetDev(bridgeName string) (Dev, error) {
 		return Dev{}, err
 	}
 
-	return LinkToNetDev(link)
+	return linkToNetDev(link)
 }

net/netns.go

@@ -1,10 +1,10 @@
+// +build go1.10
+
 package net
 
 import (
-	"bytes"
 	"errors"
 	"fmt"
-	"os/exec"
 	"path/filepath"
 	"runtime"
 
@@ -14,19 +14,7 @@ import (
 
 var ErrLinkNotFound = errors.New("Link not found")
 
-// NB: The following function is unsafe, because:
-//     - It changes a network namespace (netns) of an OS thread which runs
-//       the function. During execution, the Go runtime might clone a new OS thread
-//       for scheduling other go-routines, thus they might end up running in
-//       a "wrong" netns.
-//     - runtime.LockOSThread does not guarantee that a spawned go-routine on
-//       the locked thread will be run by it. Thus, the work function is
-//       not allowed to spawn any go-routine which is dependent on the given netns.
-
-//     Please see https://github.com/weaveworks/weave/issues/2388#issuecomment-228365069
-//     for more details and make sure that you understand the implications before
-//     using the function!
-func WithNetNSUnsafe(ns netns.NsHandle, work func() error) error {
+func WithNetNS(ns netns.NsHandle, work func() error) error {
 	runtime.LockOSThread()
 	defer runtime.UnlockOSThread()
 
@@ -45,8 +33,8 @@ func WithNetNSUnsafe(ns netns.NsHandle, work func() error) error {
 	return err
 }
 
-func WithNetNSLinkUnsafe(ns netns.NsHandle, ifName string, work func(link netlink.Link) error) error {
-	return WithNetNSUnsafe(ns, func() error {
+func WithNetNSLink(ns netns.NsHandle, ifName string, work func(link netlink.Link) error) error {
+	return WithNetNS(ns, func() error {
 		link, err := netlink.LinkByName(ifName)
 		if err != nil {
 			if err.Error() == errors.New("Link not found").Error() {
@@ -58,26 +46,12 @@ func WithNetNSLinkUnsafe(ns netns.NsHandle, ifName string, work func(link netlin
 	})
 }
 
-var WeaveUtilCmd = "weaveutil"
-
-// A safe version of WithNetNS* which creates a process executing
-// "nsenter --net=<ns-path> weaveutil <cmd> [args]".
-func WithNetNS(nsPath string, cmd string, args ...string) ([]byte, error) {
-	var stdout, stderr bytes.Buffer
-
-	args = append([]string{"--net=" + nsPath, WeaveUtilCmd, cmd}, args...)
-	c := exec.Command("nsenter", args...)
-	c.Stdout = &stdout
-	c.Stderr = &stderr
-	if err := c.Run(); err != nil {
-		return nil, fmt.Errorf("%s: %s", string(stderr.Bytes()), err)
+func WithNetNSByPath(path string, work func() error) error {
+	ns, err := netns.GetFromPath(path)
+	if err != nil {
+		return err
 	}
-
-	return stdout.Bytes(), nil
-}
-
-func WithNetNSByPid(pid int, cmd string, args ...string) ([]byte, error) {
-	return WithNetNS(NSPathByPid(pid), cmd, args...)
+	return WithNetNS(ns, work)
 }
 
 func NSPathByPid(pid int) string {

net/veth.go

@@ -104,7 +104,10 @@ const (
 )
 
 func interfaceExistsInNamespace(netNSPath string, ifName string) bool {
-	_, err := WithNetNS(netNSPath, "check-iface", ifName)
+	err := WithNetNSByPath(netNSPath, func() error {
+		_, err := netlink.LinkByName(ifName)
+		return err
+	})
 	return err == nil
 }
 
@@ -125,7 +128,9 @@ func AttachContainer(netNSPath, id, ifName, bridgeName string, mtu int, withMult
 			if err := netlink.LinkSetNsFd(veth, int(ns)); err != nil {
 				return fmt.Errorf("failed to move veth to container netns: %s", err)
 			}
-			if _, err := WithNetNS(netNSPath, "setup-iface", peerName, ifName); err != nil {
+			if err := WithNetNS(ns, func() error {
+				return setupIface(peerName, ifName)
+			}); err != nil {
 				return fmt.Errorf("error setting up interface: %s", err)
 			}
 			return nil
@@ -139,19 +144,16 @@ func AttachContainer(netNSPath, id, ifName, bridgeName string, mtu int, withMult
 
 	}
 
-	args := []string{ifName, fmt.Sprintf("%t", withMulticastRoute)}
-	for _, cidr := range cidrs {
-		args = append(args, cidr.String())
-	}
-	if _, err := WithNetNS(netNSPath, "setup-iface-addrs", args...); err != nil {
+	if err := WithNetNSLink(ns, ifName, func(veth netlink.Link) error {
+		return setupIfaceAddrs(veth, withMulticastRoute, cidrs)
+	}); err != nil {
 		return fmt.Errorf("error setting up interface addresses: %s", err)
 	}
 	return nil
 }
 
-// SetupIfaceAddrs is the implementation of the 'setup-iface-addrs' call above,
-// running in another process in the container's netns
-func SetupIfaceAddrs(veth netlink.Link, withMulticastRoute bool, cidrs []*net.IPNet) error {
+// setupIfaceAddrs expects to be called in the container's netns
+func setupIfaceAddrs(veth netlink.Link, withMulticastRoute bool, cidrs []*net.IPNet) error {
 	newAddresses, err := AddAddresses(veth, cidrs)
 	if err != nil {
 		return err
@@ -203,9 +205,8 @@ func SetupIfaceAddrs(veth netlink.Link, withMulticastRoute bool, cidrs []*net.IP
 	return nil
 }
 
-// SetupIface is the implementation of the 'setup-iface' call above,
-// running in another process in the container's netns
-func SetupIface(ifaceName, newIfName string) error {
+// setupIface expects to be called in the container's netns
+func setupIface(ifaceName, newIfName string) error {
 	ipt, err := iptables.New()
 	if err != nil {
 		return err
@@ -219,14 +220,32 @@ func SetupIface(ifaceName, newIfName string) error {
 		return err
 	}
 	// This is only called by AttachContainer which is only called in host pid namespace
-	if err := ConfigureARPCache("/proc", newIfName); err != nil {
+	if err := configureARPCache("/proc", newIfName); err != nil {
 		return err
 	}
 	return ipt.Append("filter", "INPUT", "-i", newIfName, "-d", "224.0.0.0/4", "-j", "DROP")
 }
 
-// NB: This function can be used only by a process that terminates immediately
-//     after calling the function as it changes netns via WithNetNSLinkUnsafe.
+// configureARP is a helper for the Docker plugin which doesn't set the addresses itself
+func ConfigureARP(prefix, rootPath string) error {
+	links, err := netlink.LinkList()
+	if err != nil {
+		return err
+	}
+	for _, link := range links {
+		ifName := link.Attrs().Name
+		if strings.HasPrefix(ifName, prefix) {
+			configureARPCache(rootPath+"/proc", ifName)
+			if addrs, err := netlink.AddrList(link, netlink.FAMILY_V4); err == nil {
+				for _, addr := range addrs {
+					arping.GratuitousArpOverIfaceByName(addr.IPNet.IP, ifName)
+				}
+			}
+		}
+	}
+	return nil
+}
+
 func DetachContainer(netNSPath, id, ifName string, cidrs []*net.IPNet) error {
 	ns, err := netns.GetFromPath(netNSPath)
 	if err != nil {
@@ -239,7 +258,7 @@ func DetachContainer(netNSPath, id, ifName string, cidrs []*net.IPNet) error {
 		return err
 	}
 
-	return WithNetNSLinkUnsafe(ns, ifName, func(veth netlink.Link) error {
+	return WithNetNSLink(ns, ifName, func(veth netlink.Link) error {
 		existingAddrs, err := netlink.AddrList(veth, netlink.FAMILY_V4)
 		if err != nil {
 			return fmt.Errorf("failed to get IP address for %q: %v", veth.Attrs().Name, err)

npc/analyser.go

@@ -4,173 +4,163 @@ import (
 	"fmt"
 
 	apiv1 "k8s.io/api/core/v1"
-	extnapi "k8s.io/api/extensions/v1beta1"
 	networkingv1 "k8s.io/api/networking/v1"
 	"k8s.io/apimachinery/pkg/util/intstr"
 
-	"github.com/weaveworks/weave/npc/ipset"
+	"github.com/weaveworks/weave/net/ipset"
 )
 
-// analysePolicyLegacy is used to analyse extensions/v1beta1/NetworkPolicy (legacy) and
-// implements pre-1.7 k8s netpol semantics, whilst analysePolicy - networking.k8s.io/v1/NetworkPolicy
-// and 1.7 semantics.
-
-func (ns *ns) analysePolicyLegacy(policy *extnapi.NetworkPolicy) (
-	rules map[string]*ruleSpec,
-	nsSelectors, podSelectors map[string]*selectorSpec,
-	err error) {
-
-	nsSelectors = make(map[string]*selectorSpec)
-	podSelectors = make(map[string]*selectorSpec)
-	rules = make(map[string]*ruleSpec)
-
-	dstSelector, err := newSelectorSpec(&policy.Spec.PodSelector, true, ns.name, ipset.HashIP)
-	if err != nil {
-		return nil, nil, nil, err
-	}
-	podSelectors[dstSelector.key] = dstSelector
-
-	for _, ingressRule := range policy.Spec.Ingress {
-		if ingressRule.Ports != nil && len(ingressRule.Ports) == 0 {
-			// Ports is empty, this rule matches no ports (no traffic matches).
-			continue
-		}
-
-		if ingressRule.From != nil && len(ingressRule.From) == 0 {
-			// From is empty, this rule matches no sources (no traffic matches).
-			continue
-		}
-
-		if ingressRule.From == nil {
-			// From is not provided, this rule matches all sources (traffic not restricted by source).
-			if ingressRule.Ports == nil {
-				// Ports is not provided, this rule matches all ports (traffic not restricted by port).
-				rule := newRuleSpec(nil, nil, dstSelector, nil)
-				rules[rule.key] = rule
-			} else {
-				// Ports is present and contains at least one item, then this rule allows traffic
-				// only if the traffic matches at least one port in the ports list.
-				withNormalisedProtoAndPortLegacy(ingressRule.Ports, func(proto, port string) {
-					rule := newRuleSpec(&proto, nil, dstSelector, &port)
-					rules[rule.key] = rule
-				})
-			}
-		} else {
-			// From is present and contains at least on item, this rule allows traffic only if the
-			// traffic matches at least one item in the from list.
-			for _, peer := range ingressRule.From {
-				var srcSelector *selectorSpec
-				if peer.PodSelector != nil {
-					srcSelector, err = newSelectorSpec(peer.PodSelector, false, ns.name, ipset.HashIP)
-					if err != nil {
-						return nil, nil, nil, err
-					}
-					podSelectors[srcSelector.key] = srcSelector
-				}
-				if peer.NamespaceSelector != nil {
-					srcSelector, err = newSelectorSpec(peer.NamespaceSelector, false, "", ipset.ListSet)
-					if err != nil {
-						return nil, nil, nil, err
-					}
-					nsSelectors[srcSelector.key] = srcSelector
-				}
-
-				if ingressRule.Ports == nil {
-					// Ports is not provided, this rule matches all ports (traffic not restricted by port).
-					rule := newRuleSpec(nil, srcSelector, dstSelector, nil)
-					rules[rule.key] = rule
-				} else {
-					// Ports is present and contains at least one item, then this rule allows traffic
-					// only if the traffic matches at least one port in the ports list.
-					withNormalisedProtoAndPortLegacy(ingressRule.Ports, func(proto, port string) {
-						rule := newRuleSpec(&proto, srcSelector, dstSelector, &port)
-						rules[rule.key] = rule
-					})
-				}
-			}
-		}
-	}
-
-	return rules, nsSelectors, podSelectors, nil
-}
-
 func (ns *ns) analysePolicy(policy *networkingv1.NetworkPolicy) (
 	rules map[string]*ruleSpec,
 	nsSelectors, podSelectors map[string]*selectorSpec,
+	ipBlocks map[string]*ipBlockSpec,
 	err error) {
 
 	nsSelectors = make(map[string]*selectorSpec)
 	podSelectors = make(map[string]*selectorSpec)
+	ipBlocks = make(map[string]*ipBlockSpec)
 	rules = make(map[string]*ruleSpec)
+	policyTypes := make([]policyType, 0)
 
+	for _, pt := range policy.Spec.PolicyTypes {
+		if pt == networkingv1.PolicyTypeIngress {
+			policyTypes = append(policyTypes, policyTypeIngress)
+		}
+		if pt == networkingv1.PolicyTypeEgress {
+			policyTypes = append(policyTypes, policyTypeEgress)
+		}
+	}
 	// If empty, matches all pods in a namespace
-	dstSelector, err := newSelectorSpec(&policy.Spec.PodSelector, true, ns.name, ipset.HashIP)
+	targetSelector, err := newSelectorSpec(&policy.Spec.PodSelector, policyTypes, ns.name, ipset.HashIP)
 	if err != nil {
-		return nil, nil, nil, err
-	}
-	podSelectors[dstSelector.key] = dstSelector
-
-	// If ingress is empty then this NetworkPolicy does not allow any traffic
-	if policy.Spec.Ingress == nil || len(policy.Spec.Ingress) == 0 {
-		return
+		return nil, nil, nil, nil, err
 	}
 
-	for _, ingressRule := range policy.Spec.Ingress {
-		// If Ports is empty or missing, this rule matches all ports
-		allPorts := ingressRule.Ports == nil || len(ingressRule.Ports) == 0
-		// If From is empty or missing, this rule matches all sources
-		allSources := ingressRule.From == nil || len(ingressRule.From) == 0
+	// To prevent targetSelector being overwritten by a subsequent selector with
+	// the same key, addIfNotExist MUST be used when adding to podSelectors.
+	// Otherwise, policyTypes of the selector might be lost resulting in
+	// an invalid content in any "default-allow" ipset.
+	addIfNotExist(targetSelector, podSelectors)
 
-		if allSources {
-			if allPorts {
-				rule := newRuleSpec(nil, nil, dstSelector, nil)
-				rules[rule.key] = rule
-			} else {
-				withNormalisedProtoAndPort(ingressRule.Ports, func(proto, port string) {
-					rule := newRuleSpec(&proto, nil, dstSelector, &port)
-					rules[rule.key] = rule
-				})
-			}
-		} else {
-			for _, peer := range ingressRule.From {
-				var srcSelector *selectorSpec
-
-				// NetworkPolicyPeer describes a peer to allow traffic from.
-				// Exactly one of its fields must be specified.
-				if peer.PodSelector != nil {
-					srcSelector, err = newSelectorSpec(peer.PodSelector, false, ns.name, ipset.HashIP)
-					if err != nil {
-						return nil, nil, nil, err
-					}
-					podSelectors[srcSelector.key] = srcSelector
-
-				} else if peer.NamespaceSelector != nil {
-					srcSelector, err = newSelectorSpec(peer.NamespaceSelector, false, "", ipset.ListSet)
-					if err != nil {
-						return nil, nil, nil, err
-					}
-					nsSelectors[srcSelector.key] = srcSelector
-				}
+	// If ingress is empty then this NetworkPolicy does not allow any ingress traffic
+	if policy.Spec.Ingress != nil && len(policy.Spec.Ingress) != 0 {
+		for _, ingressRule := range policy.Spec.Ingress {
+			// If Ports is empty or missing, this rule matches all ports
+			allPorts := ingressRule.Ports == nil || len(ingressRule.Ports) == 0
+			// If From is empty or missing, this rule matches all sources
+			allSources := ingressRule.From == nil || len(ingressRule.From) == 0
 
+			if allSources {
 				if allPorts {
-					rule := newRuleSpec(nil, srcSelector, dstSelector, nil)
+					rule := newRuleSpec(policyTypeIngress, nil, nil, targetSelector, nil)
 					rules[rule.key] = rule
 				} else {
 					withNormalisedProtoAndPort(ingressRule.Ports, func(proto, port string) {
-						rule := newRuleSpec(&proto, srcSelector, dstSelector, &port)
+						rule := newRuleSpec(policyTypeIngress, &proto, nil, targetSelector, &port)
 						rules[rule.key] = rule
 					})
 				}
+			} else {
+				for _, peer := range ingressRule.From {
+					var srcSelector *selectorSpec
+
+					// NetworkPolicyPeer describes a peer to allow traffic from.
+					// Exactly one of its fields must be specified.
+					if peer.PodSelector != nil {
+						srcSelector, err = newSelectorSpec(peer.PodSelector, nil, ns.name, ipset.HashIP)
+						if err != nil {
+							return nil, nil, nil, nil, err
+						}
+						addIfNotExist(srcSelector, podSelectors)
+
+					} else if peer.NamespaceSelector != nil {
+						srcSelector, err = newSelectorSpec(peer.NamespaceSelector, nil, "", ipset.ListSet)
+						if err != nil {
+							return nil, nil, nil, nil, err
+						}
+						nsSelectors[srcSelector.key] = srcSelector
+					}
+
+					if allPorts {
+						rule := newRuleSpec(policyTypeIngress, nil, srcSelector, targetSelector, nil)
+						rules[rule.key] = rule
+					} else {
+						withNormalisedProtoAndPort(ingressRule.Ports, func(proto, port string) {
+							rule := newRuleSpec(policyTypeIngress, &proto, srcSelector, targetSelector, &port)
+							rules[rule.key] = rule
+						})
+					}
+				}
 			}
 		}
 	}
 
-	return rules, nsSelectors, podSelectors, nil
+	// If egress is empty then this NetworkPolicy does not allow any egress traffic
+	if policy.Spec.Egress != nil && len(policy.Spec.Egress) != 0 {
+		for _, egressRule := range policy.Spec.Egress {
+			// If Ports is empty or missing, this rule matches all ports
+			allPorts := egressRule.Ports == nil || len(egressRule.Ports) == 0
+			// If To is empty or missing, this rule matches all destinations
+			allDestinations := egressRule.To == nil || len(egressRule.To) == 0
+
+			if allDestinations {
+				if allPorts {
+					rule := newRuleSpec(policyTypeEgress, nil, targetSelector, nil, nil)
+					rules[rule.key] = rule
+				} else {
+					withNormalisedProtoAndPort(egressRule.Ports, func(proto, port string) {
+						rule := newRuleSpec(policyTypeEgress, &proto, targetSelector, nil, &port)
+						rules[rule.key] = rule
+					})
+				}
+			} else {
+				for _, peer := range egressRule.To {
+					var dstSelector *selectorSpec
+					var dstRuleHost ruleHost
+
+					// NetworkPolicyPeer describes a peer to allow traffic to.
+					// Exactly one of its fields must be specified.
+					if peer.PodSelector != nil {
+						dstSelector, err = newSelectorSpec(peer.PodSelector, nil, ns.name, ipset.HashIP)
+						if err != nil {
+							return nil, nil, nil, nil, err
+						}
+						addIfNotExist(dstSelector, podSelectors)
+						dstRuleHost = dstSelector
+
+					} else if peer.NamespaceSelector != nil {
+						dstSelector, err = newSelectorSpec(peer.NamespaceSelector, nil, "", ipset.ListSet)
+						if err != nil {
+							return nil, nil, nil, nil, err
+						}
+						nsSelectors[dstSelector.key] = dstSelector
+						dstRuleHost = dstSelector
+					} else if peer.IPBlock != nil {
+						ipBlock := newIPBlockSpec(peer.IPBlock, ns.name)
+						ipBlocks[ipBlock.key] = ipBlock
+						dstRuleHost = ipBlock
+					}
+
+					if allPorts {
+						rule := newRuleSpec(policyTypeEgress, nil, targetSelector, dstRuleHost, nil)
+						rules[rule.key] = rule
+					} else {
+						withNormalisedProtoAndPort(egressRule.Ports, func(proto, port string) {
+							rule := newRuleSpec(policyTypeEgress, &proto, targetSelector, dstRuleHost, &port)
+							rules[rule.key] = rule
+						})
+					}
+				}
+			}
+		}
+	}
+
+	return rules, nsSelectors, podSelectors, ipBlocks, nil
 }
 
-func withNormalisedProtoAndPortLegacy(npps []extnapi.NetworkPolicyPort, f func(proto, port string)) {
-	for _, npp := range npps {
-		f(proto(npp.Protocol), port(npp.Port))
+func addIfNotExist(s *selectorSpec, ss map[string]*selectorSpec) {
+	if _, ok := ss[s.key]; !ok {
+		ss[s.key] = s
 	}
 }
 

npc/constants.go

@@ -7,6 +7,12 @@ const (
 	DefaultChain = "WEAVE-NPC-DEFAULT"
 	IngressChain = "WEAVE-NPC-INGRESS"
 
+	EgressChain        = "WEAVE-NPC-EGRESS"
+	EgressDefaultChain = "WEAVE-NPC-EGRESS-DEFAULT"
+	EgressCustomChain  = "WEAVE-NPC-EGRESS-CUSTOM"
+	EgressMarkChain    = "WEAVE-NPC-EGRESS-ACCEPT"
+	EgressMark         = "0x40000/0x40000"
+
 	IpsetNamePrefix = "weave-"
 
 	LocalIpset = IpsetNamePrefix + "local-pods"

npc/controller.go

@@ -9,7 +9,7 @@ import (
 	networkingv1 "k8s.io/api/networking/v1"
 
 	"github.com/weaveworks/weave/common"
-	"github.com/weaveworks/weave/npc/ipset"
+	"github.com/weaveworks/weave/net/ipset"
 	"github.com/weaveworks/weave/npc/iptables"
 )
 
@@ -37,19 +37,16 @@ type controller struct {
 
 	nss         map[string]*ns // ns name -> ns struct
 	nsSelectors *selectorSet   // selector string -> nsSelector
-
-	legacy bool // use legacy network policy semantics (k8s pre-1.7)
 }
 
-func New(nodeName string, legacy bool, ipt iptables.Interface, ips ipset.Interface) NetworkPolicyController {
+func New(nodeName string, ipt iptables.Interface, ips ipset.Interface) NetworkPolicyController {
 	c := &controller{
 		nodeName: nodeName,
-		legacy:   legacy,
 		ipt:      ipt,
 		ips:      ips,
 		nss:      make(map[string]*ns)}
 
-	doNothing := func(*selector) error { return nil }
+	doNothing := func(*selector, policyType) error { return nil }
 	c.nsSelectors = newSelectorSet(ips, c.onNewNsSelector, doNothing, doNothing)
 
 	return c
@@ -71,7 +68,7 @@ func (npc *controller) onNewNsSelector(selector *selector) error {
 func (npc *controller) withNS(name string, f func(ns *ns) error) error {
 	ns, found := npc.nss[name]
 	if !found {
-		newNs, err := newNS(name, npc.nodeName, npc.legacy, npc.ipt, npc.ips, npc.nsSelectors)
+		newNs, err := newNS(name, npc.nodeName, npc.ipt, npc.ips, npc.nsSelectors)
 		if err != nil {
 			return err
 		}

npc/controller_test.go

@@ -1,12 +1,14 @@
 package npc
 
 import (
+	"fmt"
 	"log"
+	"strings"
 	"testing"
 
 	"github.com/pkg/errors"
 	"github.com/stretchr/testify/require"
-	"github.com/weaveworks/weave/npc/ipset"
+	"github.com/weaveworks/weave/net/ipset"
 	coreapi "k8s.io/api/core/v1"
 	networkingv1 "k8s.io/api/networking/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@@ -103,18 +105,46 @@ func (i *mockIPSet) List(prefix string) ([]ipset.Name, error) {
 }
 
 type mockIPTables struct {
+	rules map[string]map[string]struct{} // chain -> rulespec -> struct{}
+}
+
+func newMockIPTables() *mockIPTables {
+	return &mockIPTables{rules: make(map[string]map[string]struct{})}
 }
 
 func (ipt *mockIPTables) Append(table, chain string, rulespec ...string) error {
+	if table != TableFilter {
+		return fmt.Errorf("invalid table: %q", table)
+	}
+
+	if _, found := ipt.rules[chain]; !found {
+		ipt.rules[chain] = make(map[string]struct{})
+	}
+
+	rule := strings.Join(rulespec, " ")
+	if _, found := ipt.rules[chain][rule]; found {
+		return fmt.Errorf("rule already exists. chain: %q, rule: %q", chain, rule)
+	}
+
+	ipt.rules[chain][rule] = struct{}{}
+
 	return nil
 }
 
 func (ipt *mockIPTables) Delete(table, chain string, rulespec ...string) error {
+	rule := strings.Join(rulespec, " ")
+
+	if _, found := ipt.rules[chain][rule]; !found {
+		return fmt.Errorf("rule does not exist. chain: %q rule: %q", chain, rule)
+	}
+
+	delete(ipt.rules[chain], rule)
+
 	return nil
 }
 
 func (ipt *mockIPTables) Insert(table, chain string, pos int, rulespec ...string) error {
-	return nil
+	return errors.New("Not Implemented")
 }
 
 func TestRegressionPolicyNamespaceOrdering3059(t *testing.T) {
@@ -144,6 +174,7 @@ func TestRegressionPolicyNamespaceOrdering3059(t *testing.T) {
 			Namespace: "destination",
 		},
 		Spec: networkingv1.NetworkPolicySpec{
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeIngress},
 			Ingress: []networkingv1.NetworkPolicyIngressRule{
 				{
 					From: []networkingv1.NetworkPolicyPeer{
@@ -167,7 +198,7 @@ func TestRegressionPolicyNamespaceOrdering3059(t *testing.T) {
 
 	// Namespaces first
 	m := newMockIPSet()
-	controller := New("foo", false, &mockIPTables{}, &m)
+	controller := New("foo", newMockIPTables(), &m)
 
 	const (
 		selectorIPSetName = "weave-I239Zp%sCvoVt*D6u=A!2]YEk"
@@ -183,7 +214,7 @@ func TestRegressionPolicyNamespaceOrdering3059(t *testing.T) {
 
 	// NetworkPolicy first
 	m = newMockIPSet()
-	controller = New("foo", false, &mockIPTables{}, &m)
+	controller = New("foo", newMockIPTables(), &m)
 
 	controller.AddNetworkPolicy(networkPolicy)
 
@@ -193,17 +224,18 @@ func TestRegressionPolicyNamespaceOrdering3059(t *testing.T) {
 	require.Equal(t, true, len(m.sets[selectorIPSetName].subSets[sourceIPSetName]) > 0)
 }
 
-// Tests default-allow ipset behavior when running in non-legacy mode.
+// Tests default-allow ipset behavior
 func TestDefaultAllow(t *testing.T) {
 	const (
-		defaultAllowIPSetName = "weave-E.1.0W^NGSp]0_t5WwH/]gX@L"
-		fooPodIP              = "10.32.0.10"
-		barPodIP              = "10.32.0.11"
-		barPodNewIP           = "10.32.0.12"
+		ingressDefaultAllowIPSetName = "weave-;rGqyMIl1HN^cfDki~Z$3]6!N"
+		egressDefaultAllowIPSetName  = "weave-s_+ChJId4Uy_$}G;WdH|~TK)I"
+		fooPodIP                     = "10.32.0.10"
+		barPodIP                     = "10.32.0.11"
+		barPodNewIP                  = "10.32.0.12"
 	)
 
 	m := newMockIPSet()
-	controller := New("bar", false, &mockIPTables{}, &m)
+	controller := New("bar", newMockIPTables(), &m)
 
 	defaultNamespace := &coreapi.Namespace{
 		ObjectMeta: metav1.ObjectMeta{
@@ -213,7 +245,8 @@ func TestDefaultAllow(t *testing.T) {
 	controller.AddNamespace(defaultNamespace)
 
 	// Should create an ipset for default-allow
-	require.Contains(t, m.sets, defaultAllowIPSetName)
+	require.Contains(t, m.sets, ingressDefaultAllowIPSetName)
+	require.Contains(t, m.sets, egressDefaultAllowIPSetName)
 
 	podFoo := &coreapi.Pod{
 		ObjectMeta: metav1.ObjectMeta{
@@ -225,7 +258,8 @@ func TestDefaultAllow(t *testing.T) {
 	controller.AddPod(podFoo)
 
 	// Should add the foo pod to default-allow
-	require.True(t, m.entryExists(defaultAllowIPSetName, fooPodIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, fooPodIP))
+	require.True(t, m.entryExists(egressDefaultAllowIPSetName, fooPodIP))
 
 	podBar := &coreapi.Pod{
 		ObjectMeta: metav1.ObjectMeta{
@@ -236,10 +270,12 @@ func TestDefaultAllow(t *testing.T) {
 		Status: coreapi.PodStatus{PodIP: barPodIP}}
 	podBarNoIP := &coreapi.Pod{ObjectMeta: podBar.ObjectMeta}
 	controller.AddPod(podBarNoIP)
+
 	controller.UpdatePod(podBarNoIP, podBar)
 
 	// Should add the bar pod to default-allow
-	require.True(t, m.entryExists(defaultAllowIPSetName, barPodIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, barPodIP))
+	require.True(t, m.entryExists(egressDefaultAllowIPSetName, barPodIP))
 
 	// Allow access from the bar pod to the foo pod
 	netpol := &networkingv1.NetworkPolicy{
@@ -248,6 +284,10 @@ func TestDefaultAllow(t *testing.T) {
 			Namespace: "default",
 		},
 		Spec: networkingv1.NetworkPolicySpec{
+			PolicyTypes: []networkingv1.PolicyType{
+				networkingv1.PolicyTypeIngress,
+				networkingv1.PolicyTypeEgress,
+			},
 			PodSelector: metav1.LabelSelector{MatchLabels: map[string]string{"run": "foo"}},
 			Ingress: []networkingv1.NetworkPolicyIngressRule{{
 				From: []networkingv1.NetworkPolicyPeer{{
@@ -256,55 +296,74 @@ func TestDefaultAllow(t *testing.T) {
 					},
 				}},
 			}},
+			Egress: []networkingv1.NetworkPolicyEgressRule{{
+				To: []networkingv1.NetworkPolicyPeer{{
+					IPBlock: &networkingv1.IPBlock{CIDR: "192.168.48.0/24"},
+				}},
+			}},
 		},
 	}
 	controller.AddNetworkPolicy(netpol)
 
 	// Should remove the foo pod from default-allow as the netpol selects it
-	require.False(t, m.entryExists(defaultAllowIPSetName, fooPodIP))
-	require.True(t, m.entryExists(defaultAllowIPSetName, barPodIP))
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, fooPodIP))
+	require.False(t, m.entryExists(egressDefaultAllowIPSetName, fooPodIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, barPodIP))
+	require.True(t, m.entryExists(egressDefaultAllowIPSetName, barPodIP))
 
 	podBarWithNewIP := *podBar
 	podBarWithNewIP.Status.PodIP = barPodNewIP
 	controller.UpdatePod(podBar, &podBarWithNewIP)
 
 	// Should update IP addr of the bar pod in default-allow
-	require.False(t, m.entryExists(defaultAllowIPSetName, barPodIP))
-	require.True(t, m.entryExists(defaultAllowIPSetName, barPodNewIP))
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, barPodIP))
+	require.False(t, m.entryExists(egressDefaultAllowIPSetName, barPodIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, barPodNewIP))
+	require.True(t, m.entryExists(egressDefaultAllowIPSetName, barPodNewIP))
 
 	controller.UpdatePod(&podBarWithNewIP, podBarNoIP)
 	// Should remove the bar pod from default-allow as it does not have any IP addr
-	require.False(t, m.entryExists(defaultAllowIPSetName, barPodNewIP))
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, barPodNewIP))
+	require.False(t, m.entryExists(egressDefaultAllowIPSetName, barPodNewIP))
 
 	podFooWithNewLabel := *podFoo
 	podFooWithNewLabel.ObjectMeta.Labels = map[string]string{"run": "new-foo"}
 	controller.UpdatePod(podFoo, &podFooWithNewLabel)
 
 	// Should bring back the foo pod to default-allow as it does not match dst of any netpol
-	require.True(t, m.entryExists(defaultAllowIPSetName, fooPodIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, fooPodIP))
+	require.True(t, m.entryExists(egressDefaultAllowIPSetName, fooPodIP))
 
 	controller.UpdatePod(&podFooWithNewLabel, podFoo)
 	// Should remove from default-allow as it matches the netpol after the update
-	require.False(t, m.entryExists(defaultAllowIPSetName, fooPodIP))
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, fooPodIP))
+	require.False(t, m.entryExists(egressDefaultAllowIPSetName, fooPodIP))
 
 	controller.DeleteNetworkPolicy(netpol)
 	// Should bring back the foo pod to default-allow as no netpol selects it
-	require.True(t, m.entryExists(defaultAllowIPSetName, fooPodIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, fooPodIP))
+	require.True(t, m.entryExists(egressDefaultAllowIPSetName, fooPodIP))
 
 	controller.DeletePod(podFoo)
 	// Should remove foo pod from default-allow
-	require.False(t, m.entryExists(defaultAllowIPSetName, fooPodIP))
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, fooPodIP))
+	require.False(t, m.entryExists(egressDefaultAllowIPSetName, fooPodIP))
+
+	controller.DeleteNamespace(defaultNamespace)
+	// Should remove default ipset
+	require.NotContains(t, m.sets, ingressDefaultAllowIPSetName)
+	require.NotContains(t, m.sets, egressDefaultAllowIPSetName)
 }
 
 func TestOutOfOrderPodEvents(t *testing.T) {
 	const (
-		defaultAllowIPSetName = "weave-E.1.0W^NGSp]0_t5WwH/]gX@L"
-		runBarIPSetName       = "weave-bZ~x=yBgzH)Ht()K*Uv3z{M]Y"
-		podIP                 = "10.32.0.10"
+		ingressDefaultAllowIPSetName = "weave-;rGqyMIl1HN^cfDki~Z$3]6!N"
+		runBarIPSetName              = "weave-bZ~x=yBgzH)Ht()K*Uv3z{M]Y"
+		podIP                        = "10.32.0.10"
 	)
 
 	m := newMockIPSet()
-	controller := New("qux", false, &mockIPTables{}, &m)
+	controller := New("qux", newMockIPTables(), &m)
 
 	defaultNamespace := &coreapi.Namespace{
 		ObjectMeta: metav1.ObjectMeta{
@@ -323,7 +382,7 @@ func TestOutOfOrderPodEvents(t *testing.T) {
 	controller.AddPod(podFoo)
 
 	// Should be in default-allow as no netpol selects podFoo
-	require.True(t, m.entryExists(defaultAllowIPSetName, podIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
 
 	netpol := &networkingv1.NetworkPolicy{
 		ObjectMeta: metav1.ObjectMeta{
@@ -331,6 +390,7 @@ func TestOutOfOrderPodEvents(t *testing.T) {
 			Namespace: "default",
 		},
 		Spec: networkingv1.NetworkPolicySpec{
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeIngress},
 			PodSelector: metav1.LabelSelector{MatchLabels: map[string]string{"run": "foo"}},
 			Ingress: []networkingv1.NetworkPolicyIngressRule{{
 				From: []networkingv1.NetworkPolicyPeer{{
@@ -344,7 +404,7 @@ func TestOutOfOrderPodEvents(t *testing.T) {
 	controller.AddNetworkPolicy(netpol)
 
 	// Shouldn't be in default-allow as netpol above selects podFoo
-	require.False(t, m.entryExists(defaultAllowIPSetName, podIP))
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
 
 	podBar := &coreapi.Pod{
 		ObjectMeta: metav1.ObjectMeta{
@@ -356,8 +416,8 @@ func TestOutOfOrderPodEvents(t *testing.T) {
 	controller.AddPod(podBar)
 
 	// Should be in default-allow as no netpol selects podBar
-	require.True(t, m.entryExists(defaultAllowIPSetName, podIP))
-	require.True(t, m.Exist(podBar.ObjectMeta.UID, defaultAllowIPSetName, podIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
+	require.True(t, m.Exist(podBar.ObjectMeta.UID, ingressDefaultAllowIPSetName, podIP))
 	// Should be in run=bar ipset
 	require.True(t, m.entryExists(runBarIPSetName, podIP))
 
@@ -368,11 +428,167 @@ func TestOutOfOrderPodEvents(t *testing.T) {
 
 	// Should be in default-allow as no netpol selects podBar and podFoo removal
 	// should not affect podBar in default-allow
-	require.True(t, m.entryExists(defaultAllowIPSetName, podIP))
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
 
 	controller.DeletePod(podBar)
 
 	// Should remove from default-allow and run=bar ipsets
-	require.Equal(t, 0, len(m.sets[defaultAllowIPSetName].subSets))
+	require.Equal(t, 0, len(m.sets[ingressDefaultAllowIPSetName].subSets))
 	require.False(t, m.entryExists(runBarIPSetName, podIP))
 }
+
+// Test case for https://github.com/weaveworks/weave/issues/3222
+func TestNewTargetSelector(t *testing.T) {
+	const (
+		ingressDefaultAllowIPSetName = "weave-;rGqyMIl1HN^cfDki~Z$3]6!N"
+		podIP                        = "10.32.0.10"
+	)
+
+	m := newMockIPSet()
+	controller := New("baz", newMockIPTables(), &m)
+
+	defaultNamespace := &coreapi.Namespace{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "default",
+		},
+	}
+	controller.AddNamespace(defaultNamespace)
+
+	podFoo := &coreapi.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			UID:       "foo",
+			Namespace: "default",
+			Name:      "foo",
+			Labels:    map[string]string{"run": "foo"}},
+		Status: coreapi.PodStatus{PodIP: podIP}}
+	controller.AddPod(podFoo)
+
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
+
+	netpolBar := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			UID:       "netpol-bar",
+			Name:      "allow-from-default-1",
+			Namespace: "default",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeIngress},
+			PodSelector: metav1.LabelSelector{},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				From: []networkingv1.NetworkPolicyPeer{{
+					PodSelector: &metav1.LabelSelector{},
+				}},
+			}},
+		},
+	}
+	controller.AddNetworkPolicy(netpolBar)
+
+	// netpolBar target selector selects podFoo
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
+
+	netpolFoo := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			UID:       "netpol-foo",
+			Name:      "allow-from-default-2",
+			Namespace: "default",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeIngress},
+			PodSelector: metav1.LabelSelector{},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				From: []networkingv1.NetworkPolicyPeer{{
+					PodSelector: &metav1.LabelSelector{},
+				}},
+			}},
+		},
+	}
+	controller.AddNetworkPolicy(netpolFoo)
+
+	controller.DeleteNetworkPolicy(netpolBar)
+	// netpolFoo target-selects podFoo
+	require.False(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
+	controller.DeleteNetworkPolicy(netpolFoo)
+	// No netpol target-selects podFoo
+	require.True(t, m.entryExists(ingressDefaultAllowIPSetName, podIP))
+}
+
+func TestEgressPolicyWithIPBlock(t *testing.T) {
+	const (
+		fooPodIP                    = "10.32.0.10"
+		exceptIPSetName             = "weave-j:W$5om!$8JS})buYAD7q^#sX"
+		exceptIPSetNameInNonDefault = "weave-|*P{aL2C@#N/1IZ!IvQItv_pt"
+	)
+
+	m := newMockIPSet()
+	ipt := newMockIPTables()
+	controller := New("foo", ipt, &m)
+
+	netpolFoo := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			UID:       "netpol-foo",
+			Name:      "allow-from-bar-to-foo",
+			Namespace: "default",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PolicyTypes: []networkingv1.PolicyType{
+				networkingv1.PolicyTypeEgress,
+			},
+			PodSelector: metav1.LabelSelector{MatchLabels: map[string]string{"run": "bar"}},
+			Egress: []networkingv1.NetworkPolicyEgressRule{{
+				To: []networkingv1.NetworkPolicyPeer{{
+					IPBlock: &networkingv1.IPBlock{
+						CIDR: "192.168.48.0/24",
+						Except: []string{
+							"192.168.48.1/32",
+							"192.168.48.2/32",
+						},
+					},
+				}},
+			}},
+		},
+	}
+	err := controller.AddNetworkPolicy(netpolFoo)
+	require.NoError(t, err)
+
+	require.Equal(t, 2, len(m.sets[exceptIPSetName].subSets))
+	require.True(t, m.entryExists(exceptIPSetName, "192.168.48.1/32"))
+	require.True(t, m.entryExists(exceptIPSetName, "192.168.48.2/32"))
+
+	// Each egress rule is represented as two iptables rules (-J MARK and -J RETURN).
+	require.Equal(t, 2, len(ipt.rules[EgressCustomChain]))
+	for rule := range ipt.rules[EgressCustomChain] {
+		require.Contains(t, rule, "-d 192.168.48.0/24 -m set ! --match-set "+exceptIPSetName+" dst")
+	}
+
+	// Check that we create a new ipset for the ipBlock bellow. An ipset with the
+	// same content already exists (created by netpolFoo), but we need to create
+	// a new one, as netpolBar is in a different namespace and weave-npc does
+	// object refcounting only within namespace boundaries.
+	netpolBar := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			UID:       "netpol-foo-bar",
+			Name:      "allow-from-bar-to-foo-in-non-default",
+			Namespace: "non-default",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PolicyTypes: []networkingv1.PolicyType{
+				networkingv1.PolicyTypeEgress,
+			},
+			PodSelector: metav1.LabelSelector{MatchLabels: map[string]string{"run": "bar"}},
+			Egress: []networkingv1.NetworkPolicyEgressRule{{
+				To: []networkingv1.NetworkPolicyPeer{{
+					IPBlock: &networkingv1.IPBlock{
+						CIDR: "192.168.48.0/24",
+						Except: []string{
+							"192.168.48.1/32",
+							"192.168.48.2/32",
+						},
+					},
+				}},
+			}},
+		},
+	}
+	err = controller.AddNetworkPolicy(netpolBar)
+	require.NoError(t, err)
+	require.Equal(t, 2, len(m.sets[exceptIPSetNameInNonDefault].subSets))
+}

npc/ipblock.go

@@ -0,0 +1,116 @@
+package npc
+
+import (
+	"fmt"
+	"sort"
+	"strings"
+
+	networkingv1 "k8s.io/api/networking/v1"
+	"k8s.io/apimachinery/pkg/types"
+
+	"github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/net/ipset"
+)
+
+type ipBlockSpec struct {
+	key       string
+	ipsetName ipset.Name // ipset for storing excepted CIDRs
+	ipBlock   *networkingv1.IPBlock
+	nsName    string // Namespace name
+}
+
+func newIPBlockSpec(ipb *networkingv1.IPBlock, nsName string) *ipBlockSpec {
+	spec := &ipBlockSpec{ipBlock: ipb, nsName: nsName}
+
+	if len(ipb.Except) > 0 {
+		sort.Strings(ipb.Except)
+		spec.key = strings.Join(ipb.Except, " ")
+		spec.ipsetName = ipset.Name(IpsetNamePrefix + shortName(nsName+":"+"ipblock-except:"+spec.key))
+	}
+
+	return spec
+}
+
+func (spec *ipBlockSpec) getRuleSpec(src bool) ([]string, string) {
+	dir, dirOpt := "dst", "-d"
+	if src {
+		dir, dirOpt = "src", "-s"
+	}
+
+	if spec.key == "" {
+		rule := []string{dirOpt, spec.ipBlock.CIDR}
+		comment := fmt.Sprintf("cidr: %s", spec.ipBlock.CIDR)
+		return rule, comment
+	}
+
+	rule := []string{dirOpt, spec.ipBlock.CIDR,
+		"-m", "set", "!", "--match-set", string(spec.ipsetName), dir}
+	comment := fmt.Sprintf("cidr: %s except [%s]", spec.ipBlock.CIDR, spec.key)
+	return rule, comment
+}
+
+type ipBlockSet struct {
+	ips   ipset.Interface
+	users map[string]map[types.UID]struct{}
+}
+
+func newIPBlockSet(ips ipset.Interface) *ipBlockSet {
+	return &ipBlockSet{
+		ips:   ips,
+		users: make(map[string]map[types.UID]struct{}),
+	}
+}
+
+func (s *ipBlockSet) deprovision(user types.UID, current, desired map[string]*ipBlockSpec) error {
+	for key, spec := range current {
+		if key == "" {
+			continue
+		}
+
+		if _, found := desired[key]; !found {
+			delete(s.users[key], user)
+			if len(s.users[key]) == 0 {
+				common.Log.Infof("destroying ipset: %#v", spec)
+				if err := s.ips.Destroy(spec.ipsetName); err != nil {
+					return err
+				}
+
+				delete(s.users, key)
+			}
+		}
+	}
+
+	return nil
+}
+
+func (s *ipBlockSet) provision(user types.UID, current, desired map[string]*ipBlockSpec) (err error) {
+	for key, spec := range desired {
+		if key == "" {
+			// No need to provision an ipBlock with empty list of excepted CIDRs
+			// (i.e. no ipset is needed for the related iptables rule).
+			continue
+		}
+
+		if _, found := current[key]; !found {
+			if _, found := s.users[key]; !found {
+				common.Log.Infof("creating ipset: %#v", spec)
+				if err := s.ips.Create(spec.ipsetName, ipset.HashNet); err != nil {
+					return err
+				}
+
+				// TODO(brb) Pass comment to ips.Create instead.
+				comment := "excepted from " + spec.ipBlock.CIDR
+				for _, cidr := range spec.ipBlock.Except {
+					if err = s.ips.AddEntry(user, spec.ipsetName, cidr, comment); err != nil {
+						return err
+					}
+				}
+
+				s.users[key] = make(map[types.UID]struct{})
+			}
+			s.users[key][user] = struct{}{}
+		}
+	}
+
+	return nil
+}

npc/namespace.go

@@ -1,7 +1,6 @@
 package npc
 
 import (
-	"encoding/json"
 	"errors"
 	"fmt"
 
@@ -13,7 +12,7 @@ import (
 	"k8s.io/apimachinery/pkg/util/uuid"
 
 	"github.com/weaveworks/weave/common"
-	"github.com/weaveworks/weave/npc/ipset"
+	"github.com/weaveworks/weave/net/ipset"
 	"github.com/weaveworks/weave/npc/iptables"
 )
 
@@ -32,20 +31,20 @@ type ns struct {
 	uid     types.UID     // surrogate UID to own allPods selector
 	allPods *selectorSpec // hash:ip ipset of all pod IPs in this namespace
 
-	// stores IP addrs of pods which are not selected by any dst podSelector of
-	// any netpol; used only in non-legacy mode and is used as a dst in
-	// the WEAVE-NPC-DEFAULT iptables chain.
-	defaultAllowIPSet ipset.Name
+	// stores IP addrs of pods which are not selected by any target podSelector of
+	// any netpol; used as a target in the WEAVE-NPC-{INGRESS,EGRESS}-DEFAULT
+	// iptables chains.
+	ingressDefaultAllowIPSet ipset.Name
+	egressDefaultAllowIPSet  ipset.Name
 
 	nsSelectors  *selectorSet
 	podSelectors *selectorSet
+	ipBlocks     *ipBlockSet
 	rules        *ruleSet
-
-	legacy bool
 }
 
-func newNS(name, nodeName string, legacy bool, ipt iptables.Interface, ips ipset.Interface, nsSelectors *selectorSet) (*ns, error) {
-	allPods, err := newSelectorSpec(&metav1.LabelSelector{}, false, name, ipset.HashIP)
+func newNS(name, nodeName string, ipt iptables.Interface, ips ipset.Interface, nsSelectors *selectorSet) (*ns, error) {
+	allPods, err := newSelectorSpec(&metav1.LabelSelector{}, nil, name, ipset.HashIP)
 	if err != nil {
 		return nil, err
 	}
@@ -60,19 +59,24 @@ func newNS(name, nodeName string, legacy bool, ipt iptables.Interface, ips ipset
 		uid:         uuid.NewUUID(),
 		allPods:     allPods,
 		nsSelectors: nsSelectors,
+		ipBlocks:    newIPBlockSet(ips),
 		rules:       newRuleSet(ipt),
-		legacy:      legacy}
-
-	ns.podSelectors = newSelectorSet(ips, ns.onNewPodSelector, ns.onNewDstPodSelector, ns.onDestroyDstPodSelector)
-
-	if !legacy {
-		defaultAllowIPSet := ipset.Name(IpsetNamePrefix + shortName("default-allow:"+name))
-		if err := ips.Create(defaultAllowIPSet, ipset.HashIP); err != nil {
-			return nil, err
-		}
-		ns.defaultAllowIPSet = defaultAllowIPSet
 	}
 
+	ns.podSelectors = newSelectorSet(ips, ns.onNewPodSelector, ns.onNewTargetPodSelector, ns.onDestroyTargetPodSelector)
+
+	ingressDefaultAllowIPSet := ipset.Name(IpsetNamePrefix + shortName("ingress-default-allow:"+name))
+	if err := ips.Create(ingressDefaultAllowIPSet, ipset.HashIP); err != nil {
+		return nil, err
+	}
+	ns.ingressDefaultAllowIPSet = ingressDefaultAllowIPSet
+
+	egressDefaultAllowIPSet := ipset.Name(IpsetNamePrefix + shortName("egress-default-allow:"+name))
+	if err := ips.Create(egressDefaultAllowIPSet, ipset.HashIP); err != nil {
+		return nil, err
+	}
+	ns.egressDefaultAllowIPSet = egressDefaultAllowIPSet
+
 	if err := ns.podSelectors.provision(ns.uid, nil, map[string]*selectorSpec{ns.allPods.key: ns.allPods}); err != nil {
 		return nil, err
 	}
@@ -102,16 +106,13 @@ func (ns *ns) onNewPodSelector(selector *selector) error {
 	return nil
 }
 
-func (ns *ns) onNewDstPodSelector(selector *selector) error {
-	if ns.legacy {
-		return nil
-	}
-
+func (ns *ns) onNewTargetPodSelector(selector *selector, policyType policyType) error {
 	for _, pod := range ns.pods {
 		if hasIP(pod) {
 			// Remove the pod from default-allow if dst podselector matches the pod
+			ipset := ns.defaultAllowIPSetName(policyType)
 			if selector.matches(pod.ObjectMeta.Labels) {
-				if err := ns.ips.DelEntry(pod.ObjectMeta.UID, ns.defaultAllowIPSet, pod.Status.PodIP); err != nil {
+				if err := ns.ips.DelEntry(pod.ObjectMeta.UID, ipset, pod.Status.PodIP); err != nil {
 					return err
 				}
 			}
@@ -121,15 +122,11 @@ func (ns *ns) onNewDstPodSelector(selector *selector) error {
 	return nil
 }
 
-func (ns *ns) onDestroyDstPodSelector(selector *selector) error {
-	if ns.legacy {
-		return nil
-	}
-
+func (ns *ns) onDestroyTargetPodSelector(selector *selector, policyType policyType) error {
 	for _, pod := range ns.pods {
 		if hasIP(pod) {
 			if selector.matches(pod.ObjectMeta.Labels) {
-				if err := ns.addToDefaultAllowIfNoMatching(pod); err != nil {
+				if err := ns.addToDefaultAllowIfNoMatching(pod, policyType); err != nil {
 					return err
 				}
 			}
@@ -139,18 +136,19 @@ func (ns *ns) onDestroyDstPodSelector(selector *selector) error {
 	return nil
 }
 
-// Add pod IP addr to default-allow ipset if there are no matching dst selectors
-func (ns *ns) addToDefaultAllowIfNoMatching(pod *coreapi.Pod) error {
+// Add pod IP addr to default-allow ipset if there are no matching target selectors
+func (ns *ns) addToDefaultAllowIfNoMatching(pod *coreapi.Pod, policyType policyType) error {
 	found := false
 	// TODO(mp) optimize (avoid iterating over selectors) by ref counting IP addrs.
 	for _, s := range ns.podSelectors.entries {
-		if ns.podSelectors.dstSelectorExist(s) && s.matches(pod.ObjectMeta.Labels) {
+		if ns.podSelectors.targetSelectorExist(s, policyType) && s.matches(pod.ObjectMeta.Labels) {
 			found = true
 			break
 		}
 	}
 	if !found {
-		if err := ns.ips.AddEntry(pod.ObjectMeta.UID, ns.defaultAllowIPSet, pod.Status.PodIP, podComment(pod)); err != nil {
+		ipset := ns.defaultAllowIPSetName(policyType)
+		if err := ns.ips.AddEntry(pod.ObjectMeta.UID, ipset, pod.Status.PodIP, podComment(pod)); err != nil {
 			return err
 		}
 	}
@@ -175,13 +173,18 @@ func (ns *ns) addPod(obj *coreapi.Pod) error {
 		ns.ips.AddEntry(obj.ObjectMeta.UID, LocalIpset, obj.Status.PodIP, podComment(obj))
 	}
 
-	found, err := ns.podSelectors.addToMatching(obj.ObjectMeta.UID, obj.ObjectMeta.Labels, obj.Status.PodIP, podComment(obj))
+	foundIngress, foundEgress, err := ns.podSelectors.addToMatching(obj.ObjectMeta.UID, obj.ObjectMeta.Labels, obj.Status.PodIP, podComment(obj))
 	if err != nil {
 		return err
 	}
-	// If there are no matching dst selectors, add the pod to default-allow
-	if !ns.legacy && !found {
-		if err := ns.ips.AddEntry(obj.ObjectMeta.UID, ns.defaultAllowIPSet, obj.Status.PodIP, podComment(obj)); err != nil {
+	// If there are no matching target selectors, add the pod to default-allow
+	if !foundIngress {
+		if err := ns.ips.AddEntry(obj.ObjectMeta.UID, ns.ingressDefaultAllowIPSet, obj.Status.PodIP, podComment(obj)); err != nil {
+			return err
+		}
+	}
+	if !foundEgress {
+		if err := ns.ips.AddEntry(obj.ObjectMeta.UID, ns.egressDefaultAllowIPSet, obj.Status.PodIP, podComment(obj)); err != nil {
 			return err
 		}
 	}
@@ -202,10 +205,11 @@ func (ns *ns) updatePod(oldObj, newObj *coreapi.Pod) error {
 			ns.ips.DelEntry(oldObj.ObjectMeta.UID, LocalIpset, oldObj.Status.PodIP)
 		}
 
-		if !ns.legacy {
-			if err := ns.ips.DelEntry(oldObj.ObjectMeta.UID, ns.defaultAllowIPSet, oldObj.Status.PodIP); err != nil {
-				return err
-			}
+		if err := ns.ips.DelEntry(oldObj.ObjectMeta.UID, ns.ingressDefaultAllowIPSet, oldObj.Status.PodIP); err != nil {
+			return err
+		}
+		if err := ns.ips.DelEntry(oldObj.ObjectMeta.UID, ns.egressDefaultAllowIPSet, oldObj.Status.PodIP); err != nil {
+			return err
 		}
 
 		return ns.podSelectors.delFromMatching(oldObj.ObjectMeta.UID, oldObj.ObjectMeta.Labels, oldObj.Status.PodIP)
@@ -215,13 +219,18 @@ func (ns *ns) updatePod(oldObj, newObj *coreapi.Pod) error {
 		if ns.checkLocalPod(newObj) {
 			ns.ips.AddEntry(newObj.ObjectMeta.UID, LocalIpset, newObj.Status.PodIP, podComment(newObj))
 		}
-		found, err := ns.podSelectors.addToMatching(newObj.ObjectMeta.UID, newObj.ObjectMeta.Labels, newObj.Status.PodIP, podComment(newObj))
+		foundIngress, foundEgress, err := ns.podSelectors.addToMatching(newObj.ObjectMeta.UID, newObj.ObjectMeta.Labels, newObj.Status.PodIP, podComment(newObj))
 		if err != nil {
 			return err
 		}
 
-		if !ns.legacy && !found {
-			if err := ns.ips.AddEntry(newObj.ObjectMeta.UID, ns.defaultAllowIPSet, newObj.Status.PodIP, podComment(newObj)); err != nil {
+		if !foundIngress {
+			if err := ns.ips.AddEntry(newObj.ObjectMeta.UID, ns.ingressDefaultAllowIPSet, newObj.Status.PodIP, podComment(newObj)); err != nil {
+				return err
+			}
+		}
+		if !foundEgress {
+			if err := ns.ips.AddEntry(newObj.ObjectMeta.UID, ns.egressDefaultAllowIPSet, newObj.Status.PodIP, podComment(newObj)); err != nil {
 				return err
 			}
 		}
@@ -229,17 +238,11 @@ func (ns *ns) updatePod(oldObj, newObj *coreapi.Pod) error {
 		return nil
 	}
 
-	if !ns.legacy &&
-		oldObj.Status.PodIP != newObj.Status.PodIP &&
-		ns.ips.Exist(oldObj.ObjectMeta.UID, ns.defaultAllowIPSet, oldObj.Status.PodIP) {
-		// Instead of iterating over all selectors we check whether old pod IP
-		// has been inserted into default-allow ipset to decide whether the IP
-		// in the ipset has to be updated.
-
-		if err := ns.ips.DelEntry(oldObj.ObjectMeta.UID, ns.defaultAllowIPSet, oldObj.Status.PodIP); err != nil {
+	if oldObj.Status.PodIP != newObj.Status.PodIP {
+		if err := ns.updateDefaultAllowIPSetEntry(oldObj, newObj, ns.ingressDefaultAllowIPSet); err != nil {
 			return err
 		}
-		if err := ns.ips.AddEntry(newObj.ObjectMeta.UID, ns.defaultAllowIPSet, newObj.Status.PodIP, podComment(newObj)); err != nil {
+		if err := ns.updateDefaultAllowIPSetEntry(oldObj, newObj, ns.egressDefaultAllowIPSet); err != nil {
 			return err
 		}
 	}
@@ -264,17 +267,11 @@ func (ns *ns) updatePod(oldObj, newObj *coreapi.Pod) error {
 				}
 			}
 
-			if !ns.legacy && ns.podSelectors.dstSelectorExist(ps) {
-				switch {
-				case !oldMatch && newMatch:
-					if err := ns.ips.DelEntry(oldObj.ObjectMeta.UID, ns.defaultAllowIPSet, oldObj.Status.PodIP); err != nil {
-						return err
-					}
-				case oldMatch && !newMatch:
-					if err := ns.addToDefaultAllowIfNoMatching(newObj); err != nil {
-						return err
-					}
-				}
+			if err := ns.addOrRemoveToDefaultAllowIPSet(ps, oldObj, newObj, oldMatch, newMatch, policyTypeIngress); err != nil {
+				return err
+			}
+			if err := ns.addOrRemoveToDefaultAllowIPSet(ps, oldObj, newObj, oldMatch, newMatch, policyTypeEgress); err != nil {
+				return err
 			}
 		}
 	}
@@ -282,6 +279,23 @@ func (ns *ns) updatePod(oldObj, newObj *coreapi.Pod) error {
 	return nil
 }
 
+func (ns *ns) addOrRemoveToDefaultAllowIPSet(ps *selector, oldObj, newObj *coreapi.Pod, oldMatch, newMatch bool, policyType policyType) error {
+	ipset := ns.defaultAllowIPSetName(policyType)
+	if ns.podSelectors.targetSelectorExist(ps, policyType) {
+		switch {
+		case !oldMatch && newMatch:
+			if err := ns.ips.DelEntry(oldObj.ObjectMeta.UID, ipset, oldObj.Status.PodIP); err != nil {
+				return err
+			}
+		case oldMatch && !newMatch:
+			if err := ns.addToDefaultAllowIfNoMatching(newObj, policyType); err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
 func (ns *ns) deletePod(obj *coreapi.Pod) error {
 	delete(ns.pods, obj.ObjectMeta.UID)
 
@@ -293,10 +307,11 @@ func (ns *ns) deletePod(obj *coreapi.Pod) error {
 		ns.ips.DelEntry(obj.ObjectMeta.UID, LocalIpset, obj.Status.PodIP)
 	}
 
-	if !ns.legacy {
-		if err := ns.ips.DelEntry(obj.ObjectMeta.UID, ns.defaultAllowIPSet, obj.Status.PodIP); err != nil {
-			return err
-		}
+	if err := ns.ips.DelEntry(obj.ObjectMeta.UID, ns.ingressDefaultAllowIPSet, obj.Status.PodIP); err != nil {
+		return err
+	}
+	if err := ns.ips.DelEntry(obj.ObjectMeta.UID, ns.egressDefaultAllowIPSet, obj.Status.PodIP); err != nil {
+		return err
 	}
 
 	return ns.podSelectors.delFromMatching(obj.ObjectMeta.UID, obj.ObjectMeta.Labels, obj.Status.PodIP)
@@ -305,7 +320,7 @@ func (ns *ns) deletePod(obj *coreapi.Pod) error {
 func (ns *ns) addNetworkPolicy(obj interface{}) error {
 	// Analyse policy, determine which rules and ipsets are required
 
-	uid, rules, nsSelectors, podSelectors, err := ns.analyse(obj)
+	uid, rules, nsSelectors, podSelectors, ipBlocks, err := ns.analyse(obj)
 	if err != nil {
 		return err
 	}
@@ -317,16 +332,19 @@ func (ns *ns) addNetworkPolicy(obj interface{}) error {
 	if err := ns.podSelectors.provision(uid, nil, podSelectors); err != nil {
 		return err
 	}
+	if err := ns.ipBlocks.provision(uid, nil, ipBlocks); err != nil {
+		return err
+	}
 	return ns.rules.provision(uid, nil, rules)
 }
 
 func (ns *ns) updateNetworkPolicy(oldObj, newObj interface{}) error {
 	// Analyse the old and the new policy so we can determine differences
-	oldUID, oldRules, oldNsSelectors, oldPodSelectors, err := ns.analyse(oldObj)
+	oldUID, oldRules, oldNsSelectors, oldPodSelectors, oldIPBlocks, err := ns.analyse(oldObj)
 	if err != nil {
 		return err
 	}
-	newUID, newRules, newNsSelectors, newPodSelectors, err := ns.analyse(newObj)
+	newUID, newRules, newNsSelectors, newPodSelectors, newIPBlocks, err := ns.analyse(newObj)
 	if err != nil {
 		return err
 	}
@@ -344,18 +362,24 @@ func (ns *ns) updateNetworkPolicy(oldObj, newObj interface{}) error {
 	if err := ns.podSelectors.deprovision(oldUID, oldPodSelectors, newPodSelectors); err != nil {
 		return err
 	}
+	if err := ns.ipBlocks.deprovision(oldUID, oldIPBlocks, newIPBlocks); err != nil {
+		return err
+	}
 	if err := ns.nsSelectors.provision(oldUID, oldNsSelectors, newNsSelectors); err != nil {
 		return err
 	}
 	if err := ns.podSelectors.provision(oldUID, oldPodSelectors, newPodSelectors); err != nil {
 		return err
 	}
+	if err := ns.ipBlocks.provision(oldUID, oldIPBlocks, newIPBlocks); err != nil {
+		return err
+	}
 	return ns.rules.provision(oldUID, oldRules, newRules)
 }
 
 func (ns *ns) deleteNetworkPolicy(obj interface{}) error {
 	// Analyse network policy to free resources
-	uid, rules, nsSelectors, podSelectors, err := ns.analyse(obj)
+	uid, rules, nsSelectors, podSelectors, ipBlocks, err := ns.analyse(obj)
 	if err != nil {
 		return err
 	}
@@ -369,73 +393,89 @@ func (ns *ns) deleteNetworkPolicy(obj interface{}) error {
 	if err := ns.nsSelectors.deprovision(uid, nsSelectors, nil); err != nil {
 		return err
 	}
-	return ns.podSelectors.deprovision(uid, podSelectors, nil)
-}
-
-func bypassRule(nsIpsetName ipset.Name, namespace string) []string {
-	return []string{"-m", "set", "--match-set", string(nsIpsetName), "dst", "-j", "ACCEPT", "-m", "comment", "--comment", "DefaultAllow isolation for namespace: " + namespace}
-}
-
-func (ns *ns) ensureBypassRule() error {
-	var ipset ipset.Name
-	if ns.legacy {
-		ipset = ns.allPods.ipsetName
-	} else {
-		ipset = ns.defaultAllowIPSet
+	if err := ns.podSelectors.deprovision(uid, podSelectors, nil); err != nil {
+		return err
+	}
+	if err := ns.ipBlocks.deprovision(uid, ipBlocks, nil); err != nil {
+		return err
 	}
 
-	common.Log.Debugf("ensuring rule for DefaultAllow in namespace: %s, set %s", ns.name, ipset)
-	return ns.ipt.Append(TableFilter, DefaultChain, bypassRule(ipset, ns.name)...)
+	return nil
 }
 
-func (ns *ns) deleteBypassRule() error {
-	var ipset ipset.Name
-	if ns.legacy {
-		ipset = ns.allPods.ipsetName
-	} else {
-		ipset = ns.defaultAllowIPSet
+func (ns *ns) updateDefaultAllowIPSetEntry(oldObj, newObj *coreapi.Pod, ipsetName ipset.Name) error {
+	// Instead of iterating over all selectors we check whether old pod IP
+	// has been inserted into default-allow ipset to decide whether the IP
+	// in the ipset has to be updated.
+	if ns.ips.Exist(oldObj.ObjectMeta.UID, ipsetName, oldObj.Status.PodIP) {
+
+		if err := ns.ips.DelEntry(oldObj.ObjectMeta.UID, ipsetName, oldObj.Status.PodIP); err != nil {
+			return err
+		}
+		if err := ns.ips.AddEntry(newObj.ObjectMeta.UID, ipsetName, newObj.Status.PodIP, podComment(newObj)); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func bypassRules(namespace string, ingress, egress ipset.Name) map[string][][]string {
+	return map[string][][]string{
+		DefaultChain: {
+			{"-m", "set", "--match-set", string(ingress), "dst", "-j", "ACCEPT",
+				"-m", "comment", "--comment", "DefaultAllow ingress isolation for namespace: " + namespace},
+		},
+		EgressDefaultChain: {
+			{"-m", "set", "--match-set", string(egress), "src", "-j", EgressMarkChain,
+				"-m", "comment", "--comment", "DefaultAllow egress isolation for namespace: " + namespace},
+			{"-m", "set", "--match-set", string(egress), "src", "-j", "RETURN",
+				"-m", "comment", "--comment", "DefaultAllow egress isolation for namespace: " + namespace},
+		},
+	}
+}
+
+func (ns *ns) ensureBypassRules() error {
+	for chain, rules := range bypassRules(ns.name, ns.ingressDefaultAllowIPSet, ns.egressDefaultAllowIPSet) {
+		for _, rule := range rules {
+			common.Log.Debugf("adding rule for DefaultAllow in namespace: %s, chain: %s, %s", ns.name, chain, rule)
+			if err := ns.ipt.Append(TableFilter, chain, rule...); err != nil {
+				return err
+			}
+		}
 	}
 
-	common.Log.Debugf("removing default rule in namespace: %s, set %s", ns.name, ipset)
-	return ns.ipt.Delete(TableFilter, DefaultChain, bypassRule(ipset, ns.name)...)
+	return nil
+}
+
+func (ns *ns) deleteBypassRules() error {
+	for chain, rules := range bypassRules(ns.name, ns.ingressDefaultAllowIPSet, ns.egressDefaultAllowIPSet) {
+		for _, rule := range rules {
+			common.Log.Debugf("removing rule for DefaultAllow in namespace: %s, chain: %s, %s", ns.name, chain, rule)
+			if err := ns.ipt.Delete(TableFilter, chain, rule...); err != nil {
+				return err
+			}
+		}
+	}
+
+	return nil
 }
 
 func (ns *ns) addNamespace(obj *coreapi.Namespace) error {
 	ns.namespace = obj
 
-	// Insert a rule to bypass policies if namespace is DefaultAllow
-	if !ns.isDefaultDeny(obj) {
-		if err := ns.ensureBypassRule(); err != nil {
-			return err
-		}
+	// Insert a rule to bypass policies
+	if err := ns.ensureBypassRules(); err != nil {
+		return err
 	}
 
 	// Add namespace ipset to matching namespace selectors
-	_, err := ns.nsSelectors.addToMatching(obj.ObjectMeta.UID, obj.ObjectMeta.Labels, string(ns.allPods.ipsetName), namespaceComment(ns))
+	_, _, err := ns.nsSelectors.addToMatching(obj.ObjectMeta.UID, obj.ObjectMeta.Labels, string(ns.allPods.ipsetName), namespaceComment(ns))
 	return err
 }
 
 func (ns *ns) updateNamespace(oldObj, newObj *coreapi.Namespace) error {
 	ns.namespace = newObj
 
-	// Update bypass rule if ingress default has changed
-	oldDefaultDeny := ns.isDefaultDeny(oldObj)
-	newDefaultDeny := ns.isDefaultDeny(newObj)
-
-	if oldDefaultDeny != newDefaultDeny {
-		common.Log.Infof("namespace DefaultDeny changed from %t to %t", oldDefaultDeny, newDefaultDeny)
-		if oldDefaultDeny {
-			if err := ns.ensureBypassRule(); err != nil {
-				return err
-			}
-		}
-		if newDefaultDeny {
-			if err := ns.deleteBypassRule(); err != nil {
-				return err
-			}
-		}
-	}
-
 	// Re-evaluate namespace selector membership if labels have changed
 	if !equals(oldObj.ObjectMeta.Labels, newObj.ObjectMeta.Labels) {
 		for _, selector := range ns.nsSelectors.entries {
@@ -464,37 +504,24 @@ func (ns *ns) deleteNamespace(obj *coreapi.Namespace) error {
 	ns.namespace = nil
 
 	// Remove bypass rule
-	if !ns.isDefaultDeny(obj) {
-		if err := ns.deleteBypassRule(); err != nil {
-			return err
-		}
+	if err := ns.deleteBypassRules(); err != nil {
+		return err
 	}
 
 	// Remove namespace ipset from any matching namespace selectors
-	return ns.nsSelectors.delFromMatching(obj.ObjectMeta.UID, obj.ObjectMeta.Labels, string(ns.allPods.ipsetName))
-}
-
-func (ns *ns) isDefaultDeny(namespace *coreapi.Namespace) bool {
-	nnpJSON, found := namespace.ObjectMeta.Annotations["net.beta.kubernetes.io/network-policy"]
-	if !found {
-		return false
+	err := ns.nsSelectors.delFromMatching(obj.ObjectMeta.UID, obj.ObjectMeta.Labels, string(ns.allPods.ipsetName))
+	if err != nil {
+		return err
 	}
 
-	if !ns.legacy {
-		common.Log.Warn("DefaultDeny annotation is supported only in legacy mode (--use-legacy-netpol)")
-		return false
+	if err := ns.ips.Destroy(ns.ingressDefaultAllowIPSet); err != nil {
+		return err
+	}
+	if err := ns.ips.Destroy(ns.egressDefaultAllowIPSet); err != nil {
+		return err
 	}
 
-	var nnp NamespaceNetworkPolicy
-	if err := json.Unmarshal([]byte(nnpJSON), &nnp); err != nil {
-		common.Log.Warn("Ignoring network policy annotation: unmarshal failed:", err)
-		// If we can't understand the annotation, behave as if it isn't present
-		return false
-	}
-
-	return nnp.Ingress != nil &&
-		nnp.Ingress.Isolation != nil &&
-		*(nnp.Ingress.Isolation) == DefaultDeny
+	return nil
 }
 
 func hasIP(pod *coreapi.Pod) bool {
@@ -523,10 +550,19 @@ func podComment(pod *coreapi.Pod) string {
 	return fmt.Sprintf("namespace: %s, pod: %s", pod.ObjectMeta.Namespace, pod.ObjectMeta.Name)
 }
 
+func (ns *ns) defaultAllowIPSetName(pt policyType) ipset.Name {
+	ipset := ns.ingressDefaultAllowIPSet
+	if pt == policyTypeEgress {
+		ipset = ns.egressDefaultAllowIPSet
+	}
+	return ipset
+}
+
 func (ns *ns) analyse(obj interface{}) (
 	uid types.UID,
 	rules map[string]*ruleSpec,
 	nsSelectors, podSelectors map[string]*selectorSpec,
+	ipBlocks map[string]*ipBlockSpec,
 	err error) {
 
 	switch p := obj.(type) {
@@ -541,16 +577,9 @@ func (ns *ns) analyse(obj interface{}) (
 	ns.policies[uid] = obj
 
 	// Analyse policy, determine which rules and ipsets are required
-	if ns.legacy {
-		rules, nsSelectors, podSelectors, err = ns.analysePolicyLegacy(obj.(*extnapi.NetworkPolicy))
-		if err != nil {
-			return
-		}
-	} else {
-		rules, nsSelectors, podSelectors, err = ns.analysePolicy(obj.(*networkingv1.NetworkPolicy))
-		if err != nil {
-			return
-		}
+	rules, nsSelectors, podSelectors, ipBlocks, err = ns.analysePolicy(obj.(*networkingv1.NetworkPolicy))
+	if err != nil {
+		return
 	}
 
 	return

npc/rule.go

@@ -2,6 +2,7 @@ package npc
 
 import (
 	"fmt"
+	"reflect"
 	"strings"
 
 	"k8s.io/apimachinery/pkg/types"
@@ -10,38 +11,71 @@ import (
 	"github.com/weaveworks/weave/npc/iptables"
 )
 
-type ruleSpec struct {
-	key  string
-	args []string
+type ruleHost interface {
+	// getRuleSpec returns source or destination specification and comment which
+	// are used in an iptables rule.
+	//
+	// If src=true then the rulespec is for source, otherwise - for destination.
+	getRuleSpec(src bool) ([]string, string)
 }
 
-func newRuleSpec(proto *string, srcHost *selectorSpec, dstHost *selectorSpec, dstPort *string) *ruleSpec {
+type ruleSpec struct {
+	key        string
+	args       []string
+	policyType policyType
+}
+
+func newRuleSpec(policyType policyType, proto *string, srcHost, dstHost ruleHost, dstPort *string) *ruleSpec {
 	args := []string{}
 	if proto != nil {
 		args = append(args, "-p", *proto)
 	}
 	srcComment := "anywhere"
-	if srcHost != nil {
-		args = append(args, "-m", "set", "--match-set", string(srcHost.ipsetName), "src")
-		if srcHost.nsName != "" {
-			srcComment = fmt.Sprintf("pods: namespace: %s, selector: %s", srcHost.nsName, srcHost.key)
-		} else {
-			srcComment = fmt.Sprintf("namespaces: selector: %s", srcHost.key)
-		}
+	if srcHost != nil && !reflect.ValueOf(srcHost).IsNil() {
+		rule, comment := srcHost.getRuleSpec(true)
+		args = append(args, rule...)
+		srcComment = comment
 	}
 	dstComment := "anywhere"
-	if dstHost != nil {
-		args = append(args, "-m", "set", "--match-set", string(dstHost.ipsetName), "dst")
-		dstComment = fmt.Sprintf("pods: namespace: %s, selector: %s", dstHost.nsName, dstHost.key)
+	if dstHost != nil && !reflect.ValueOf(dstHost).IsNil() {
+		rule, comment := dstHost.getRuleSpec(false)
+		args = append(args, rule...)
+		dstComment = comment
 	}
 	if dstPort != nil {
 		args = append(args, "--dport", *dstPort)
 	}
-	args = append(args, "-j", "ACCEPT")
-	args = append(args, "-m", "comment", "--comment", fmt.Sprintf("%s -> %s", srcComment, dstComment))
+	// NOTE: if you remove the comment bellow, then embed `policyType` into `key`.
+	// Otherwise, the rule won't be provisioned if it exists for other policy type.
+	args = append(args, "-m", "comment", "--comment", fmt.Sprintf("%s -> %s (%s)", srcComment, dstComment, policyTypeStr(policyType)))
 	key := strings.Join(args, " ")
 
-	return &ruleSpec{key, args}
+	return &ruleSpec{key, args, policyType}
+}
+
+func (spec *ruleSpec) iptChain() string {
+	if spec.policyType == policyTypeEgress {
+		return EgressCustomChain
+	}
+	return IngressChain
+}
+
+func (spec *ruleSpec) iptRuleSpecs() [][]string {
+	if spec.policyType == policyTypeIngress {
+		rule := make([]string, len(spec.args))
+		copy(rule, spec.args)
+		rule = append(rule, "-j", "ACCEPT")
+		return [][]string{rule}
+	}
+
+	// policyTypeEgress
+	ruleMark := make([]string, len(spec.args))
+	copy(ruleMark, spec.args)
+	ruleMark = append(ruleMark, "-j", EgressMarkChain)
+	ruleReturn := make([]string, len(spec.args))
+	copy(ruleReturn, spec.args)
+	ruleReturn = append(ruleReturn, "-j", "RETURN")
+	return [][]string{ruleMark, ruleReturn}
 }
 
 type ruleSet struct {
@@ -58,10 +92,14 @@ func (rs *ruleSet) deprovision(user types.UID, current, desired map[string]*rule
 		if _, found := desired[key]; !found {
 			delete(rs.users[key], user)
 			if len(rs.users[key]) == 0 {
-				common.Log.Infof("deleting rule: %v", spec.args)
-				if err := rs.ipt.Delete(TableFilter, IngressChain, spec.args...); err != nil {
-					return err
+				chain := spec.iptChain()
+				for _, rule := range spec.iptRuleSpecs() {
+					common.Log.Infof("deleting rule %v from %q chain", rule, chain)
+					if err := rs.ipt.Delete(TableFilter, chain, rule...); err != nil {
+						return err
+					}
 				}
+
 				delete(rs.users, key)
 			}
 		}
@@ -74,9 +112,12 @@ func (rs *ruleSet) provision(user types.UID, current, desired map[string]*ruleSp
 	for key, spec := range desired {
 		if _, found := current[key]; !found {
 			if _, found := rs.users[key]; !found {
-				common.Log.Infof("adding rule: %v", spec.args)
-				if err := rs.ipt.Append(TableFilter, IngressChain, spec.args...); err != nil {
-					return err
+				chain := spec.iptChain()
+				for _, rule := range spec.iptRuleSpecs() {
+					common.Log.Infof("adding rule %v to %q chain", rule, chain)
+					if err := rs.ipt.Append(TableFilter, chain, rule...); err != nil {
+						return err
+					}
 				}
 				rs.users[key] = make(map[types.UID]struct{})
 			}

npc/rule_test.go

@@ -9,7 +9,7 @@ func TestRegressionHandleNil3095(t *testing.T) {
 	// Test for handling nil values
 	// https://github.com/weaveworks/weave/issues/3095
 
-	rs := newRuleSpec(nil, nil, nil, nil)
+	rs := newRuleSpec(policyTypeIngress, nil, nil, nil, nil)
 
 	require.NotEqual(t, rs.key, "")
 

npc/selector.go

@@ -1,34 +1,36 @@
 package npc
 
 import (
+	"fmt"
+
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/labels"
 	"k8s.io/apimachinery/pkg/types"
 
 	"github.com/weaveworks/weave/common"
-	"github.com/weaveworks/weave/npc/ipset"
+	"github.com/weaveworks/weave/net/ipset"
 )
 
 type selectorSpec struct {
-	key      string          // string representation (for hash keying/equality comparison)
-	selector labels.Selector // k8s Selector object (for matching)
-	dst      bool            // destination selector
+	key         string          // string representation (for hash keying/equality comparison)
+	selector    labels.Selector // k8s Selector object (for matching)
+	policyTypes []policyType    // If non-empty, then selectorSpec is a target selector for given policyTypes.
 
 	ipsetType ipset.Type // type of ipset to provision
 	ipsetName ipset.Name // generated ipset name
 	nsName    string     // Namespace name
 }
 
-func newSelectorSpec(json *metav1.LabelSelector, dst bool, nsName string, ipsetType ipset.Type) (*selectorSpec, error) {
+func newSelectorSpec(json *metav1.LabelSelector, policyType []policyType, nsName string, ipsetType ipset.Type) (*selectorSpec, error) {
 	selector, err := metav1.LabelSelectorAsSelector(json)
 	if err != nil {
 		return nil, err
 	}
 	key := selector.String()
 	return &selectorSpec{
-		key:      key,
-		selector: selector,
-		dst:      dst,
+		key:         key,
+		selector:    selector,
+		policyTypes: policyType,
 		// We prefix the selector string with the namespace name when generating
 		// the shortname because you can specify the same selector in multiple
 		// namespaces - we need those to map to distinct ipsets
@@ -37,6 +39,23 @@ func newSelectorSpec(json *metav1.LabelSelector, dst bool, nsName string, ipsetT
 		nsName:    nsName}, nil
 }
 
+func (spec *selectorSpec) getRuleSpec(src bool) ([]string, string) {
+	dir := "dst"
+	if src {
+		dir = "src"
+	}
+	rule := []string{"-m", "set", "--match-set", string(spec.ipsetName), dir}
+
+	comment := "anywhere"
+	if spec.nsName != "" {
+		comment = fmt.Sprintf("pods: namespace: %s, selector: %s", spec.nsName, spec.key)
+	} else {
+		comment = fmt.Sprintf("namespaces: selector: %s", spec.key)
+	}
+
+	return rule, comment
+}
+
 type selector struct {
 	ips  ipset.Interface
 	spec *selectorSpec
@@ -55,49 +74,54 @@ func (s *selector) delEntry(user types.UID, entry string) error {
 }
 
 type selectorFn func(selector *selector) error
+type selectorWithPolicyTypeFn func(selector *selector, policyType policyType) error
 
 type selectorSet struct {
 	ips           ipset.Interface
 	onNewSelector selectorFn
 
-	// invoked after dst selector has been provisioned for the first time
-	onNewDstSelector selectorFn
-	// invoked after the last instance of dst selector has been deprovisioned
-	onDestroyDstSelector selectorFn
+	// invoked after target selector has been provisioned for the first time
+	onNewTargetSelector selectorWithPolicyTypeFn
+	// invoked after the last instance of target selector has been deprovisioned
+	onDestroyTargetSelector selectorWithPolicyTypeFn
 
 	users   map[string]map[types.UID]struct{} // list of users per selector
 	entries map[string]*selector
 
-	// We need to keep track of dst selector instances to be able to invoke
-	// onNewDstSelector and onDestroyDstSelector callbacks at the right time;
-	// selectorSpec.Key -> count
-	dstSelectorsCount map[string]int
+	// We need to keep track of target selector instances to be able to invoke
+	// onNewTargetSelector and onDestroyTargetSelector callbacks at the right time;
+	// selectorSpec.Key -> policyType -> count
+	targetSelectorsCount map[string]map[policyType]int
 }
 
-func newSelectorSet(ips ipset.Interface, onNewSelector, onNewDstSelector selectorFn, onDestroyDstSelector selectorFn) *selectorSet {
+func newSelectorSet(ips ipset.Interface, onNewSelector selectorFn, onNewTargetSelector, onDestroyTargetSelector selectorWithPolicyTypeFn) *selectorSet {
 	return &selectorSet{
-		ips:                  ips,
-		onNewSelector:        onNewSelector,
-		onNewDstSelector:     onNewDstSelector,
-		onDestroyDstSelector: onDestroyDstSelector,
+		ips:                     ips,
+		onNewSelector:           onNewSelector,
+		onNewTargetSelector:     onNewTargetSelector,
+		onDestroyTargetSelector: onDestroyTargetSelector,
 		users:                make(map[string]map[types.UID]struct{}),
 		entries:              make(map[string]*selector),
-		dstSelectorsCount:    make(map[string]int)}
+		targetSelectorsCount: make(map[string]map[policyType]int)}
 }
 
-func (ss *selectorSet) addToMatching(user types.UID, labelMap map[string]string, entry string, comment string) (bool, error) {
-	found := false
+func (ss *selectorSet) addToMatching(user types.UID, labelMap map[string]string, entry string, comment string) (bool, bool, error) {
+	foundIngress := false
+	foundEgress := false
 	for _, s := range ss.entries {
 		if s.matches(labelMap) {
-			if ss.dstSelectorExist(s) {
-				found = true
+			if ss.targetSelectorExist(s, policyTypeIngress) {
+				foundIngress = true
+			}
+			if ss.targetSelectorExist(s, policyTypeEgress) {
+				foundEgress = true
 			}
 			if err := s.addEntry(user, entry, comment); err != nil {
-				return found, err
+				return foundIngress, foundEgress, err
 			}
 		}
 	}
-	return found, nil
+	return foundIngress, foundEgress, nil
 }
 
 func (ss *selectorSet) delFromMatching(user types.UID, labelMap map[string]string, entry string) error {
@@ -111,8 +135,8 @@ func (ss *selectorSet) delFromMatching(user types.UID, labelMap map[string]strin
 	return nil
 }
 
-func (ss *selectorSet) dstSelectorExist(s *selector) bool {
-	return ss.dstSelectorsCount[s.spec.key] > 0
+func (ss *selectorSet) targetSelectorExist(s *selector, policyType policyType) bool {
+	return ss.targetSelectorsCount[s.spec.key][policyType] > 0
 }
 
 func (ss *selectorSet) deprovision(user types.UID, current, desired map[string]*selectorSpec) error {
@@ -129,13 +153,14 @@ func (ss *selectorSet) deprovision(user types.UID, current, desired map[string]*
 				delete(ss.users, key)
 			}
 
-			if spec.dst {
-				ss.dstSelectorsCount[key]--
-				if ss.dstSelectorsCount[key] == 0 {
-					if err := ss.onDestroyDstSelector(&selector{ss.ips, spec}); err != nil {
+			for _, policyType := range spec.policyTypes {
+				ss.targetSelectorsCount[key][policyType]--
+				if ss.targetSelectorsCount[key][policyType] == 0 {
+					if err := ss.onDestroyTargetSelector(&selector{ss.ips, spec}, policyType); err != nil {
 						return err
 					}
 				}
+				// TODO(brb) delete(...)
 			}
 		}
 	}
@@ -160,10 +185,13 @@ func (ss *selectorSet) provision(user types.UID, current, desired map[string]*se
 			}
 			ss.users[key][user] = struct{}{}
 
-			if spec.dst {
-				ss.dstSelectorsCount[key]++
-				if ss.dstSelectorsCount[key] == 1 {
-					if err := ss.onNewDstSelector(selector); err != nil {
+			for _, pt := range spec.policyTypes {
+				if _, found := ss.targetSelectorsCount[key]; !found {
+					ss.targetSelectorsCount[key] = make(map[policyType]int)
+				}
+				ss.targetSelectorsCount[key][pt]++
+				if ss.targetSelectorsCount[key][pt] == 1 {
+					if err := ss.onNewTargetSelector(selector, pt); err != nil {
 						return err
 					}
 				}

npc/types.go

@@ -0,0 +1,15 @@
+package npc
+
+type policyType byte
+
+const (
+	policyTypeIngress policyType = iota
+	policyTypeEgress
+)
+
+func policyTypeStr(policyType policyType) string {
+	if policyType == policyTypeEgress {
+		return "egress"
+	}
+	return "ingress"
+}

plugin/ipam/driver.go

@@ -110,6 +110,10 @@ func (i *Ipam) ReleaseAddress(poolID string, address net.IP) error {
 	return i.weave.ReleaseIPsFor(address.String())
 }
 
+func (i *Ipam) IsBuiltIn() bool {
+	return false
+}
+
 // Functions required by ipamapi "contract" but not actually used.
 
 func (i *Ipam) DiscoverNew(discoverapi.DiscoveryType, interface{}) error {

plugin/net/cni.go

@@ -117,6 +117,16 @@ func (c *CNIPlugin) CmdAdd(args *skel.CmdArgs) error {
 	}
 	defer ns.Close()
 
+	hostNs, err := netns.Get()
+	if err != nil {
+		return fmt.Errorf("error accessing host network namespace: %s", err)
+	}
+	defer hostNs.Close()
+
+	if ns.Equal(hostNs) {
+		return fmt.Errorf("can not specify host network namespace as network namespace to which container should be added")
+	}
+
 	id := args.ContainerID
 	if len(id) < 5 {
 		data := make([]byte, 5)
@@ -130,7 +140,7 @@ func (c *CNIPlugin) CmdAdd(args *skel.CmdArgs) error {
 	if err := weavenet.AttachContainer(args.Netns, id, args.IfName, conf.BrName, conf.MTU, false, []*net.IPNet{&ip.Address}, false, conf.HairpinMode); err != nil {
 		return err
 	}
-	if err := weavenet.WithNetNSLinkUnsafe(ns, args.IfName, func(link netlink.Link) error {
+	if err := weavenet.WithNetNSLink(ns, args.IfName, func(link netlink.Link) error {
 		return setupRoutes(link, args.IfName, ip.Address, ip.Gateway, result.Routes)
 	}); err != nil {
 		return fmt.Errorf("error setting up routes: %s", err)
@@ -177,7 +187,13 @@ func (c *CNIPlugin) CmdDel(args *skel.CmdArgs) error {
 
 	// As of CNI 0.3 spec, runtimes can send blank if they just want the address deallocated
 	if args.Netns != "" {
-		if _, err = weavenet.WithNetNS(args.Netns, "del-iface", args.IfName); err != nil {
+		if err = weavenet.WithNetNSByPath(args.Netns, func() error {
+			link, err := netlink.LinkByName(args.IfName)
+			if err != nil {
+				return err
+			}
+			return netlink.LinkDel(link)
+		}); err != nil {
 			// We log the error and carry on instead of returning nil, as there may still be resources to free in IPAM.
 			logOnStderr(fmt.Errorf("error removing interface %q: %s", args.IfName, err))
 		}

plugin/net/watcher.go

@@ -49,7 +49,9 @@ func (w *watcher) ContainerStarted(id string) {
 				rootDir = "/host"
 			}
 			netNSPath := weavenet.NSPathByPidWithRoot(rootDir, info.State.Pid)
-			if _, err := weavenet.WithNetNS(netNSPath, "configure-arp", weavenet.VethName, rootDir); err != nil {
+			if err := weavenet.WithNetNSByPath(netNSPath, func() error {
+				return weavenet.ConfigureARP(weavenet.VethName, rootDir)
+			}); err != nil {
 				w.driver.warn("ContainerStarted", "unable to configure interfaces: %s", err)
 			}
 		}

prog/kube-peers/main.go

@@ -1,236 +0,0 @@
-/*
-Package main deals with weave-net peers on the cluster.
-
-This involves peer management, such as getting the latest peers or removing defunct peers from the cluster
-*/
-package main
-
-import (
-	"flag"
-	"fmt"
-	"net"
-	"os"
-
-	api "k8s.io/apimachinery/pkg/apis/meta/v1"
-	"k8s.io/client-go/kubernetes"
-	"k8s.io/client-go/rest"
-
-	weaveapi "github.com/weaveworks/weave/api"
-	"github.com/weaveworks/weave/common"
-)
-
-type nodeInfo struct {
-	name string
-	addr string
-}
-
-// return the IP addresses of all nodes in the cluster
-func getKubePeers(c *kubernetes.Clientset) ([]nodeInfo, error) {
-	nodeList, err := c.CoreV1().Nodes().List(api.ListOptions{})
-	if err != nil {
-		return nil, err
-	}
-	addresses := make([]nodeInfo, 0, len(nodeList.Items))
-	for _, peer := range nodeList.Items {
-		var internalIP, externalIP string
-		for _, addr := range peer.Status.Addresses {
-			// Check it's a valid ipv4 address
-			ip := net.ParseIP(addr.Address)
-			if ip == nil || ip.To4() == nil {
-				continue
-			}
-			if addr.Type == "InternalIP" {
-				internalIP = ip.To4().String()
-			} else if addr.Type == "ExternalIP" {
-				externalIP = ip.To4().String()
-			}
-		}
-
-		// Fallback for cases where a Node has an ExternalIP but no InternalIP
-		if internalIP != "" {
-			addresses = append(addresses, nodeInfo{name: peer.Name, addr: internalIP})
-		} else if externalIP != "" {
-			addresses = append(addresses, nodeInfo{name: peer.Name, addr: externalIP})
-		}
-	}
-	return addresses, nil
-}
-
-const (
-	configMapName      = "weave-net"
-	configMapNamespace = "kube-system"
-)
-
-// update the list of all peers that have gone through this code path
-func addMyselfToPeerList(cml *configMapAnnotations, c *kubernetes.Clientset, peerName, name string) (*peerList, error) {
-	var list *peerList
-	err := cml.LoopUpdate(func() error {
-		var err error
-		list, err = cml.GetPeerList()
-		if err != nil {
-			return err
-		}
-		if !list.contains(peerName) {
-			list.add(peerName, name)
-			err = cml.UpdatePeerList(*list)
-			if err != nil {
-				return err
-			}
-		}
-		return nil
-	})
-	return list, err
-}
-
-func checkIamInPeerList(cml *configMapAnnotations, c *kubernetes.Clientset, peerName string) (bool, error) {
-	if err := cml.Init(); err != nil {
-		return false, err
-	}
-	list, err := cml.GetPeerList()
-	if err != nil {
-		return false, err
-	}
-	common.Log.Debugf("[kube-peers] Checking peer %q against list %v", peerName, list)
-	return list.contains(peerName), nil
-}
-
-// For each of those peers that is no longer listed as a node by
-// Kubernetes, remove it from Weave IPAM
-func reclaimRemovedPeers(weave *weaveapi.Client, cml *configMapAnnotations, nodes []nodeInfo, myPeerName string) error {
-	for {
-		if err := cml.Init(); err != nil {
-			return err
-		}
-		// 1. Compare peers stored in the peerList against all peers reported by k8s now.
-		storedPeerList, err := cml.GetPeerList()
-		if err != nil {
-			return err
-		}
-		peerMap := make(map[string]peerInfo, len(storedPeerList.Peers))
-		for _, peer := range storedPeerList.Peers {
-			peerMap[peer.NodeName] = peer
-		}
-		for _, node := range nodes {
-			delete(peerMap, node.name)
-		}
-		common.Log.Debugln("[kube-peers] Nodes that have disappeared:", peerMap)
-		if len(peerMap) == 0 {
-			break
-		}
-		// 2. Loop for each X in the first set and not in the second - we wish to remove X from our data structures
-		for _, peer := range peerMap {
-			common.Log.Debugln("[kube-peers] Preparing to remove disappeared peer", peer)
-			okToRemove := false
-			// 3. Check if there is an existing annotation with key X
-			if existingAnnotation, found := cml.GetAnnotation(KubePeersPrefix + peer.PeerName); found {
-				common.Log.Debugln("[kube-peers] Existing annotation", existingAnnotation)
-				// 4.   If annotation already contains my identity, ok;
-				if existingAnnotation == myPeerName {
-					okToRemove = true
-				}
-			} else {
-				// 5.   If non-existent, write an annotation with key X and contents "my identity"
-				common.Log.Debugln("[kube-peers] Noting I plan to remove ", peer.PeerName)
-				if err := cml.UpdateAnnotation(KubePeersPrefix+peer.PeerName, myPeerName); err == nil {
-					okToRemove = true
-				} else {
-					common.Log.Debugln("[kube-peers] error from UpdateAnnotation: ", err)
-				}
-			}
-			if okToRemove {
-				// 6.   If step 4 or 5 succeeded, rmpeer X
-				result, err := weave.RmPeer(peer.PeerName)
-				common.Log.Infof("[kube-peers] rmpeer of %s: %s", peer.PeerName, result)
-				if err != nil {
-					return err
-				}
-				cml.LoopUpdate(func() error {
-					// 7aa.   Remove any annotations Z* that have contents X
-					if err := cml.RemoveAnnotationsWithValue(peer.PeerName); err != nil {
-						return err
-					}
-					// 7a.    Remove X from peerList
-					storedPeerList.remove(peer.PeerName)
-					if err := cml.UpdatePeerList(*storedPeerList); err != nil {
-						return err
-					}
-					// 7b.    Remove annotation with key X
-					return cml.RemoveAnnotation(KubePeersPrefix + peer.PeerName)
-				})
-			}
-			// 8.   If step 5 failed due to optimistic lock conflict, stop: someone else is handling X
-
-			// Step 3-5 is to protect against two simultaneous rmpeers of X
-			// Step 4 is to pick up again after a restart between step 5 and step 7b
-			// If the peer doing the reclaim disappears between steps 5 and 7a, then someone will clean it up in step 7aa
-			// If peer doing the reclaim disappears forever between 7a and 7b then we get a dangling annotation
-			// This should be sufficiently rare that we don't care.
-		}
-
-		// 9. Go back to step 1 until there is no difference between the two sets
-	}
-	// Question: Should we narrow step 2 by checking against Weave Net IPAM?
-	// i.e. If peer X owns any address space and is marked unreachable, we want to rmpeer X
-	return nil
-}
-
-func main() {
-	var (
-		justReclaim bool
-		justCheck   bool
-		peerName    string
-		nodeName    string
-		logLevel    string
-	)
-	flag.BoolVar(&justReclaim, "reclaim", false, "reclaim IP space from dead peers")
-	flag.BoolVar(&justCheck, "check-peer-new", false, "return success if peer name is not stored in annotation")
-	flag.StringVar(&peerName, "peer-name", "unknown", "name of this Weave Net peer")
-	flag.StringVar(&nodeName, "node-name", "unknown", "name of this Kubernetes node")
-	flag.StringVar(&logLevel, "log-level", "info", "logging level (debug, info, warning, error)")
-	flag.Parse()
-
-	common.SetLogLevel(logLevel)
-	config, err := rest.InClusterConfig()
-	if err != nil {
-		common.Log.Fatalf("[kube-peers] Could not get cluster config: %v", err)
-	}
-	c, err := kubernetes.NewForConfig(config)
-	if err != nil {
-		common.Log.Fatalf("[kube-peers] Could not make Kubernetes connection: %v", err)
-	}
-	if justCheck {
-		cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
-		exists, err := checkIamInPeerList(cml, c, peerName)
-		if err != nil {
-			common.Log.Fatalf("[kube-peers] Could not check peer list: %v", err)
-		}
-		if exists {
-			os.Exit(9)
-		} else {
-			os.Exit(0)
-		}
-	}
-	peers, err := getKubePeers(c)
-	if err != nil {
-		common.Log.Fatalf("[kube-peers] Could not get peers: %v", err)
-	}
-	if justReclaim {
-		cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
-
-		list, err := addMyselfToPeerList(cml, c, peerName, nodeName)
-		if err != nil {
-			common.Log.Fatalf("[kube-peers] Could not update peer list: %v", err)
-		}
-		common.Log.Infoln("[kube-peers] Added myself to peer list", list)
-
-		weave := weaveapi.NewClient(os.Getenv("WEAVE_HTTP_ADDR"), common.Log)
-		err = reclaimRemovedPeers(weave, cml, peers, peerName)
-		if err != nil {
-			common.Log.Fatalf("[kube-peers] Error while reclaiming space: %v", err)
-		}
-		return
-	}
-	for _, node := range peers {
-		fmt.Println(node.addr)
-	}
-}

prog/kube-utils/annotations.go

@@ -2,13 +2,10 @@
 In order to keep track of active weave peers, we use annotations on the Kubernetes cluster.
 
 Kubernetes uses etcd to distribute and synchronise these annotations so we don't have to.
-
-This module deals with operations on the peerlist backed by Kubernetes' annotation mechanism.
 */
 package main
 
 import (
-	"encoding/json"
 	"log"
 	"time"
 
@@ -29,43 +26,11 @@ type configMapAnnotations struct {
 	cm            *v1.ConfigMap
 }
 
-func newConfigMapAnnotations(ns string, configMapName string, clientset *kubernetes.Clientset) *configMapAnnotations {
+func newConfigMapAnnotations(ns string, configMapName string, c kubernetes.Interface) *configMapAnnotations {
 	return &configMapAnnotations{
 		Namespace:     ns,
 		ConfigMapName: configMapName,
-		Client:        clientset.CoreV1(),
-	}
-}
-
-type peerList struct {
-	Peers []peerInfo
-}
-
-type peerInfo struct {
-	PeerName string // Weave internal unique ID
-	NodeName string // Kubernetes node name
-}
-
-func (pl peerList) contains(peerName string) bool {
-	for _, peer := range pl.Peers {
-		if peer.PeerName == peerName {
-			return true
-		}
-	}
-	return false
-}
-
-func (pl *peerList) add(peerName string, name string) {
-	pl.Peers = append(pl.Peers, peerInfo{PeerName: peerName, NodeName: name})
-}
-
-func (pl *peerList) remove(peerNameToRemove string) {
-	for i := 0; i < len(pl.Peers); {
-		if pl.Peers[i].PeerName == peerNameToRemove {
-			pl.Peers = append(pl.Peers[:i], pl.Peers[i+1:]...)
-		} else {
-			i++
-		}
+		Client:        c.CoreV1(),
 	}
 }
 
@@ -111,27 +76,6 @@ func (cml *configMapAnnotations) Init() error {
 	return nil
 }
 
-func (cml *configMapAnnotations) GetPeerList() (*peerList, error) {
-	var record peerList
-	if cml.cm == nil {
-		return nil, errors.New("endpoint not initialized, call Init first")
-	}
-	if recordBytes, found := cml.cm.Annotations[KubePeersAnnotationKey]; found {
-		if err := json.Unmarshal([]byte(recordBytes), &record); err != nil {
-			return nil, err
-		}
-	}
-	return &record, nil
-}
-
-func (cml *configMapAnnotations) UpdatePeerList(list peerList) error {
-	recordBytes, err := json.Marshal(list)
-	if err != nil {
-		return err
-	}
-	return cml.UpdateAnnotation(KubePeersAnnotationKey, string(recordBytes))
-}
-
 // Clean up a string so it meets the Kubernetes requiremements for Annotation keys:
 // name part must consist of alphanumeric characters, '-', '_' or '.', and must
 // start and end with an alphanumeric character (e.g. 'MyName', or 'my.name', or '123-abc')

prog/kube-utils/kube_test.go

@@ -0,0 +1,30 @@
+// Adapted from k8s.io/client-go/kubernetes/fake to use a modified ObjectTracker
+package main
+
+import (
+	"k8s.io/apimachinery/pkg/runtime"
+	serializer "k8s.io/apimachinery/pkg/runtime/serializer"
+	"k8s.io/apimachinery/pkg/watch"
+	"k8s.io/client-go/kubernetes/fake"
+	"k8s.io/client-go/testing"
+
+	weavetracker "github.com/weaveworks/weave/testing/kubernetes/testing"
+)
+
+var scheme = runtime.NewScheme()
+var codecs = serializer.NewCodecFactory(scheme)
+
+func NewSimpleClientset(objects ...runtime.Object) *fake.Clientset {
+	o := weavetracker.NewObjectTracker(scheme, codecs.UniversalDecoder())
+	for _, obj := range objects {
+		if err := o.Add(obj); err != nil {
+			panic(err)
+		}
+	}
+
+	ret := fake.Clientset{}
+	ret.AddReactor("*", "*", weavetracker.ObjectReaction(o))
+	ret.AddWatchReactor("*", testing.DefaultWatchReactor(watch.NewFake(), nil))
+
+	return &ret
+}

prog/kube-utils/main.go

@@ -0,0 +1,254 @@
+/*
+Package main deals with weave-net peers on the cluster.
+
+This involves peer management, such as getting the latest peers or removing defunct peers from the cluster
+*/
+package main
+
+import (
+	"flag"
+	"fmt"
+	"net"
+	"os"
+
+	api "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/client-go/kubernetes"
+	"k8s.io/client-go/rest"
+
+	weaveapi "github.com/weaveworks/weave/api"
+	"github.com/weaveworks/weave/common"
+)
+
+type nodeInfo struct {
+	name string
+	addr string
+}
+
+// return the IP addresses of all nodes in the cluster
+func getKubePeers(c kubernetes.Interface, includeWithNoIPAddr bool) ([]nodeInfo, error) {
+	nodeList, err := c.CoreV1().Nodes().List(api.ListOptions{})
+	if err != nil {
+		return nil, err
+	}
+	addresses := make([]nodeInfo, 0, len(nodeList.Items))
+	for _, peer := range nodeList.Items {
+		var internalIP, externalIP string
+		for _, addr := range peer.Status.Addresses {
+			// Check it's a valid ipv4 address
+			ip := net.ParseIP(addr.Address)
+			if ip == nil || ip.To4() == nil {
+				continue
+			}
+			if addr.Type == "InternalIP" {
+				internalIP = ip.To4().String()
+			} else if addr.Type == "ExternalIP" {
+				externalIP = ip.To4().String()
+			}
+		}
+
+		// Fallback for cases where a Node has an ExternalIP but no InternalIP
+		if internalIP != "" {
+			addresses = append(addresses, nodeInfo{name: peer.Name, addr: internalIP})
+		} else if externalIP != "" {
+			addresses = append(addresses, nodeInfo{name: peer.Name, addr: externalIP})
+		} else if includeWithNoIPAddr {
+			addresses = append(addresses, nodeInfo{name: peer.Name, addr: ""})
+		}
+	}
+	return addresses, nil
+}
+
+// (minimal, incomplete) interface so weaver can be mocked for testing.
+type weaveClient interface {
+	RmPeer(peerName string) (string, error)
+}
+
+// For each of those peers that is no longer listed as a node by
+// Kubernetes, remove it from Weave IPAM
+func reclaimRemovedPeers(kube kubernetes.Interface, weave weaveClient, cml *configMapAnnotations, myPeerName, myNodeName string) error {
+	for loopsWhenNothingChanged := 0; loopsWhenNothingChanged < 3; loopsWhenNothingChanged++ {
+		if err := cml.Init(); err != nil {
+			return err
+		}
+		// 1. Compare peers stored in the peerList against all peers reported by k8s now.
+		storedPeerList, err := cml.GetPeerList()
+		if err != nil {
+			return err
+		}
+		nodes, err := getKubePeers(kube, true)
+		nodeSet := make(map[string]struct{}, len(nodes))
+		for _, node := range nodes {
+			nodeSet[node.name] = struct{}{}
+		}
+		peerMap := make(map[string]peerInfo, len(storedPeerList.Peers))
+		for _, peer := range storedPeerList.Peers {
+			peerMap[peer.PeerName] = peer
+		}
+		// remove entries from the peer map that are current nodes
+		for key, peer := range peerMap {
+			if _, found := nodeSet[peer.NodeName]; found {
+				// unless they have a duplicate of my NodeName but are not me
+				if peer.NodeName == myNodeName && peer.PeerName != myPeerName {
+					continue
+				}
+				delete(peerMap, key)
+			}
+		}
+		// so the remainder is everything we want to clean up
+		common.Log.Debugln("[kube-peers] Nodes that have disappeared:", peerMap)
+		if len(peerMap) == 0 {
+			break
+		}
+		// 2. Loop for each X in the first set and not in the second - we wish to remove X from our data structures
+		for _, peer := range peerMap {
+			if peer.PeerName == myPeerName { // Don't remove myself.
+				common.Log.Warnln("[kube-peers] not removing myself", peer)
+				continue
+			}
+			changed, err := reclaimPeer(weave, cml, peer.PeerName, myPeerName)
+			if err != nil {
+				return err
+			}
+			if changed {
+				loopsWhenNothingChanged = 0
+			}
+		}
+
+		// 9. Go back to step 1 until there is no difference between the two sets
+		// (or we hit the counter that says we've been round the loop 3 times and nothing happened)
+	}
+	return nil
+}
+
+// Attempt to reclaim the IP addresses owned by peerName, using the
+// Kubernetes api-server as a point of consensus so that only one peer
+// actions the reclaim.
+// Return a bool to show whether we attempted to change anything,
+// and an error if something went wrong.
+func reclaimPeer(weave weaveClient, cml *configMapAnnotations, peerName string, myPeerName string) (changed bool, err error) {
+	common.Log.Debugln("[kube-peers] Preparing to remove disappeared peer", peerName)
+	okToRemove := false
+	// 3. Check if there is an existing annotation with key X
+	if existingAnnotation, found := cml.GetAnnotation(KubePeersPrefix + peerName); found {
+		common.Log.Debugln("[kube-peers] Existing annotation", existingAnnotation)
+		// 4.   If annotation already contains my identity, ok;
+		if existingAnnotation == myPeerName {
+			okToRemove = true
+		}
+	} else {
+		// 5.   If non-existent, write an annotation with key X and contents "my identity"
+		common.Log.Debugln("[kube-peers] Noting I plan to remove ", peerName)
+		if err := cml.UpdateAnnotation(KubePeersPrefix+peerName, myPeerName); err == nil {
+			okToRemove = true
+		} else {
+			common.Log.Errorln("[kube-peers] error from UpdateAnnotation: ", err)
+		}
+	}
+	if !okToRemove {
+		return false, nil
+	}
+	// 6.   If step 4 or 5 succeeded, rmpeer X
+	result, err := weave.RmPeer(peerName)
+	common.Log.Infof("[kube-peers] rmpeer of %s: %s", peerName, result)
+	if err != nil {
+		return false, err
+	}
+	err = cml.LoopUpdate(func() error {
+		// 7aa.   Remove any annotations Z* that have contents X
+		if err := cml.RemoveAnnotationsWithValue(peerName); err != nil {
+			return err
+		}
+		// 7a.    Remove X from peerList
+		storedPeerList, err := cml.GetPeerList()
+		if err != nil {
+			return err
+		}
+		storedPeerList.remove(peerName)
+		if err := cml.UpdatePeerList(*storedPeerList); err != nil {
+			return err
+		}
+		// 7b.    Remove annotation with key X
+		return cml.RemoveAnnotation(KubePeersPrefix + peerName)
+	})
+	// 8.   If step 5 failed due to optimistic lock conflict, stop: someone else is handling X
+
+	// Step 3-5 is to protect against two simultaneous rmpeers of X
+	// Step 4 is to pick up again after a restart between step 5 and step 7b
+	// If the peer doing the reclaim disappears between steps 5 and 7a, then someone will clean it up in step 7aa
+	// If peer doing the reclaim disappears forever between 7a and 7b then we get a dangling annotation
+	// This should be sufficiently rare that we don't care.
+
+	// Question: Should we narrow step 2 by checking against Weave Net IPAM?
+	// i.e. If peer X owns any address space and is marked unreachable, we want to rmpeer X
+	return true, err
+}
+
+func main() {
+	var (
+		justReclaim       bool
+		justCheck         bool
+		justSetNodeStatus bool
+		peerName          string
+		nodeName          string
+		logLevel          string
+	)
+	flag.BoolVar(&justReclaim, "reclaim", false, "reclaim IP space from dead peers")
+	flag.BoolVar(&justCheck, "check-peer-new", false, "return success if peer name is not stored in annotation")
+	flag.BoolVar(&justSetNodeStatus, "set-node-status", false, "set NodeNetworkUnavailable to false")
+	flag.StringVar(&peerName, "peer-name", "unknown", "name of this Weave Net peer")
+	flag.StringVar(&nodeName, "node-name", "unknown", "name of this Kubernetes node")
+	flag.StringVar(&logLevel, "log-level", "info", "logging level (debug, info, warning, error)")
+	flag.Parse()
+
+	common.SetLogLevel(logLevel)
+	config, err := rest.InClusterConfig()
+	if err != nil {
+		common.Log.Fatalf("[kube-peers] Could not get cluster config: %v", err)
+	}
+	c, err := kubernetes.NewForConfig(config)
+	if err != nil {
+		common.Log.Fatalf("[kube-peers] Could not make Kubernetes connection: %v", err)
+	}
+	if justCheck {
+		cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
+		exists, err := checkIamInPeerList(cml, c, peerName)
+		if err != nil {
+			common.Log.Fatalf("[kube-peers] Could not check peer list: %v", err)
+		}
+		if exists {
+			os.Exit(9)
+		} else {
+			os.Exit(0)
+		}
+	}
+	if justSetNodeStatus {
+		err := setNodeNetworkUnavailableFalse(c, nodeName)
+		if err != nil {
+			common.Log.Fatalf("[kube-peers] could not set node status: %v", err)
+		}
+		return
+	}
+	if err != nil {
+		common.Log.Fatalf("[kube-peers] Could not get peers: %v", err)
+	}
+	if justReclaim {
+		cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
+
+		list, err := addMyselfToPeerList(cml, c, peerName, nodeName)
+		if err != nil {
+			common.Log.Fatalf("[kube-peers] Could not update peer list: %v", err)
+		}
+		common.Log.Infoln("[kube-peers] Added myself to peer list", list)
+
+		weave := weaveapi.NewClient(os.Getenv("WEAVE_HTTP_ADDR"), common.Log)
+		err = reclaimRemovedPeers(c, weave, cml, peerName, nodeName)
+		if err != nil {
+			common.Log.Fatalf("[kube-peers] Error while reclaiming space: %v", err)
+		}
+		return
+	}
+	peers, err := getKubePeers(c, false)
+	for _, node := range peers {
+		fmt.Println(node.addr)
+	}
+}

prog/kube-utils/nodestatus.go

@@ -0,0 +1,29 @@
+package main
+
+import (
+	"encoding/json"
+	"fmt"
+
+	"k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/client-go/kubernetes"
+)
+
+// based on k8s.io/kubernetes/pkg/controller/nodeipam/ipam/adapter.go
+func setNodeNetworkUnavailableFalse(c kubernetes.Interface, nodeName string) error {
+	condition := v1.NodeCondition{
+		Type:               v1.NodeNetworkUnavailable,
+		Status:             v1.ConditionFalse,
+		Reason:             "WeaveIsUp",
+		Message:            "Weave pod has set this",
+		LastTransitionTime: metav1.Now(),
+		LastHeartbeatTime:  metav1.Now(),
+	}
+	raw, err := json.Marshal(&[]v1.NodeCondition{condition})
+	if err != nil {
+		return err
+	}
+	patch := []byte(fmt.Sprintf(`{"status":{"conditions":%s}}`, raw))
+	_, err = c.CoreV1().Nodes().PatchStatus(nodeName, patch)
+	return err
+}

prog/kube-utils/peerlist.go

@@ -0,0 +1,104 @@
+/*
+This module deals with operations on the peerlist backed by Kubernetes' annotation mechanism.
+*/
+package main
+
+import (
+	"encoding/json"
+
+	"github.com/pkg/errors"
+	"k8s.io/client-go/kubernetes"
+
+	"github.com/weaveworks/weave/common"
+)
+
+const (
+	configMapName      = "weave-net"
+	configMapNamespace = "kube-system"
+)
+
+type peerList struct {
+	Peers []peerInfo
+}
+
+type peerInfo struct {
+	PeerName string // Weave internal unique ID
+	NodeName string // Kubernetes node name
+}
+
+func (pl peerList) contains(peerName string) bool {
+	for _, peer := range pl.Peers {
+		if peer.PeerName == peerName {
+			return true
+		}
+	}
+	return false
+}
+
+func (pl *peerList) add(peerName string, name string) {
+	pl.Peers = append(pl.Peers, peerInfo{PeerName: peerName, NodeName: name})
+}
+
+func (pl *peerList) remove(peerNameToRemove string) {
+	for i := 0; i < len(pl.Peers); {
+		if pl.Peers[i].PeerName == peerNameToRemove {
+			pl.Peers = append(pl.Peers[:i], pl.Peers[i+1:]...)
+		} else {
+			i++
+		}
+	}
+}
+
+func (cml *configMapAnnotations) GetPeerList() (*peerList, error) {
+	var record peerList
+	if cml.cm == nil {
+		return nil, errors.New("endpoint not initialized, call Init first")
+	}
+	if recordBytes, found := cml.cm.Annotations[KubePeersAnnotationKey]; found {
+		if err := json.Unmarshal([]byte(recordBytes), &record); err != nil {
+			return nil, err
+		}
+	}
+	return &record, nil
+}
+
+func (cml *configMapAnnotations) UpdatePeerList(list peerList) error {
+	recordBytes, err := json.Marshal(list)
+	if err != nil {
+		return err
+	}
+	return cml.UpdateAnnotation(KubePeersAnnotationKey, string(recordBytes))
+}
+
+// update the list of all peers that have gone through this code path
+func addMyselfToPeerList(cml *configMapAnnotations, c kubernetes.Interface, peerName, name string) (*peerList, error) {
+	var list *peerList
+	err := cml.LoopUpdate(func() error {
+		var err error
+		list, err = cml.GetPeerList()
+		if err != nil {
+			return err
+		}
+		if !list.contains(peerName) {
+			list.add(peerName, name)
+			err = cml.UpdatePeerList(*list)
+			if err != nil {
+				return err
+			}
+		}
+		return nil
+	})
+	return list, err
+}
+
+func checkIamInPeerList(cml *configMapAnnotations, c kubernetes.Interface, peerName string) (bool, error) {
+	if err := cml.Init(); err != nil {
+		return false, err
+	}
+	list, err := cml.GetPeerList()
+	if err != nil {
+		return false, err
+	}
+	common.Log.Debugf("[kube-peers] Checking peer %q against list %v", peerName, list)
+	return list.contains(peerName), nil
+}

prog/kube-utils/peerlist_test.go

@@ -0,0 +1,181 @@
+package main
+
+import (
+	"fmt"
+	"math/rand"
+	"sync"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+const (
+	peerName1 = "01:00:00:00:00:00"
+	peerName2 = "02:00:00:00:00:00"
+	peerName3 = "03:00:00:00:00:00"
+	nodeName1 = "fake-node-1"
+	nodeName2 = "fake-node-2"
+
+	testIters = 1000
+)
+
+func TestPeerListBasic(t *testing.T) {
+	c := NewSimpleClientset()
+
+	cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
+	err := cml.Init()
+	require.NoError(t, err)
+
+	list1, err := addMyselfToPeerList(cml, c, peerName1, nodeName1)
+	require.NoError(t, err)
+	require.Equal(t, 1, len(list1.Peers))
+	list2, err := addMyselfToPeerList(cml, c, peerName2, nodeName2)
+	require.NoError(t, err)
+	require.Equal(t, 2, len(list2.Peers))
+
+	check1, err := checkIamInPeerList(cml, c, peerName1)
+	require.NoError(t, err)
+	require.Equal(t, true, check1)
+	check3, err := checkIamInPeerList(cml, c, peerName3)
+	require.NoError(t, err)
+	require.Equal(t, false, check3)
+
+	storedPeerList, err := cml.GetPeerList()
+	require.NoError(t, err)
+
+	storedPeerList.remove(peerName1)
+	require.Equal(t, 1, len(storedPeerList.Peers))
+	err = cml.UpdatePeerList(*storedPeerList)
+	require.NoError(t, err)
+	check1, err = checkIamInPeerList(cml, c, peerName1)
+	require.NoError(t, err)
+	require.Equal(t, false, check1)
+}
+
+// Run function _f_ _iterations_ times, in _concurrency_
+// number of goroutines
+func doConcurrentIterations(iterations, concurrency int, f func(int)) {
+	iterationsPerThread := iterations / concurrency
+
+	wg := sync.WaitGroup{}
+	for i := 0; i < concurrency; i++ {
+		wg.Add(1)
+		go func(j int) {
+			defer wg.Done()
+			for k := 0; k < iterationsPerThread; k++ {
+				f((j * iterationsPerThread) + k)
+			}
+		}(i)
+	}
+	wg.Wait()
+}
+
+func TestPeerListFuzz(t *testing.T) {
+	const (
+		testIters   = 1000
+		maxNodes    = 100
+		concurrency = 5
+	)
+	peerName := func(i int) string { return fmt.Sprintf("%02d:00:00:02:00:00", i) }
+	nodeName := func(i int) string { return fmt.Sprintf("node%d", i) }
+
+	stateLock := sync.Mutex{}
+	nodes := []int{}
+	nodeMap := map[int]bool{} // true if an operation is pending
+
+	c := NewSimpleClientset()
+
+	addNode := func() {
+		i := rand.Intn(maxNodes)
+		// add to test state first, then to Kubernetes
+		stateLock.Lock()
+		pending, found := nodeMap[i]
+		if !found {
+			nodes = append(nodes, i)
+		}
+		nodeMap[i] = true
+		stateLock.Unlock()
+		if pending { // someone else is adding or deleting this node
+			return
+		}
+
+		cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
+		err := cml.Init()
+		require.NoError(t, err)
+		_, err = addMyselfToPeerList(cml, c, peerName(i), nodeName(i))
+		require.NoError(t, err)
+
+		stateLock.Lock()
+		nodeMap[i] = false
+		stateLock.Unlock()
+	}
+
+	removeNode := func() {
+		stateLock.Lock()
+		if len(nodes) == 0 { // can't remove if there aren't any nodes
+			stateLock.Unlock()
+			return
+		}
+		pos := rand.Intn(len(nodes))
+		i := nodes[pos]
+		if pending, found := nodeMap[i]; found {
+			if pending {
+				stateLock.Unlock()
+				return // someone else is adding or deleting this node
+			}
+			nodeMap[i] = true
+		} else {
+			t.Errorf("missing entry from node map: %d", i)
+		}
+		stateLock.Unlock()
+
+		// remove from Kubernetes first, then from test state
+		cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
+		require.NoError(t, cml.Init())
+		storedPeerList, err := cml.GetPeerList()
+		require.NoError(t, err)
+		found := storedPeerList.contains(peerName(i))
+		require.True(t, found, "peer %d not found in stored list", i)
+
+		_, err = reclaimPeer(mockWeave{}, cml, peerName(i), fmt.Sprintf("deleter-%d", i))
+		require.NoError(t, err)
+
+		storedPeerList, err = cml.GetPeerList()
+		require.NoError(t, err)
+		stateLock.Lock()
+		delete(nodeMap, i)
+		for pos = 0; pos < len(nodes); pos++ {
+			if nodes[pos] == i {
+				// Remove item from list by swapping it with last and reducing slice length by 1
+				nodes[pos] = nodes[len(nodes)-1]
+				nodes = nodes[:len(nodes)-1]
+				break
+			}
+		}
+		stateLock.Unlock()
+	}
+
+	doConcurrentIterations(testIters, concurrency, func(iteration int) {
+		r := rand.Float32()
+		switch {
+		case 0.0 <= r && r < 0.4:
+			addNode()
+
+		case (0.4 <= r && r < 0.8):
+			removeNode()
+		}
+	})
+
+	cml := newConfigMapAnnotations(configMapNamespace, configMapName, c)
+	err := cml.Init()
+	require.NoError(t, err)
+	storedPeerList, err := cml.GetPeerList()
+	require.NoError(t, err)
+	require.Equal(t, len(storedPeerList.Peers), len(nodes))
+}
+
+type mockWeave struct{}
+
+func (mockWeave) RmPeer(peerName string) (string, error) {
+	return "", nil
+}

prog/weave-kube/Dockerfile.template

@@ -1,3 +1,14 @@
 FROM DOCKERHUB_USER/weaveARCH_EXT
-ADD ./launch.sh ./kube-peers /home/weave/
+
+# These labels are pretty static, and can therefore be added early on:
+LABEL maintainer="Weaveworks <help@weave.works>" \
+      org.opencontainers.image.title="weave-kube" \
+      org.opencontainers.image.source="https://github.com/weaveworks/weave" \
+      org.opencontainers.image.vendor="Weaveworks"
+
+ADD ./launch.sh ./kube-utils /home/weave/
 ENTRYPOINT ["/home/weave/launch.sh"]
+
+# This label will change for every build, and should therefore be the last layer of the image:
+ARG revision
+LABEL org.opencontainers.image.revision="${revision}"

prog/weave-kube/launch.sh

@@ -27,9 +27,9 @@ xt_set_exists() {
 # Default if not supplied - same as weave net default
 IPALLOC_RANGE=${IPALLOC_RANGE:-10.32.0.0/12}
 HTTP_ADDR=${WEAVE_HTTP_ADDR:-127.0.0.1:6784}
-STATUS_ADDR=${WEAVE_STATUS_ADDR:-0.0.0.0:6782}
+METRICS_ADDR=${WEAVE_METRICS_ADDR:-0.0.0.0:6782}
 HOST_ROOT=${HOST_ROOT:-/host}
-CONN_LIMIT=${CONN_LIMIT:-30}
+CONN_LIMIT=${CONN_LIMIT:-100}
 DB_PREFIX=${DB_PREFIX:-/weavedb/weave-net}
 
 # Check if the IP range overlaps anything existing on the host
@@ -54,12 +54,23 @@ else
     fi
 fi
 
+STATUS_OPTS="--metrics-addr=$METRICS_ADDR"
+# --status-addr exposes internal information, so only turn it on if asked to.
+if [ -n "${WEAVE_STATUS_ADDR}" ]; then
+    STATUS_OPTS="$STATUS_OPTS --status-addr=$WEAVE_STATUS_ADDR"
+fi
+
 # Default is that npc will be running; allow caller to override
 WEAVE_NPC_OPTS="--expect-npc"
 if [ "${EXPECT_NPC}" = "0" ]; then
     WEAVE_NPC_OPTS=""
 fi
 
+NO_MASQ_LOCAL_OPT=""
+if [ -n "${NO_MASQ_LOCAL}" ]; then
+    NO_MASQ_LOCAL_OPT="--no-masq-local"
+fi
+
 # Kubernetes sets HOSTNAME to the host's hostname
 # when running a pod in host namespace.
 NICKNAME_ARG=""
@@ -74,7 +85,7 @@ router_bridge_opts() {
 }
 
 if [ -z "$KUBE_PEERS" ]; then
-    if ! KUBE_PEERS=$(/home/weave/kube-peers) || [ -z "$KUBE_PEERS" ]; then
+    if ! KUBE_PEERS=$(/home/weave/kube-utils) || [ -z "$KUBE_PEERS" ]; then
         echo Failed to get peers >&2
         exit 1
     fi
@@ -99,7 +110,7 @@ fi
 if [ -f ${DB_PREFIX}data.db ]; then
     if /usr/bin/weaveutil get-db-flag "$DB_PREFIX" peer-reclaim >/dev/null ; then
         # If this peer name is not stored in the list then we were removed by another peer.
-        if /home/weave/kube-peers -check-peer-new -peer-name="$PEERNAME" -log-level=debug ; then
+        if /home/weave/kube-utils -check-peer-new -peer-name="$PEERNAME" -log-level=debug ; then
             # In order to avoid a CRDT clash, clean up
             echo "Peer not in list; removing persisted data" >&2
             rm -f ${DB_PREFIX}data.db
@@ -133,10 +144,13 @@ post_start_actions() {
     /home/weave/weave --local setup-cni
 
     # Attempt to run the reclaim process, but don't halt the script if it fails
-    /home/weave/kube-peers -reclaim -node-name="$HOSTNAME" -peer-name="$PEERNAME" -log-level=debug || true
+    /home/weave/kube-utils -reclaim -node-name="$HOSTNAME" -peer-name="$PEERNAME" -log-level=debug || true
 
     # Expose the weave network so host processes can communicate with pods
     /home/weave/weave --local expose $WEAVE_EXPOSE_IP
+
+    # Mark network as up
+    /home/weave/kube-utils -set-node-status -node-name="$HOSTNAME"
 }
 
 post_start_actions &
@@ -144,11 +158,12 @@ post_start_actions &
 /home/weave/weaver $EXTRA_ARGS --port=6783 $(router_bridge_opts) \
      --name="$PEERNAME" \
      --host-root=$HOST_ROOT \
-     --http-addr=$HTTP_ADDR --status-addr=$STATUS_ADDR --docker-api='' --no-dns \
+     --http-addr=$HTTP_ADDR $STATUS_OPTS --docker-api='' --no-dns \
      --db-prefix="$DB_PREFIX" \
      --ipalloc-range=$IPALLOC_RANGE $NICKNAME_ARG \
      --ipalloc-init $IPALLOC_INIT \
      --conn-limit=$CONN_LIMIT \
      $WEAVE_NPC_OPTS \
+     $NO_MASQ_LOCAL_OPT \
      "$@" \
      $KUBE_PEERS

prog/weave-kube/weave-daemonset-k8s-1.7.yaml

@@ -25,14 +25,6 @@ items:
           - get
           - list
           - watch
-      - apiGroups:
-          - extensions
-        resources:
-          - networkpolicies
-        verbs:
-          - get
-          - list
-          - watch
       - apiGroups:
           - 'networking.k8s.io'
         resources:
@@ -41,6 +33,13 @@ items:
           - get
           - list
           - watch
+      - apiGroups:
+        - ''
+        resources:
+        - nodes/status
+        verbs:
+        - patch
+        - update
   - apiVersion: rbac.authorization.k8s.io/v1beta1
     kind: ClusterRoleBinding
     metadata:
@@ -101,6 +100,8 @@ items:
         name: weave-net
       namespace: kube-system
     spec:
+      # Wait 5 seconds to let pod connect before rolling next pod
+      minReadySeconds: 5
       template:
         metadata:
           labels:
@@ -154,7 +155,6 @@ items:
                       fieldPath: spec.nodeName
               image: 'weaveworks/weave-npc:latest'
               imagePullPolicy: Always
-#npc-args
               resources:
                 requests:
                   cpu: 10m

prog/weave-kube/weave-daemonset-k8s-1.8.yaml

@@ -0,0 +1,209 @@
+apiVersion: v1
+kind: List
+items:
+  - apiVersion: v1
+    kind: ServiceAccount
+    metadata:
+      name: weave-net
+      labels:
+        name: weave-net
+      namespace: kube-system
+  - apiVersion: rbac.authorization.k8s.io/v1beta1
+    kind: ClusterRole
+    metadata:
+      name: weave-net
+      labels:
+        name: weave-net
+    rules:
+      - apiGroups:
+          - ''
+        resources:
+          - pods
+          - namespaces
+          - nodes
+        verbs:
+          - get
+          - list
+          - watch
+      - apiGroups:
+          - extensions
+        resources:
+          - networkpolicies
+        verbs:
+          - get
+          - list
+          - watch
+      - apiGroups:
+          - 'networking.k8s.io'
+        resources:
+          - networkpolicies
+        verbs:
+          - get
+          - list
+          - watch
+      - apiGroups:
+        - ''
+        resources:
+        - nodes/status
+        verbs:
+        - patch
+        - update
+  - apiVersion: rbac.authorization.k8s.io/v1beta1
+    kind: ClusterRoleBinding
+    metadata:
+      name: weave-net
+      labels:
+        name: weave-net
+    roleRef:
+      kind: ClusterRole
+      name: weave-net
+      apiGroup: rbac.authorization.k8s.io
+    subjects:
+      - kind: ServiceAccount
+        name: weave-net
+        namespace: kube-system
+  - apiVersion: rbac.authorization.k8s.io/v1beta1
+    kind: Role
+    metadata:
+      name: weave-net
+      namespace: kube-system
+      labels:
+        name: weave-net
+    rules:
+      - apiGroups:
+          - ''
+        resources:
+          - configmaps
+        resourceNames:
+          - weave-net
+        verbs:
+          - get
+          - update
+      - apiGroups:
+          - ''
+        resources:
+          - configmaps
+        verbs:
+          - create
+  - apiVersion: rbac.authorization.k8s.io/v1beta1
+    kind: RoleBinding
+    metadata:
+      name: weave-net
+      namespace: kube-system
+      labels:
+        name: weave-net
+    roleRef:
+      kind: Role
+      name: weave-net
+      apiGroup: rbac.authorization.k8s.io
+    subjects:
+      - kind: ServiceAccount
+        name: weave-net
+        namespace: kube-system
+  - apiVersion: extensions/v1beta1
+    kind: DaemonSet
+    metadata:
+      name: weave-net
+      labels:
+        name: weave-net
+      namespace: kube-system
+    spec:
+      # Wait 5 seconds to let pod connect before rolling next pod
+      minReadySeconds: 5
+      template:
+        metadata:
+          labels:
+            name: weave-net
+        spec:
+          containers:
+            - name: weave
+              command:
+                - /home/weave/launch.sh
+              env:
+                - name: HOSTNAME
+                  valueFrom:
+                    fieldRef:
+                      apiVersion: v1
+                      fieldPath: spec.nodeName
+              image: 'weaveworks/weave-kube:latest'
+              imagePullPolicy: Always
+              livenessProbe:
+                httpGet:
+                  host: 127.0.0.1
+                  path: /status
+                  port: 6784
+                initialDelaySeconds: 30
+              resources:
+                requests:
+                  cpu: 10m
+              securityContext:
+                privileged: true
+              volumeMounts:
+                - name: weavedb
+                  mountPath: /weavedb
+                - name: cni-bin
+                  mountPath: /host/opt
+                - name: cni-bin2
+                  mountPath: /host/home
+                - name: cni-conf
+                  mountPath: /host/etc
+                - name: dbus
+                  mountPath: /host/var/lib/dbus
+                - name: lib-modules
+                  mountPath: /lib/modules
+                - name: xtables-lock
+                  mountPath: /run/xtables.lock
+                  readOnly: false
+            - name: weave-npc
+              env:
+                - name: HOSTNAME
+                  valueFrom:
+                    fieldRef:
+                      apiVersion: v1
+                      fieldPath: spec.nodeName
+              image: 'weaveworks/weave-npc:latest'
+              imagePullPolicy: Always
+#npc-args
+              resources:
+                requests:
+                  cpu: 10m
+              securityContext:
+                privileged: true
+              volumeMounts:
+                - name: xtables-lock
+                  mountPath: /run/xtables.lock
+                  readOnly: false
+          hostNetwork: true
+          hostPID: true
+          restartPolicy: Always
+          securityContext:
+            seLinuxOptions: {}
+          serviceAccountName: weave-net
+          tolerations:
+            - effect: NoSchedule
+              operator: Exists
+          volumes:
+            - name: weavedb
+              hostPath:
+                path: /var/lib/weave
+            - name: cni-bin
+              hostPath:
+                path: /opt
+            - name: cni-bin2
+              hostPath:
+                path: /home
+            - name: cni-conf
+              hostPath:
+                path: /etc
+            - name: dbus
+              hostPath:
+                path: /var/lib/dbus
+            - name: lib-modules
+              hostPath:
+                path: /lib/modules
+            - name: xtables-lock
+              hostPath:
+                path: /run/xtables.lock
+                type: FileOrCreate
+      updateStrategy:
+        type: RollingUpdate

prog/weave-npc/Dockerfile.template

@@ -1,17 +1,24 @@
 FROM ALPINE_BASEIMAGE
 
-# If we're building for another architecture than amd64, the CROSS_BUILD_ placeholder is removed so e.g. CROSS_BUILD_COPY turns into COPY
-# If we're building normally, for amd64, CROSS_BUILD lines are removed
-CROSS_BUILD_COPY qemu-QEMUARCH-static /usr/bin/
-
-LABEL maintainer "Weaveworks Inc <help@weave.works>"
-LABEL works.weave.role=system \
+# These labels are pretty static, and can therefore be added early on:
+LABEL works.weave.role="system" \
+      maintainer="Weaveworks <help@weave.works>" \
+      org.opencontainers.image.title="Weave Net" \
+      org.opencontainers.image.description="Weave Net creates a virtual network that connects Docker containers across multiple hosts and enables their automatic discovery" \
+      org.opencontainers.image.url="https://weave.works" \
+      org.opencontainers.image.source="https://github.com/weaveworks/weave" \
+      org.opencontainers.image.vendor="Weaveworks" \
+      org.label-schema.schema-version="1.0" \
       org.label-schema.name="Weave Net" \
       org.label-schema.description="Weave Net creates a virtual network that connects Docker containers across multiple hosts and enables their automatic discovery" \
       org.label-schema.url="https://weave.works" \
       org.label-schema.vcs-url="https://github.com/weaveworks/weave" \
       org.label-schema.vendor="Weaveworks"
 
+# If we're building for another architecture than amd64, the CROSS_BUILD_ placeholder is removed so e.g. CROSS_BUILD_COPY turns into COPY
+# If we're building normally, for amd64, CROSS_BUILD lines are removed
+CROSS_BUILD_COPY qemu-QEMUARCH-static /usr/bin/
+
 RUN apk add --update \
     iptables \
     ipset \
@@ -21,3 +28,8 @@ RUN apk add --update \
 COPY ./weave-npc /usr/bin/weave-npc
 COPY ./ulogd.conf /etc/ulogd.conf
 ENTRYPOINT ["/usr/bin/weave-npc"]
+
+# These labels will change for every build, and should therefore be the last layer of the image:
+ARG revision
+LABEL org.opencontainers.image.revision="${revision}" \
+      org.label-schema.vcs-ref="${revision}"

prog/weave-npc/main.go

@@ -8,7 +8,6 @@ import (
 	"github.com/coreos/go-iptables/iptables"
 	"github.com/spf13/cobra"
 	coreapi "k8s.io/api/core/v1"
-	extnapi "k8s.io/api/extensions/v1beta1"
 	networkingv1 "k8s.io/api/networking/v1"
 	"k8s.io/apimachinery/pkg/fields"
 	"k8s.io/apimachinery/pkg/runtime"
@@ -18,8 +17,9 @@ import (
 	"k8s.io/client-go/tools/cache"
 
 	"github.com/weaveworks/weave/common"
+	"github.com/weaveworks/weave/net"
+	"github.com/weaveworks/weave/net/ipset"
 	"github.com/weaveworks/weave/npc"
-	"github.com/weaveworks/weave/npc/ipset"
 	"github.com/weaveworks/weave/npc/metrics"
 	"github.com/weaveworks/weave/npc/ulogd"
 )
@@ -30,7 +30,7 @@ var (
 	logLevel    string
 	allowMcast  bool
 	nodeName    string
-	legacy      bool
+	maxList     int
 )
 
 func handleError(err error) { common.CheckFatal(err) }
@@ -52,11 +52,31 @@ func resetIPTables(ipt *iptables.IPTables) error {
 		return err
 	}
 
-	return ipt.ClearChain(npc.TableFilter, npc.MainChain)
+	if err := ipt.ClearChain(npc.TableFilter, npc.MainChain); err != nil {
+		return err
+	}
+
+	if err := ipt.ClearChain(npc.TableFilter, npc.EgressMarkChain); err != nil {
+		return err
+	}
+
+	if err := ipt.ClearChain(npc.TableFilter, npc.EgressCustomChain); err != nil {
+		return err
+	}
+
+	if err := ipt.ClearChain(npc.TableFilter, npc.EgressDefaultChain); err != nil {
+		return err
+	}
+
+	// We do not clear npc.EgressChain here because otherwise, in the case of restarting
+	// weave-npc process, all egress traffic is allowed for a short period of time.
+	// The chain is created in createBaseRules.
+
+	return nil
 }
 
 func resetIPSets(ips ipset.Interface) error {
-	// Remove ipsets prefixed `weave-` only
+	// Remove ipsets prefixed `weave-` only.
 
 	sets, err := ips.List(npc.IpsetNamePrefix)
 	if err != nil {
@@ -76,6 +96,11 @@ func resetIPSets(ips ipset.Interface) error {
 	}
 
 	for _, s := range sets {
+		// LocalIPset might be used by WEAVE-NPC-EGRESS chain which we do not
+		// flush, so we cannot destroy it.
+		if s == npc.LocalIpset {
+			continue
+		}
 		common.Log.Debugf("Destroying ipset '%s'", string(s))
 		if err := ips.Destroy(s); err != nil {
 			common.Log.Errorf("Failed to destroy ipset '%s'", string(s))
@@ -111,11 +136,81 @@ func createBaseRules(ipt *iptables.IPTables, ips ipset.Interface) error {
 	}
 
 	// If the destination address is not any of the local pods, let it through
-	if err := ips.Create(npc.LocalIpset, ipset.HashIP); err != nil {
+	found, err := ipsetExist(ips, npc.LocalIpset)
+	if err != nil {
 		return err
 	}
-	return ipt.Append(npc.TableFilter, npc.MainChain,
-		"-m", "set", "!", "--match-set", npc.LocalIpset, "dst", "-j", "ACCEPT")
+	if !found {
+		if err := ips.Create(npc.LocalIpset, ipset.HashIP); err != nil {
+			return err
+		}
+	}
+	if err := ipt.Append(npc.TableFilter, npc.MainChain,
+		"-m", "set", "!", "--match-set", npc.LocalIpset, "dst", "-j", "ACCEPT"); err != nil {
+		return err
+	}
+
+	if err := ipt.Append(npc.TableFilter, npc.EgressMarkChain,
+		"-j", "MARK", "--set-xmark", npc.EgressMark); err != nil {
+		return err
+	}
+
+	// Egress rules:
+	//
+	// -A WEAVE-NPC-EGRESS -m state --state RELATED,ESTABLISHED -j ACCEPT
+	// -A WEAVE-NPC-EGRESS -m set ! --match-set weave-local-pods src -j RETURN
+	// -A WEAVE-NPC-EGRESS -d 224.0.0.0/4 -j RETURN
+	// -A WEAVE-NPC-EGRESS -m state --state NEW -j WEAVE-NPC-EGRESS-DEFAULT
+	// -A WEAVE-NPC-EGRESS -m state --state NEW -m mark ! --mark 0x40000/0x40000 -j WEAVE-NPC-EGRESS-CUSTOM
+	// -A WEAVE-NPC-EGRESS -m state --state NEW -m mark ! --mark 0x40000/0x40000 -j NFLOG --nflog-group 86
+	// -A WEAVE-NPC-EGRESS -m mark ! --mark 0x40000/0x40000 -j DROP
+	//
+	// -A WEAVE-NPC-EGRESS-CUSTOM <rulespec> -j MARK --set-xmark 0x40000/0x40000
+	// -A WEAVE-NPC-EGRESS-CUSTOM <rulespec> -j RETURN
+	//
+	// -A WEAVE-NPC-EGRESS-DEFAULT <rulespec> -j MARK --set-xmark 0x40000/0x40000
+	// -A WEAVE-NPC-EGRESS-DEFAULT <rulespec> -j RETURN
+	//
+	// For each rule we create two (mark and return). We cannot just accept
+	// a packet if it matches any rule, as a packet might need to traverse
+	// the ingress npc as well which happens later in the chain (in some cases
+	// we cannot detect whether packet is ingress or egress, so we need to
+	// check both chains).
+
+	ruleSpecs := [][]string{
+		{"-m", "state", "--state", "RELATED,ESTABLISHED", "-j", "ACCEPT"},
+		{"-m", "state", "--state", "NEW", "-m", "set", "!", "--match-set", npc.LocalIpset, "src", "-j", "RETURN"},
+	}
+	if allowMcast {
+		ruleSpecs = append(ruleSpecs, []string{"-d", "224.0.0.0/4", "-j", "RETURN"})
+	}
+	ruleSpecs = append(ruleSpecs, [][]string{
+		{"-m", "state", "--state", "NEW", "-j", string(npc.EgressDefaultChain)},
+		{"-m", "state", "--state", "NEW", "-m", "mark", "!", "--mark", npc.EgressMark, "-j", string(npc.EgressCustomChain)},
+		{"-m", "state", "--state", "NEW", "-m", "mark", "!", "--mark", npc.EgressMark, "-j", "NFLOG", "--nflog-group", "86"},
+		{"-m", "mark", "!", "--mark", npc.EgressMark, "-j", "DROP"},
+	}...)
+	if err := net.AddChainWithRules(ipt, npc.TableFilter, npc.EgressChain, ruleSpecs); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+// Dummy way to check whether a given ipset exists.
+// TODO(brb) Use "ipset -exist create <..>" for our purpose instead (for some reasons
+// creating an ipset with -exist fails).
+func ipsetExist(ips ipset.Interface, name ipset.Name) (bool, error) {
+	sets, err := ips.List(string(name))
+	if err != nil {
+		return false, err
+	}
+	for _, s := range sets {
+		if s == name {
+			return true, nil
+		}
+	}
+	return false, nil
 }
 
 func root(cmd *cobra.Command, args []string) {
@@ -131,10 +226,6 @@ func root(cmd *cobra.Command, args []string) {
 	}
 	common.Log.Infof("Starting Weaveworks NPC %s; node name %q", version, nodeName)
 
-	if legacy {
-		common.Log.Info("Running in legacy mode (k8s pre-1.7 network policy semantics)")
-	}
-
 	if err := metrics.Start(metricsAddr); err != nil {
 		common.Log.Fatalf("Failed to start metrics: %v", err)
 	}
@@ -152,13 +243,13 @@ func root(cmd *cobra.Command, args []string) {
 	ipt, err := iptables.New()
 	handleError(err)
 
-	ips := ipset.New(common.LogLogger())
+	ips := ipset.New(common.LogLogger(), maxList)
 
 	handleError(resetIPTables(ipt))
 	handleError(resetIPSets(ips))
 	handleError(createBaseRules(ipt, ips))
 
-	npc := npc.New(nodeName, legacy, ipt, ips)
+	npc := npc.New(nodeName, ipt, ips)
 
 	nsController := makeController(client.Core().RESTClient(), "namespaces", &coreapi.Namespace{},
 		cache.ResourceEventHandlerFuncs{
@@ -219,11 +310,7 @@ func root(cmd *cobra.Command, args []string) {
 			handleError(npc.UpdateNetworkPolicy(old, new))
 		},
 	}
-	if legacy {
-		npController = makeController(client.Extensions().RESTClient(), "networkpolicies", &extnapi.NetworkPolicy{}, npHandlers)
-	} else {
-		npController = makeController(client.NetworkingV1().RESTClient(), "networkpolicies", &networkingv1.NetworkPolicy{}, npHandlers)
-	}
+	npController = makeController(client.NetworkingV1().RESTClient(), "networkpolicies", &networkingv1.NetworkPolicy{}, npHandlers)
 
 	go nsController.Run(wait.NeverStop)
 	go podController.Run(wait.NeverStop)
@@ -244,7 +331,7 @@ func main() {
 	rootCmd.PersistentFlags().StringVar(&logLevel, "log-level", "debug", "logging level (debug, info, warning, error)")
 	rootCmd.PersistentFlags().BoolVar(&allowMcast, "allow-mcast", true, "allow all multicast traffic")
 	rootCmd.PersistentFlags().StringVar(&nodeName, "node-name", "", "only generate rules that apply to this node")
-	rootCmd.PersistentFlags().BoolVar(&legacy, "use-legacy-netpol", false, "use legacy network policies (pre k8s 1.7 vsn)")
+	rootCmd.PersistentFlags().IntVar(&maxList, "max-list-size", 1024, "maximum size of ipset list (for namespaces)")
 
 	handleError(rootCmd.Execute())
 }

prog/weavedb/Dockerfile

@@ -1,7 +1,17 @@
 # This is a nearly-empty image that we use to create a data-only container for persistence
 FROM scratch
-LABEL maintainer "Weaveworks Inc <help@weave.works>"
-LABEL works.weave.role=system
+
+# These labels are pretty static, and can therefore be added early on:
+LABEL works.weave.role="system" \
+      maintainer="Weaveworks <help@weave.works>" \
+      org.opencontainers.image.title="weavedb" \
+      org.opencontainers.image.source="https://github.com/weaveworks/weave" \
+      org.opencontainers.image.vendor="Weaveworks"
+
 ENTRYPOINT ["data-only"]
 # Work round Docker refusing to save an empty image
 COPY Dockerfile /
+
+# This label will change for every build, and should therefore be the last layer of the image:
+ARG revision
+LABEL org.opencontainers.image.revision="${revision}"

prog/weaveexec/Dockerfile.template

@@ -1,5 +1,11 @@
 FROM DOCKERHUB_USER/weaveARCH_EXT
 
+# These labels are pretty static, and can therefore be added early on:
+LABEL maintainer="Weaveworks <help@weave.works>" \
+      org.opencontainers.image.title="weaveexec" \
+      org.opencontainers.image.source="https://github.com/weaveworks/weave" \
+      org.opencontainers.image.vendor="Weaveworks"
+
 ENTRYPOINT ["/home/weave/sigproxy", "/home/weave/weave"]
 
 ADD ./sigproxy ./symlink /home/weave/
@@ -7,3 +13,7 @@ ADD ./weavewait /w/w
 ADD ./weavewait_noop /w-noop/w
 ADD ./weavewait_nomcast /w-nomcast/w
 WORKDIR /home/weave
+
+# This label will change for every build, and should therefore be the last layer of the image:
+ARG revision
+LABEL org.opencontainers.image.revision="${revision}"

prog/weaver/Dockerfile.template

@@ -1,17 +1,24 @@
 FROM ALPINE_BASEIMAGE
 
-# If we're building for another architecture than amd64, the CROSS_BUILD_ placeholder is removed so e.g. CROSS_BUILD_COPY turns into COPY
-# If we're building normally, for amd64, CROSS_BUILD lines are removed
-CROSS_BUILD_COPY qemu-QEMUARCH-static /usr/bin/
-
-LABEL maintainer "Weaveworks Inc <help@weave.works>"
-LABEL works.weave.role=system \
+# These labels are pretty static, and can therefore be added early on:
+LABEL works.weave.role="system" \
+      maintainer="Weaveworks <help@weave.works>" \
+      org.opencontainers.image.title="Weave Net" \
+      org.opencontainers.image.description="Weave Net creates a virtual network that connects Docker containers across multiple hosts and enables their automatic discovery" \
+      org.opencontainers.image.url="https://weave.works" \
+      org.opencontainers.image.source="https://github.com/weaveworks/weave" \
+      org.opencontainers.image.vendor="Weaveworks" \
+      org.label-schema.schema-version="1.0" \
       org.label-schema.name="Weave Net" \
       org.label-schema.description="Weave Net creates a virtual network that connects Docker containers across multiple hosts and enables their automatic discovery" \
       org.label-schema.url="https://weave.works" \
       org.label-schema.vcs-url="https://github.com/weaveworks/weave" \
       org.label-schema.vendor="Weaveworks"
 
+# If we're building for another architecture than amd64, the CROSS_BUILD_ placeholder is removed so e.g. CROSS_BUILD_COPY turns into COPY
+# If we're building normally, for amd64, CROSS_BUILD lines are removed
+CROSS_BUILD_COPY qemu-QEMUARCH-static /usr/bin/
+
 RUN apk add --update \
     curl \
     ethtool \
@@ -21,11 +28,16 @@ RUN apk add --update \
     util-linux \
     conntrack-tools \
     bind-tools \
+    ca-certificates \
   && rm -rf /var/cache/apk/*
 
 ADD ./weave ./weaver /home/weave/
 ADD ./weaveutil /usr/bin/
 ADD weavedata.db /weavedb/
-COPY ca-certificates.crt /etc/ssl/certs/
 ENTRYPOINT ["/home/weave/weaver"]
 WORKDIR /home/weave
+
+# These labels will change for every build, and should therefore be the last layer of the image:
+ARG revision
+LABEL org.opencontainers.image.revision="${revision}" \
+      org.label-schema.vcs-ref="${revision}"

prog/weaver/http.go

@@ -329,7 +329,7 @@ func HandleHTTP(muxRouter *mux.Router, version string, router *weave.NetworkRout
 			w.Write(json)
 		})
 
-	muxRouter.Methods("GET").Path("/report").Queries("format", "{format}").HandlerFunc(
+	muxRouter.Methods("GET").Path("/report").HandlerFunc(
 		func(w http.ResponseWriter, r *http.Request) {
 			funcs := template.FuncMap{
 				"json": func(v interface{}) string {
@@ -337,7 +337,12 @@ func HandleHTTP(muxRouter *mux.Router, version string, router *weave.NetworkRout
 					return string(a)
 				},
 			}
-			formatTemplate, err := template.New("format").Funcs(funcs).Parse(mux.Vars(r)["format"])
+			format, ok := r.URL.Query()["format"]
+			if !ok || len(format) != 1 {
+				http.Error(w, "error: must supply one format parameter", http.StatusBadRequest)
+				return
+			}
+			formatTemplate, err := template.New("format").Funcs(funcs).Parse(format[0])
 			if err != nil {
 				http.Error(w, err.Error(), http.StatusBadRequest)
 				return

prog/weaver/main.go

@@ -26,6 +26,7 @@ import (
 	"github.com/weaveworks/weave/nameserver"
 	weavenet "github.com/weaveworks/weave/net"
 	"github.com/weaveworks/weave/net/address"
+	"github.com/weaveworks/weave/net/ipset"
 	"github.com/weaveworks/weave/plugin"
 	weaveproxy "github.com/weaveworks/weave/proxy"
 	weave "github.com/weaveworks/weave/router"
@@ -155,6 +156,7 @@ func main() {
 		noDiscovery        bool
 		httpAddr           string
 		statusAddr         string
+		metricsAddr        string
 		ipamConfig         ipamConfig
 		dockerAPI          string
 		peers              []string
@@ -186,12 +188,13 @@ func main() {
 	mflag.StringVar(&logLevel, []string{"-log-level"}, "info", "logging level (debug, info, warning, error)")
 	mflag.BoolVar(&pktdebug, []string{"-pkt-debug"}, false, "enable per-packet debug logging")
 	mflag.StringVar(&prof, []string{"-profile"}, "", "enable profiling and write profiles to given path")
-	mflag.IntVar(&config.ConnLimit, []string{"-conn-limit"}, 30, "connection limit (0 for unlimited)")
+	mflag.IntVar(&config.ConnLimit, []string{"-conn-limit"}, 100, "connection limit (0 for unlimited)")
 	mflag.BoolVar(&noDiscovery, []string{"-no-discovery"}, false, "disable peer discovery")
 	mflag.IntVar(&bufSzMB, []string{"-bufsz"}, 8, "capture buffer size in MB")
 	mflag.IntVar(&bridgeConfig.MTU, []string{"-mtu"}, 0, "MTU size")
 	mflag.StringVar(&httpAddr, []string{"-http-addr"}, "", "address to bind HTTP interface to (disabled if blank, absolute path indicates unix domain socket)")
 	mflag.StringVar(&statusAddr, []string{"-status-addr"}, "", "address to bind status+metrics interface to (disabled if blank, absolute path indicates unix domain socket)")
+	mflag.StringVar(&metricsAddr, []string{"-metrics-addr"}, "", "address to bind metrics interface to (disabled if blank, absolute path indicates unix domain socket)")
 	mflag.StringVar(&ipamConfig.Mode, []string{"-ipalloc-init"}, "", "allocator initialisation strategy (consensus, seed or observer)")
 	mflag.StringVar(&ipamConfig.IPRangeCIDR, []string{"-ipalloc-range"}, "", "IP address range reserved for automatic allocation, in CIDR notation")
 	mflag.StringVar(&ipamConfig.IPSubnetCIDR, []string{"-ipalloc-default-subnet"}, "", "subnet to allocate within by default, in CIDR notation")
@@ -219,6 +222,7 @@ func main() {
 	mflag.BoolVar(&pluginConfig.EnableV2Multicast, []string{"-plugin-v2-multicast"}, false, "enable multicast for Docker plugin (v2)")
 	mflag.StringVar(&pluginConfig.Socket, []string{"-plugin-socket"}, "/run/docker/plugins/weave.sock", "plugin socket on which to listen")
 	mflag.StringVar(&pluginConfig.MeshSocket, []string{"-plugin-mesh-socket"}, "/run/docker/plugins/weavemesh.sock", "plugin socket on which to listen in mesh mode")
+	mflag.BoolVar(&bridgeConfig.NoMasqLocal, []string{"-no-masq-local"}, false, "do not SNAT external traffic sent to containers running on this node")
 
 	proxyConfig := newProxyConfig()
 
@@ -303,7 +307,8 @@ func main() {
 
 	bridgeConfig.Mac = name.String()
 	bridgeConfig.Port = config.Port
-	bridgeType, err := weavenet.EnsureBridge(procPath, &bridgeConfig, Log)
+	ips := ipset.New(common.LogLogger(), 0)
+	bridgeType, err := weavenet.EnsureBridge(procPath, &bridgeConfig, Log, ips)
 	checkFatal(err)
 	Log.Println("Bridge type is", bridgeType)
 
@@ -334,6 +339,9 @@ func main() {
 	if bridgeConfig.AWSVPC && !ipamConfig.Enabled() {
 		Log.Fatalf("--awsvpc mode requires IPAM enabled")
 	}
+	if bridgeConfig.AWSVPC && bridgeConfig.NoMasqLocal {
+		Log.Fatalf("--awsvpc mode is not compatible with the --no-masq-local option")
+	}
 
 	db, err := db.NewBoltDB(dbPrefix)
 	checkFatal(err)
@@ -393,17 +401,19 @@ func main() {
 	var (
 		allocator     *ipam.Allocator
 		defaultSubnet address.CIDR
-		trackerName   string
 	)
 	if ipamConfig.Enabled() {
 		var t tracker.LocalRangeTracker
 		if bridgeConfig.AWSVPC {
-			Log.Infoln("Creating AWSVPC LocalRangeTracker")
 			t, err = tracker.NewAWSVPCTracker(bridgeConfig.WeaveBridgeName)
 			if err != nil {
 				Log.Fatalf("Cannot create AWSVPC LocalRangeTracker: %s", err)
 			}
-			trackerName = "awsvpc"
+		} else if bridgeConfig.NoMasqLocal {
+			t = weavenet.NewNoMasqLocalTracker(ips)
+		}
+		if t != nil {
+			Log.Infof("Using %q LocalRangeTracker", t)
 		}
 
 		preClaims, err := findExistingAddresses(dockerCli, bridgeConfig.WeaveBridgeName)
@@ -451,7 +461,7 @@ func main() {
 	if httpAddr != "" {
 		muxRouter := mux.NewRouter()
 		if allocator != nil {
-			allocator.HandleHTTP(muxRouter, defaultSubnet, trackerName, dockerCli)
+			allocator.HandleHTTP(muxRouter, defaultSubnet, dockerCli)
 		}
 		if ns != nil {
 			ns.HandleHTTP(muxRouter, dockerCli)
@@ -474,10 +484,17 @@ func main() {
 		muxRouter.Methods("GET").Path("/metrics").Handler(metricsHandler(router, allocator, ns, dnsserver))
 		statusMux := http.NewServeMux()
 		statusMux.Handle("/", muxRouter)
-		Log.Println("Listening for metrics requests on", statusAddr)
+		Log.Println("Listening for status+metrics requests on", statusAddr)
 		go listenAndServeHTTP(statusAddr, statusMux)
 	}
 
+	if metricsAddr != "" {
+		metricsMux := http.NewServeMux()
+		metricsMux.Handle("/metrics", metricsHandler(router, allocator, ns, dnsserver))
+		Log.Println("Listening for metrics requests on", metricsAddr)
+		go listenAndServeHTTP(metricsAddr, metricsMux)
+	}
+
 	if plugin != nil {
 		go plugin.Start(httpAddr, dockerCli, waitReady.Add())
 	}

prog/weaveutil/container.go

@@ -89,8 +89,9 @@ func containerFQDN(args []string) error {
 	return nil
 }
 
-func runContainer(args []string) error {
+func parseContainerArgs(args []string) docker.CreateContainerOptions {
 	env := []string{}
+	labels := map[string]string{}
 	name := ""
 	net := ""
 	pid := ""
@@ -109,6 +110,10 @@ func runContainer(args []string) error {
 				env = append(env, args[i+1])
 			}
 			args = append(args[:i], args[i+2:]...)
+		case "-l", "--label":
+			key, value := parseLabel(args[i+1])
+			labels[key] = value
+			args = append(args[:i], args[i+2:]...)
 		case "--name":
 			name = args[i+1]
 			args = append(args[:i], args[i+2:]...)
@@ -161,19 +166,25 @@ func runContainer(args []string) error {
 	image := args[0]
 	cmds := args[1:]
 
+	config := docker.Config{Image: image, Env: env, Cmd: cmds, Labels: labels}
+	hostConfig := docker.HostConfig{NetworkMode: net, PidMode: pid, Privileged: privileged, RestartPolicy: restart, Binds: volumes, VolumesFrom: volumesFrom}
+	return docker.CreateContainerOptions{Name: name, Config: &config, HostConfig: &hostConfig}
+}
+
+func runContainer(args []string) error {
+	containerOptions := parseContainerArgs(args)
+
 	c, err := docker.NewVersionedClientFromEnv("1.18")
 	if err != nil {
 		return fmt.Errorf("unable to connect to docker: %s", err)
 	}
 
-	config := docker.Config{Image: image, Env: env, Cmd: cmds}
-	hostConfig := docker.HostConfig{NetworkMode: net, PidMode: pid, Privileged: privileged, RestartPolicy: restart, Binds: volumes, VolumesFrom: volumesFrom}
-	container, err := c.CreateContainer(docker.CreateContainerOptions{Name: name, Config: &config, HostConfig: &hostConfig})
+	container, err := c.CreateContainer(containerOptions)
 	if err != nil {
 		return fmt.Errorf("unable to create container: %s", err)
 	}
 
-	err = c.StartContainer(container.ID, &hostConfig)
+	err = c.StartContainer(container.ID, containerOptions.HostConfig)
 	if err != nil {
 		return fmt.Errorf("unable to start container: %s", err)
 	}
@@ -182,6 +193,14 @@ func runContainer(args []string) error {
 	return nil
 }
 
+func parseLabel(s string) (key, value string) {
+	pos := strings.Index(s, "=")
+	if pos == -1 { // no value - set it to blank
+		return s, ""
+	}
+	return s[:pos], s[pos+1:]
+}
+
 func listContainers(args []string) error {
 	if len(args) > 1 {
 		cmdUsage("list-containers", "[<label>]")

prog/weaveutil/container_test.go

@@ -0,0 +1,36 @@
+package main
+
+import (
+	"testing"
+
+	docker "github.com/fsouza/go-dockerclient"
+	"github.com/stretchr/testify/require"
+)
+
+func TestParseContainerArgs(t *testing.T) {
+	args := []string{"--name", "weave", "--privileged", "--net", "host",
+		"-l", "foo=bar", "-l", "foobar",
+		"-v", "/var/run/docker.sock:/var/run/docker.sock", "-v", "/etc:/host/etc",
+		"--restart", "always", "--pid", "host", "--volumes-from", "weavedb",
+		"-e", "WEAVE_DEBUG", "-e", "EXEC_IMAGE=weaveworks/weaveexec:latest",
+		"weaveworks/weave:latest", "cmd", "arg1", "arg2"}
+	expected := docker.CreateContainerOptions{
+		Name: "weave",
+		Config: &docker.Config{
+			Image:  "weaveworks/weave:latest",
+			Env:    []string{"WEAVE_DEBUG", "EXEC_IMAGE=weaveworks/weaveexec:latest"},
+			Cmd:    []string{"cmd", "arg1", "arg2"},
+			Labels: map[string]string{"foo": "bar", "foobar": ""},
+		},
+		HostConfig: &docker.HostConfig{
+			NetworkMode:   "host",
+			PidMode:       "host",
+			Privileged:    true,
+			RestartPolicy: docker.RestartPolicy{Name: "always"},
+			Binds:         []string{"/var/run/docker.sock:/var/run/docker.sock", "/etc:/host/etc"},
+			VolumesFrom:   []string{"weavedb"},
+		},
+	}
+	have := parseContainerArgs(args)
+	require.Equal(t, expected, have)
+}

prog/weaveutil/main.go

@@ -4,10 +4,7 @@ package main
 import (
 	"fmt"
 	"os"
-	"path/filepath"
 	"strings"
-
-	weavenet "github.com/weaveworks/weave/net"
 )
 
 var commands map[string]func([]string) error
@@ -36,12 +33,6 @@ func init() {
 		"create-volume-container":  createVolumeContainer,
 		"attach-container":         attach,
 		"detach-container":         detach,
-		"configure-arp":            configureARP,
-		"check-iface":              checkIface,
-		"del-iface":                delIface,
-		"setup-iface":              setupIface,
-		"setup-iface-addrs":        setupIfaceAddrs,
-		"list-netdevs":             listNetDevs,
 		"cni-net":                  cniNet,
 		"cni-ipam":                 cniIPAM,
 		"bridge-ip":                bridgeIP,
@@ -55,14 +46,6 @@ func init() {
 }
 
 func main() {
-	// force re-exec of this binary
-	selfPath, err := filepath.EvalSymlinks("/proc/self/exe")
-	if err != nil {
-		fmt.Fprintln(os.Stderr, err)
-		os.Exit(1)
-	}
-	weavenet.WeaveUtilCmd = selfPath
-
 	// If no args passed, act as CNI plugin based on executable name
 	switch {
 	case len(os.Args) == 1 && strings.HasSuffix(os.Args[0], "weave-ipam"):

prog/weaveutil/netdev.go

@@ -1,141 +0,0 @@
-package main
-
-import (
-	"encoding/json"
-	"fmt"
-	"strconv"
-	"strings"
-
-	"github.com/j-keck/arping"
-	"github.com/vishvananda/netlink"
-
-	weavenet "github.com/weaveworks/weave/net"
-)
-
-// checkIface returns an error if the given interface cannot be found.
-func checkIface(args []string) error {
-	if len(args) != 1 {
-		cmdUsage("check-iface", "<iface-name>")
-	}
-	ifaceName := args[0]
-
-	if _, err := netlink.LinkByName(ifaceName); err != nil {
-		return err
-	}
-
-	return nil
-}
-
-func delIface(args []string) error {
-	if len(args) != 1 {
-		cmdUsage("del-iface", "<iface-name>")
-	}
-	ifName := args[0]
-
-	link, err := netlink.LinkByName(ifName)
-	if err != nil {
-		return err
-	}
-	return netlink.LinkDel(link)
-}
-
-// setupIface renames the given iface and configures its ARP cache settings.
-func setupIface(args []string) error {
-	if len(args) != 2 {
-		cmdUsage("setup-iface", "<iface-name> <new-iface-name>")
-	}
-	ifaceName := args[0]
-	newIfName := args[1]
-
-	return weavenet.SetupIface(ifaceName, newIfName)
-}
-
-// setupIfaceAddrs sets up addresses on an interface. It expects to be called inside the container's netns.
-func setupIfaceAddrs(args []string) error {
-	if len(args) < 1 {
-		cmdUsage("setup-iface-addrs", "<iface-name> <with-multicast> <cidr>...")
-	}
-	link, err := netlink.LinkByName(args[0])
-	if err != nil {
-		return err
-	}
-	withMulticastRoute, err := strconv.ParseBool(args[1])
-	if err != nil {
-		return err
-	}
-	cidrs, err := parseCIDRs(args[2:])
-	if err != nil {
-		return err
-	}
-	return weavenet.SetupIfaceAddrs(link, withMulticastRoute, cidrs)
-}
-
-func configureARP(args []string) error {
-	if len(args) != 2 {
-		cmdUsage("configure-arp", "<iface-name-prefix> <root-path>")
-	}
-	prefix := args[0]
-	rootPath := args[1]
-
-	links, err := netlink.LinkList()
-	if err != nil {
-		return err
-	}
-	for _, link := range links {
-		ifName := link.Attrs().Name
-		if strings.HasPrefix(ifName, prefix) {
-			weavenet.ConfigureARPCache(rootPath+"/proc", ifName)
-			if addrs, err := netlink.AddrList(link, netlink.FAMILY_V4); err == nil {
-				for _, addr := range addrs {
-					arping.GratuitousArpOverIfaceByName(addr.IPNet.IP, ifName)
-				}
-			}
-		}
-	}
-
-	return nil
-}
-
-// listNetDevs outputs network ifaces identified by the given indexes
-// in the format of weavenet.Dev.
-func listNetDevs(args []string) error {
-	if len(args) == 0 {
-		cmdUsage("list-netdevs", "<iface-index>[ <iface-index>]")
-	}
-
-	indexes := make(map[int]struct{})
-	for _, index := range args {
-		if index != "" {
-			id, err := strconv.Atoi(index)
-			if err != nil {
-				return err
-			}
-			indexes[id] = struct{}{}
-		}
-	}
-
-	links, err := netlink.LinkList()
-	if err != nil {
-		return err
-	}
-
-	var netdevs []weavenet.Dev
-
-	for _, link := range links {
-		if _, found := indexes[link.Attrs().Index]; found {
-			netdev, err := weavenet.LinkToNetDev(link)
-			if err != nil {
-				return err
-			}
-			netdevs = append(netdevs, netdev)
-		}
-	}
-
-	nds, err := json.Marshal(netdevs)
-	if err != nil {
-		return err
-	}
-	fmt.Println(string(nds))
-
-	return nil
-}

prog/weaveutil/plugin_network.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"strings"
 
+	"github.com/docker/docker/api/types/filters"
 	"github.com/docker/docker/client"
 	docker "github.com/fsouza/go-dockerclient"
 )
@@ -58,7 +59,7 @@ func isDockerPluginEnabled(args []string) error {
 
 	ctx := context.Background()
 
-	plugins, err := c.PluginList(ctx)
+	plugins, err := c.PluginList(ctx, filters.Args{})
 	if err != nil {
 		return err
 	}

site/building.md

@@ -6,27 +6,28 @@ search_type: Documentation
 You only need to build Weave Net if you want to work on the Weave Net codebase
 (or you just enjoy building software).
 
-Apart from the `weave` shell script, weave is delivered as a set of
+Apart from the `weave` shell script, Weave Net is delivered as a set of
 container images.  There is no distribution-specific packaging, so in
 principle it shouldn't matter which Linux distribution you build
 under.  But naturally, Docker is a prerequisite (version 1.6.0 or
 later).
 
 The only way to build is by using the the build container; the
-Makefile is setup to make this transparent.  This method is
+`Makefile` is setup to make this transparent.  This method is
 documented below, and can be run directly on your machine or using
 a Vagrant VM.
 
 ## <a name="ubuntu"></a>Building directly on your machine
 
 The weave git repository should be cloned into
-`$GOPATH/src/github.com/weaveworks/weave`, in accordance with [the go
+`$GOPATH/src/github.com/weaveworks/weave`, in accordance with [the Go
 workspace conventions](https://golang.org/doc/code.html#Workspaces):
 
 ```
 $ WEAVE=github.com/weaveworks/weave
 $ git clone https://$WEAVE $GOPATH/src/$WEAVE
 $ cd $GOPATH/src/$WEAVE
+$ git submodule update --init
 ```
 
 Next install Docker if you haven't already, by following the instructions
@@ -44,7 +45,7 @@ On a fresh repository, the Makefile will do the following:
 - build the weave components in the build container
 - package them into two Docker images (`weaveworks/weave`,
 `weaveworks/weaveexec`)
-- Exported these images as `weave.tar.gz`
+- export these images as `weave.tar.gz`.
 
 The first two steps may take a while - don't worry, they are
 are cached and should not need to be redone very often.
@@ -63,18 +64,19 @@ First, check out the code:
 ```
 $ git clone https://github.com/weaveworks/weave
 $ cd weave
+$ git submodule update --init
 ```
 
-The `Vagrantfile` in the top directory constructs a VM that has
+The `Vagrantfile` in the top directory constructs a VM that has:
 
- * docker installed
- * go tools installed
- * weave dependencies installed
- * $GOPATH set to ~
+ * Docker installed
+ * Go tools installed
+ * Weave's dependencies installed
+ * `$GOPATH` set to `~`
  * the local working directory mapped as a synced folder into the
-   right place in $GOPATH
+   right place in `$GOPATH`.
 
-Once you are in the working directory you can issue
+Once you are in the working directory you can issue:
 
 ```
 $ vagrant up
@@ -86,7 +88,7 @@ installation is done just once). The working directory is sync'ed with
 use git and so on in the regular filesystem.
 
 To build and run the code, you need to use the VM. To log in and build
-the weave image, do
+the weave image, do:
 
 ```
 $ vagrant ssh
@@ -94,7 +96,7 @@ vm$ cd src/github.com/weaveworks/weave
 vm$ make
 ```
 
-The Docker daemon is also running in this VM, so you can then do
+The Docker daemon is also running in this VM, so you can then do:
 
 ```
 vm$ sudo ./weave launch
@@ -104,7 +106,7 @@ vm$ sudo docker ps
 and so on.
 
 If you are looking to just do a build and not run anything on this VM,
-you can do so with
+you can do so with:
 
 ```
 $ vagrant ssh -c 'make -C src/github.com/weaveworks/weave'
@@ -112,7 +114,7 @@ $ vagrant ssh -c 'make -C src/github.com/weaveworks/weave'
 
 you should then find a `weave.tar.gz` container snapshot tarball in the
 top-level directory. You can use that snapshot with `docker load`
-against a different host, e.g.
+against a different host, e.g.:
 
 ```
 $ export DOCKER_HOST=tcp://<HOST:PORT>

site/faq.md

@@ -40,6 +40,18 @@ A simple way to accomplish this would be to run Weave on the host and then run,
 Yet another option is to expose a port from the container on host B and then connect to it. You can read about exposing ports in [Exporting Services](/site/tasks/manage/service-management.md#exporting).
 
 
+<a name="duplicate-peer"></a>
+**Q: Why am I seeing "peer names collision" and failed connections?**
+
+This sometimes happens when machines are cloned; we require each
+machine in your cluster to have a unique identity.
+
+For more information see [Peer Names](/site/operational-guide/concepts.md#peer-name).
+
+Depending on your Linux distribution you may need to [set up
+machine-id](https://www.freedesktop.org/software/systemd/man/machine-id.html)
+or [generate a dbus id](https://dbus.freedesktop.org/doc/dbus-uuidgen.1.html).
+
 <a name="duplicate-ip"></a>
 **Q: Why am I seeing the same IP address assigned to two different containers on different hosts?**
 
@@ -92,7 +104,8 @@ Yes, of course!  Weave allows you to run isolated networks and still allow open
 You must permit traffic to flow through TCP 6783 and UDP 6783/6784,
 which are Weave’s control and data ports.
 
-The daemon also uses TCP port 6782 for [metrics](/site/tasks/manage/metrics.md), but
+The daemon also uses TCP 6781/6782 for
+[metrics](/site/tasks/manage/metrics.md#metrics-endpoint-addresses), but
 you would only need to open up this port if you wish to collect metrics
 from another host.
 

site/install.md

@@ -31,7 +31,6 @@ Install Weave Net from the command line on its own or if you are using Docker, K
 * [Integrating Kubernetes and Mesos via the CNI Plugin](https://www.weave.works/docs/net/latest/kubernetes/)
 * [Integrating Kubernetes via the Addon](https://www.weave.works/docs/net/latest/kubernetes/kube-addon/)
      * [Installation](https://www.weave.works/docs/net/latest/kubernetes/kube-addon/#install)
-     * [Upgrading Kubernetes to version 1.6](https://www.weave.works/docs/net/latest/kubernetes/kube-addon/#kube-1.6-upgrade)
      * [Upgrading the Daemon Sets](https://www.weave.works/docs/net/latest/kubernetes/kube-addon/#daemon-sets)
      * [CPU and Memory Requirements](https://www.weave.works/docs/net/latest/kubernetes/kube-addon/#resources)
      * [Pod Eviction](https://www.weave.works/docs/net/latest/kubernetes/kube-addon/#eviction)

site/install/installing-weave.md

@@ -48,10 +48,8 @@ To disable this check, run the following before launching Weave Net:
 
 ### Guides for Specific Platforms
 
-CoreOS users see [here](/guides/networking-docker-containers-with-weave-on-coreos/) for an example of installing Weave using cloud-config.
-
-Amazon ECS users see [here](https://github.com/weaveworks/integrations/blob/master/aws/ecs/README.md)
-for the latest Weave AMIs and [here](https://www.weave.works/docs/tutorials/old-guides/ecs/) to get started with Weave Net on ECS.
+Amazon ECS users see [here](https://www.weave.works/docs/scope/latest/ami/)
+for the latest Weave AMIs.
 
 If you're on Amazon EC2, the standard installation instructions at the
 top of this page, provide the simplest setup and the most flexibility.

site/kubernetes/kube-addon.md

@@ -7,21 +7,25 @@ search_type: Documentation
 The following topics are discussed:
 
 * [Installation](#install)
- * [Upgrading Kubernetes to version 1.6](#kube-1.6-upgrade)
- * [Upgrading the Daemon Sets](#daemon-sets)
- * [CPU and Memory Requirements](#resources)
- * [Pod Eviction](#eviction)
+   * [Upgrading the Daemon Sets](#daemon-sets)
+   * [CPU and Memory Requirements](#resources)
+   * [Pod Eviction](#eviction)
 * [Features](#features)
- * [Pod Network](#pod-network)
- * [Network Policy](#npc)
+   * [Pod Network](#pod-network)
+   * [Network Policy](#npc)
 * [Troubleshooting](#troubleshooting)
- * [Troubleshooting Blocked Connections](#blocked-connections)
- * [Things to watch out for](#key-points)
+   * [Reading the logs](#logs)
+   * [Troubleshooting Blocked Connections](#blocked-connections)
+   * [Things to watch out for](#key-points)
 * [Changing Configuration Options](#configuration-options)
 
 
 ## <a name="install"></a> Installation
 
+*Before installing Weave Net, you should make sure the following ports are not
+blocked by your firewall: TCP 6783 and UDP 6783/6784.
+For more details, see the [FAQ](/site/faq.md#ports).*
+
 Weave Net can be installed onto your CNI-enabled Kubernetes cluster
 with a single command:
 
@@ -59,38 +63,13 @@ Shut down Kubernetes, and _on all nodes_ perform the following:
 Then relaunch Kubernetes and install the addon as described
 above.
 
-## <a name="kube-1.6-upgrade"></a> Upgrading Kubernetes to version 1.6
-
-In version 1.6, Kubernetes has increased security, so we need to
-create a special service account to run Weave Net. This is done in
-the file `weave-daemonset-k8s-1.6.yaml` attached to the [Weave Net
-release](https://github.com/weaveworks/weave/releases/latest).
-
-Also, the
-[toleration](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/taint-toleration-dedicated.md)
-required to let Weave Net run on master nodes has moved from an
-annotation to a field on the DaemonSet spec object.
-
-If you have edited the Weave Net DaemonSet from a previous release,
-you will need to re-make your changes against the new version.
-
 ### <a name="daemon-sets"></a> Upgrading the Daemon Sets
 
-For Kubernetes 1.6 and above the DaemonSet definition specifies
-[Rolling Updates](https://kubernetes.io/docs/tasks/manage-daemon/update-daemon-set/),
+The DaemonSet definition specifies [Rolling
+Updates](https://kubernetes.io/docs/tasks/manage-daemon/update-daemon-set/),
 so when you apply a new version Kubernetes will automatically restart
 the Weave Net pods one by one.
 
-Kubernetes v1.5 and below does not support rolling upgrades of daemon sets,
-and so you will need to perform the procedure manually:
-
-* Apply the updated addon manifest `kubectl apply -f "https://cloud.weave.works/k8s/net?k8s-version=$(kubectl version | base64 | tr -d '\n')"`
-* Kill each Weave Net pod with `kubectl delete` and then wait for it to reboot before moving on to the next pod.
-
-**Note:** In versions prior to Weave Net 2.0, deleting all Weave Net pods at the same time
-  will result in them losing track of IP address range ownership, possibly leading to
-  duplicate IP addresses if you then start a new copy of Weave Net.
-
 ## <a name="resources"></a>CPU and Memory Requirements
 
 Kubernetes manages
@@ -186,12 +165,58 @@ definition](https://v1-7.docs.kubernetes.io/docs/api-reference/v1.7/#networkpoli
 
 ## <a name="troubleshooting"></a> Troubleshooting
 
+The first thing to check is whether Weave Net is up and
+running. The `kubectl apply` command you used to install it only
+_requests_ that it be downloaded and started; if anything goes wrong
+at startup, those details will only be visible in the [logs](#logs) of
+the container(s).
+
+To check what is running:
+
+```
+$ kubectl get pods -n kube-system -l name=weave-net
+NAME              READY  STATUS   RESTARTS  AGE
+weave-net-1jkl6   2/2    Running  0         1d
+weave-net-bskbv   2/2    Running  0         1d
+weave-net-m4x1b   2/2    Running  0         1d
+```
+
+You should see one line for each node in your cluster; each line
+should have STATUS "Running", and READY should be 2 out of 2. If you
+see a STATUS like "Error" or "CrashLoopBackoff", look in the logs of
+the container with that status.
+
+### <a name="logs"></a> Reading the logs
+
+Pick one of the pods from the list output by `kubectl get pods` and
+ask for the logs like this:
+
+```
+$ kubectl logs -n kube-system weave-net-1jkl6 weave
+```
+
+For easier viewing, pipe the output into a file , especially if
+it is long.
+
+By default log level of `weave` container is set to `info` level. If you wish to see more detailed logs you can set the desired log level for the `-log-level` flag through the `EXTRA_ARGS` environment variable for the `weave` container in the weave-net daemon set. Add environment variable as below.
+
+```yaml
+      containers:
+      - command:
+        - /home/weave/launch.sh
+        name: weave
+        env:
+        - name: EXTRA_ARGS
+          value: --log-level=debug
+```
+
 Many Kubernetes network issues occur at a higher level than Weave Net.
 The [Kubernetes Service Debugging Guide]
 (https://kubernetes.io/docs/tasks/debug-application-cluster/debug-service/)
 has a detailed step-by-step guide.
 
-The status of Weave Net can be checked by running its CLI commands. This can be done in various ways:
+Once it is up and running, the status of Weave Net can be checked by
+running its CLI commands. This can be done in various ways:
 
 1\. [Install the `weave` script](/site/installing-weave.md) and run:
 
@@ -302,7 +327,15 @@ UDP connection from 10.32.0.7:56648 to 10.32.0.11:80 blocked by Weave NPC.
   impossible to correctly enforce network policies that restrict which
   pods can talk.
 - If you do set the `--cluster-cidr` option on kube-proxy, make sure
-  it matches the `IPALLOC_RANGE` given to Weave Net (see below)
+  it matches the `IPALLOC_RANGE` given to Weave Net (see below).
+- IP forwarding must be enabled on each node, in order for pods to
+  access Kubernetes services or other IP addresses on another
+  network. Check this with `sysctl net.ipv4.ip_forward`; the result
+  should be `1`. (Be aware that there can be security implications of
+  enabling IP forwarding).
+- Weave Net can be run on minikube v0.28 or later with the default CNI config shipped with minikube
+  being disabled. See [#3124](https://github.com/weaveworks/weave/issues/3124#issuecomment-397820940)
+  for more details.
 
 ## <a name="configuration-options"></a> Changing Configuration Options
 
@@ -311,11 +344,11 @@ UDP connection from 10.32.0.7:56648 to 10.32.0.11:80 blocked by Weave NPC.
 You can customise the YAML you get from `cloud.weave.works` by passing some of Weave Net's options, arguments and environment variables as query parameters:
 
   - `version`: Weave Net's version. Default: `latest`, i.e. latest release. *N.B.*: This only changes the specified version inside the generated YAML file, it does not ensure that the rest of the YAML is compatible with that version. To freeze the YAML version save a copy of the YAML file from the [release page](https://github.com/weaveworks/weave/releases) and use that copy instead of downloading it each time from `cloud.weave.works`.
-  - `password-secret`: name of the Kubernetes secret containing your password.
+  - `password-secret`: name of the Kubernetes secret containing your password.  *N.B*: The Kubernetes secret name must correspond to a name of a file containing your password.
      Example:
 
-        $ echo "s3cr3tp4ssw0rd" > /var/lib/weave/weave-passwd.txt
-        $ kubectl create secret -n kube-system generic weave-passwd --from-file=/var/lib/weave/weave-passwd.txt
+        $ echo "s3cr3tp4ssw0rd" > /var/lib/weave/weave-passwd
+        $ kubectl create secret -n kube-system generic weave-passwd --from-file=/var/lib/weave/weave-passwd
         $ kubectl apply -f "https://cloud.weave.works/k8s/net?k8s-version=$(kubectl version | base64 | tr -d '\n')&password-secret=weave-passwd"
 
   - `known-peers`: comma-separated list of hosts. Default: empty.
@@ -323,7 +356,6 @@ You can customise the YAML you get from `cloud.weave.works` by passing some of W
   - `disable-npc`: boolean (`true|false`). Default: `false`.
   - `env.NAME=VALUE`: add environment variable `NAME` and set it to `VALUE`.
   - `seLinuxOptions.NAME=VALUE`: add SELinux option `NAME` and set it to `VALUE`, e.g. `seLinuxOptions.type=spc_t`
-  - `use-legacy-netpol`: use [legacy NetworkPolicy semantics](https://v1-6.docs.kubernetes.io/docs/api-reference/v1.6/#networkpolicy-v1beta1-extensions), boolean (`true|false`). Default: `true` for Kubernetes version <= 1.6, `false` for > 1.6.
 
 The list of variables you can set is:
 
@@ -346,11 +378,18 @@ The list of variables you can set is:
 * `WEAVE_EXPOSE_IP` - set the IP address used as a gateway from the
   Weave network to the host network - this is useful if you are
   configuring the addon as a static pod.
+* `WEAVE_METRICS_ADDR` - address and port that the Weave Net
+  daemon will serve Prometheus-style metrics on (defaults to 0.0.0.0:6782)
+* `WEAVE_STATUS_ADDR` - address and port that the Weave Net
+  daemon will serve status requests on (defaults to disabled)
 * `WEAVE_MTU` - Weave Net defaults to 1376 bytes, but you can set a
   smaller size if your underlying network has a tighter limit, or set
   a larger size for better performance if your network supports jumbo
   frames - see [here](/site/tasks/manage/fastdp.md#mtu) for more
   details.
+* `NO_MASQ_LOCAL` - set to 1 to preserve the client source IP address when
+  accessing Service annotated with `service.spec.externalTrafficPolicy=Local`.
+  The feature works only with Weave IPAM (default).
 
 Example:
 ```

site/tasks/ipam/troubleshooting-ipam.md

@@ -67,3 +67,18 @@ Columns are as follows:
   partition, it may be because the peer has failed and needs to be
   removed administratively - see [Starting, Stopping and Removing
   Peers](/site/tasks/ipam/stop-remove-peers-ipam.md) for more details.
+
+
+### <a name="seeded-different-peers"></a>Seeded by Different Peers
+
+If you see the message: `IP allocation was seeded by different peers`,
+this means that some Weave Net peers were initialized into one cluster
+and some into another cluster; Weave Net cannot operate in this state.
+
+To recover, you need to eliminate the IPAM data from the affected
+nodes and restart.  If you installed via the Kubernetes Addon, this
+data will be in a file under `/var/lib/weave` on the node - delete
+this file and restart the node.
+
+For other installations, run `weave reset` and restart. Any existing
+connections to containers will be lost.

site/tasks/manage/metrics.md

@@ -10,9 +10,6 @@ controller](/site/kubernetes/kube-addon.md#npc).
 
 ### Router Metrics
 
-The endpoint address is `localhost:6782`; the following metrics are
-exposed:
-
 * `weave_connections` - Number of peer-to-peer connections.
 * `weave_connection_terminations_total` - Number of peer-to-peer
   connections terminated.
@@ -25,27 +22,35 @@ exposed:
 * `weave_ipam_pending_allocates` - Number of pending allocates.
 * `weave_ipam_pending_claims` - Number of pending claims.
 
-#### Publish Router Metrics Endpoint
-
-By default, when started via `weave launch`, weave listens on its local
-interface to serve metrics. To publish your metrics throughout your cluster,
-e.g. if your prometheus server is installed on a different host machine,
-you need to set `WEAVE_STATUS_ADDR` to your corresponding IP and port.
-Default port is 6782.
-
-`WEAVE_STATUS_ADDR=X.X.X.X:PORT`
-
-You can set `WEAVE_STATUS_ADDR=0.0.0.0:6782` to listen on all interfaces,
-but be aware, this may expose your metrics to the public internet.
-
 ### Kubernetes Network Policy Controller Metrics
 
-The endpoint address is `localhost:6781`; the following metric is
+The following metric is
 exposed:
 
 * `weavenpc_blocked_connections_total` - Connection attempts blocked
   by policy controller.
 
+### Metrics Endpoint Addresses
+
+When installed as a Kubernetes Addon, the router listens for metrics
+requests on 0.0.0.0:6782 and the Network Policy Controller listens on
+0.0.0.0:6781. No other requests are served on these endpoints.
+
+>Note: If your Kubernetes hosts are exposed to the public internet
+then these metrics endpoints will also be exposed.
+
+When started via `weave launch`, by default weave listens on its local
+interface to serve metrics and other read-only status requests. To
+publish your metrics throughout your cluster, you can set
+`WEAVE_STATUS_ADDR`:
+
+`WEAVE_STATUS_ADDR=X.X.X.X:PORT`
+
+Set it to an empty string to disable.
+
+You can also pass the parameter `--metrics-addr=X.X.X.X:PORT` to
+`weave launch` to specify an address to listen for metrics only.
+
 # Static Configuration for Weave Net
 
 The following YAML fragment can be used in your Prometheus configuration to

site/troubleshooting.md

@@ -157,6 +157,9 @@ The columns are as follows:
    the encryption mode, data transport method, remote peer name and
    nickname for pending and established connections, mtu if known
 
+Specific error messages:
+* `IP allocation was seeded by different peers` - [more details here](/site/ipam/troubleshooting-ipam.md#seeded-different-peers)
+
 ### <a name="weave-status-peers"></a>List Peers
 
 Detailed information on peers can be obtained with `weave status
@@ -329,5 +332,6 @@ and the container image versions as git hashes.
 
 **See Also**
 
+ * [Troubleshooting the Kubernetes Addon](/site/kubernetes/kube-addon/#troubleshooting)
  * [Troubleshooting IPAM](/site/tasks/ipam/troubleshooting-ipam.md)
  * [Troubleshooting the Proxy](/site/tasks/weave-docker-api/using-proxy.md)

test/840_weave_kube_3_test.sh

@@ -4,7 +4,7 @@
 
 tear_down_kubeadm() {
     for host in $HOSTS; do
-        run_on $host "sudo kubeadm reset && sudo rm -r -f /opt/cni/bin/*weave*"
+        run_on $host "sudo kubeadm reset --force && sudo rm -r -f /opt/cni/bin/*weave*"
         # Remove the rule to ensure that api-server is not accessible while kube-proxy
         # is not ready. Otherwise, we have a race between "-j KUBE-FORWARD" and "-j WEAVE-NPC"
         # which makes the NodePort isolation test case to fail.
@@ -34,27 +34,25 @@ docker_on $HOST1 run --rm --privileged --net=host --entrypoint=/usr/sbin/ipset w
 
 # kubeadm init upgrades to latest Kubernetes version by default, therefore we try to lock the version using the below option:
 k8s_version="$(run_on $HOST1 "kubelet --version" | grep -oP "(?<=Kubernetes )v[\d\.\-beta]+")"
-k8s_version_option="$([[ "$k8s_version" > "v1.6" ]] && echo "kubernetes-version" || echo "use-kubernetes-version")"
 
 for host in $HOSTS; do
     if [ $host = $HOST1 ] ; then
-	run_on $host "sudo systemctl start kubelet && sudo kubeadm init --$k8s_version_option=$k8s_version --token=$TOKEN --pod-network-cidr=$WEAVE_NETWORK"
+	run_on $host "sudo systemctl start kubelet && sudo kubeadm init --kubernetes-version=$k8s_version --token=$TOKEN --pod-network-cidr=$WEAVE_NETWORK"
     else
 	run_on $host "sudo systemctl start kubelet && sudo kubeadm join --token=$TOKEN --discovery-token-unsafe-skip-ca-verification $HOST1IP:$KUBE_PORT"
     fi
 done
 
 if [ -n "$COVERAGE" ]; then
-    COVERAGE_ARGS="env:\\n                - name: EXTRA_ARGS\\n                  value: \"-test.coverprofile=/home/weave/cover.prof --\""
+    WEAVE_ENV_VARS="env:\\n                - name: EXTRA_ARGS\\n                  value: \"-test.coverprofile=/home/weave/cover.prof --\""
 else
-    COVERAGE_ARGS="env:"
+    WEAVE_ENV_VARS="env:"
 fi
 
 # Ensure Kubernetes uses locally built container images and inject code coverage environment variable (or do nothing depending on $COVERAGE):
 sed -e "s%imagePullPolicy: Always%imagePullPolicy: Never%" \
-    -e "s%env:%$COVERAGE_ARGS%" \
-    -e "s%#npc-args%              args:\n                - '--use-legacy-netpol'%" \
-    "$(dirname "$0")/../prog/weave-kube/weave-daemonset-k8s-1.7.yaml" | run_on "$HOST1" "$KUBECTL apply -n kube-system -f -"
+    -e "s%env:%$WEAVE_ENV_VARS%" \
+    "$(dirname "$0")/../prog/weave-kube/weave-daemonset-k8s-1.8.yaml" | run_on "$HOST1" "$KUBECTL apply -n kube-system -f -"
 
 sleep 5
 
@@ -64,7 +62,7 @@ check_connections() {
 
 assert_raises 'wait_for_x check_connections "connections to establish"'
 
-# Check we can ping between the Weave bridg IPs on each host
+# Check we can ping between the Weave bridge IPs on each host
 HOST1EXPIP=$($SSH $HOST1 "weave expose" || true)
 HOST2EXPIP=$($SSH $HOST2 "weave expose" || true)
 HOST3EXPIP=$($SSH $HOST3 "weave expose" || true)
@@ -77,10 +75,19 @@ assert "run_on $HOST1 ps aux | grep -c '[d]efunct'" "0"
 assert "run_on $HOST2 ps aux | grep -c '[d]efunct'" "0"
 assert "run_on $HOST3 ps aux | grep -c '[d]efunct'" "0"
 
-# Set up a simple network policy so all our test pods can talk to each other
-run_on $HOST1 "$KUBECTL annotate ns default net.beta.kubernetes.io/network-policy='{\"ingress\":{\"isolation\":\"DefaultDeny\"}}'"
+# Set up simple network policies so all our test pods can talk to each other
 run_on $HOST1 "$KUBECTL apply -f -" <<EOF
-apiVersion: extensions/v1beta1
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: default-deny
+spec:
+  podSelector: {}
+  policyTypes:
+  - Ingress
+EOF
+run_on $HOST1 "$KUBECTL apply -f -" <<EOF
+apiVersion: networking.k8s.io/v1
 kind: NetworkPolicy
 metadata:
   name: test840
@@ -103,7 +110,7 @@ EOF
 
 # Another policy, this time with no 'from' section, just to check that doesn't cause a crash
 run_on $HOST1 "$KUBECTL apply -f -" <<EOF
-apiVersion: extensions/v1beta1
+apiVersion: networking.k8s.io/v1
 kind: NetworkPolicy
 metadata:
   name: test840f
@@ -180,17 +187,6 @@ run_on $HOST1 "$KUBECTL run nettest-deny --labels=\"access=deny,run=nettest-deny
 denyPodName=$($SSH $HOST1 "$KUBECTL get pods -l run=nettest-deny -o go-template='{{(index .items 0).metadata.name}}'")
 assert_raises "! $SSH $HOST1 $KUBECTL exec $denyPodName -- curl -s -S -f -m 2 http://$DOMAIN:8080/status >/dev/null"
 
-# Restart weave-net with npc in non-legacy mode
-$SSH $HOST1 "$KUBECTL delete ds weave-net -n=kube-system"
-sed -e "s%imagePullPolicy: Always%imagePullPolicy: Never%" \
-    -e "s%env:%$COVERAGE_ARGS%" \
-    "$(dirname "$0")/../prog/weave-kube/weave-daemonset-k8s-1.7.yaml" | run_on "$HOST1" "$KUBECTL apply -n kube-system -f -"
-
-assert_raises 'wait_for_x check_all_pods_communicate pods'
-
-# nettest-deny should still not be able to reach nettest pods
-assert_raises "! $SSH $HOST1 $KUBECTL exec $denyPodName -- curl -s -S -f -m 2 http://$DOMAIN:8080/status >/dev/null"
-
 # check access via virtual IP
 VIRTUAL_IP=$($SSH $HOST1 $KUBECTL get service netvirt -o template --template={{.spec.clusterIP}})
 assert_raises   "$SSH $HOST1 $KUBECTL exec     $podName -- curl -s -S -f -m 2 http://$VIRTUAL_IP/status >/dev/null"
@@ -226,12 +222,39 @@ run_on $HOST1 "$KUBECTL delete netpol allow-nettest-deny"
 # nettest-deny should still not be able to reach nettest pods
 assert_raises "! $SSH $HOST1 $KUBECTL exec $denyPodName -- curl -s -S -f -m 2 http://$DOMAIN:8080/status >/dev/null"
 
-# allow access for all
+# Create many namespaces to stress namespaceSelector
+for n in 1 2 3 4 5 6 7 8 9 10; do
+    run_on $HOST1 "$KUBECTL create namespace namespace${n}"
+done
+
+# allow access from any namespace
 run_on $HOST1 "$KUBECTL apply -f -" <<EOF
 apiVersion: networking.k8s.io/v1
 kind: NetworkPolicy
 metadata:
-  name: allow-nettest-deny
+  name: allow-any-namespace
+  namespace: default
+spec:
+  podSelector: {}
+  ingress:
+    - from:
+      - namespaceSelector: {}
+EOF
+
+# Should be able to access from the "deny" pod now
+assert_raises "$SSH $HOST1 $KUBECTL exec $denyPodName -- curl -s -S -f -m 2 http://$DOMAIN:8080/status >/dev/null"
+
+# host should still not be able to reach pods via service virtual IP or NodePort
+# because host is not in a namespace
+assert_raises "! $SSH $HOST1 curl -s -S -f -m 2 http://$VIRTUAL_IP/status >/dev/null"
+assert_raises "! $SSH $HOST1 curl -s -S -f -m 2 http://$HOST2:31138/status >/dev/null"
+
+# allow access from anywhere
+run_on $HOST1 "$KUBECTL apply -f -" <<EOF
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: allow-anywhere
   namespace: default
 spec:
   podSelector: {}
@@ -239,12 +262,85 @@ spec:
     - {}
 EOF
 
-assert_raises "$SSH $HOST1 $KUBECTL exec $denyPodName -- curl -s -S -f -m 2 http://$DOMAIN:8080/status >/dev/null"
-
 # Virtual IP and NodePort should now work
 assert_raises "$SSH $HOST1 curl -s -S -f -m 2 http://$VIRTUAL_IP/status >/dev/null"
 assert_raises "$SSH $HOST1 curl -s -S -f -m 2 http://$HOST2:31138/status >/dev/null"
 
+# allow nettest-deny to access only to 8.8.8.0/24
+run_on $HOST1 "$KUBECTL apply -f -" <<EOF
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: egress-nettest-deny-dns
+  namespace: default
+spec:
+  podSelector:
+    matchLabels:
+      run: nettest-deny
+  policyTypes:
+  - Egress
+  egress:
+  - to:
+    - ipBlock:
+        cidr: 8.8.8.0/24
+        except:
+        - 8.8.8.4/32
+EOF
+
+assert_raises "! $SSH $HOST1 $KUBECTL exec $denyPodName -- curl -s -S -f -m 2 http://$DOMAIN:8080/status >/dev/null"
+assert_raises "! $SSH $HOST1 $KUBECTL exec $denyPodName -- dig @8.8.8.4 google.com >/dev/null"
+assert_raises "$SSH $HOST1 $KUBECTL exec $denyPodName -- dig @8.8.8.8 google.com >/dev/null"
+
+# also, allow nettest-deny to access nettest and kube-system (for kube-dns)
+run_on $HOST1 "$KUBECTL label namespace kube-system name=kube-system"
+run_on $HOST1 "$KUBECTL replace -f -" <<EOF
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: egress-nettest-deny-dns
+  namespace: default
+spec:
+  podSelector:
+    matchLabels:
+      run: nettest-deny
+  policyTypes:
+  - Egress
+  egress:
+  - to:
+    - ipBlock:
+        cidr: 8.8.8.0/24
+        except:
+        - 8.8.8.4/32
+    - podSelector:
+        matchLabels:
+          run: nettest
+    - namespaceSelector:
+        matchLabels:
+          name: kube-system
+EOF
+
+assert_raises "$SSH $HOST1 $KUBECTL exec $denyPodName -- curl -s -S -f -m 2 http://$DOMAIN:8080/status >/dev/null"
+
+# Passing --no-masq-local and setting externalTrafficPolicy to Local must preserve
+# the client source IP addr
+
+CLIENT_IP_MASQ="$($SSH $HOST1 curl -sS http://$HOST2:31138/client_ip)"
+
+WEAVE_ENV_VARS="${WEAVE_ENV_VARS}\\n                - name: NO_MASQ_LOCAL\\n                  value: \"1\""
+$SSH $HOST1 "$KUBECTL delete ds weave-net -n=kube-system"
+sed -e "s%imagePullPolicy: Always%imagePullPolicy: Never%" \
+    -e "s%env:%$WEAVE_ENV_VARS%" \
+    "$(dirname "$0")/../prog/weave-kube/weave-daemonset-k8s-1.8.yaml" | run_on "$HOST1" "$KUBECTL apply -n kube-system -f -"
+
+sleep 5
+
+run_on $HOST1 "$KUBECTL patch svc netvirt -p '{\"spec\":{\"externalTrafficPolicy\":\"Local\"}}'"
+
+CLIENT_IP_NO_MASQ="$($SSH $HOST1 curl -sS http://$HOST2:31138/client_ip)"
+
+assert_raises "[ $CLIENT_IP_NO_MASQ != $CLIENT_IP_MASQ ]"
+assert_raises "[ $HOST1IP == $CLIENT_IP_NO_MASQ ]"
+
 tear_down_kubeadm
 
 # Destroy our test ipset, and implicitly check it is still there

test/870_weave_recovers_unreachable_ips_on_relaunch_3_test.sh

@@ -27,7 +27,7 @@ fi
 function teardown_kubernetes_cluster {
     greyly echo "Tearing down kubernetes cluster"
     for host in $HOSTS; do
-        run_on $host "sudo kubeadm reset && sudo rm -r -f /opt/cni/bin/*weave*"
+        run_on $host "sudo kubeadm reset --force && sudo rm -r -f /opt/cni/bin/*weave*"
     done
 }
 
@@ -38,11 +38,10 @@ function setup_kubernetes_cluster {
 
     # kubeadm init upgrades to latest Kubernetes version by default, therefore we try to lock the version using the below option:
     k8s_version="$(run_on $HOST1 "kubelet --version" | grep -oP "(?<=Kubernetes )v[\d\.\-beta]+")"
-    k8s_version_option="$([[ "$k8s_version" > "v1.6" ]] && echo "kubernetes-version" || echo "use-kubernetes-version")"
 
     for host in $HOSTS; do
         if [ $host = $HOST1 ] ; then
-        run_on $host "sudo systemctl start kubelet && sudo kubeadm init --$k8s_version_option=$k8s_version --token=$TOKEN"
+        run_on $host "sudo systemctl start kubelet && sudo kubeadm init --kubernetes-version=$k8s_version --token=$TOKEN"
         else
         run_on $host "sudo systemctl start kubelet && sudo kubeadm join --token=$TOKEN --discovery-token-unsafe-skip-ca-verification $HOST1IP:$KUBE_PORT"
         fi

test/config.sh

@@ -337,7 +337,7 @@ start_suite() {
 # Common postconditions to assert on each host, after each test:
 assert_common_postconditions() {
     # Ensure we do not generate any defunct (a.k.a. zombie) process:
-    assert "run_on $1 ps aux | grep -c '[d]efunct'" "0"
+    assert "run_on $1 ps aux | grep '[d]efunct'" ""
 }
 
 end_suite() {

test/images/network-tester/Dockerfile

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 FROM bash
-RUN apk add --no-cache curl
+RUN apk add --no-cache curl bind-tools
 ADD webserver webserver
 EXPOSE 8080
 ENTRYPOINT ["/webserver"]

test/images/network-tester/webserver.go

@@ -13,6 +13,8 @@
 // Visit /status to see pass/running/fail determination. (literally, it will
 // return one of those words.)
 //
+// Visit /client_ip to see the source IP addr of the request.
+//
 // /write is used by other network test servers to register connectivity.
 package main
 
@@ -78,6 +80,20 @@ func (s *State) serveStatus(w http.ResponseWriter, r *http.Request) {
 	fmt.Fprintf(w, "fail")
 }
 
+// serveClientIP returns the client source IP addr.
+func (s *State) serveClientIP(w http.ResponseWriter, r *http.Request) {
+	s.lock.Lock()
+	defer s.lock.Unlock()
+
+	host, _, err := net.SplitHostPort(r.RemoteAddr)
+	if err != nil {
+		go s.Logf("Warning: unable to parse remote addr: %s", err)
+		w.WriteHeader(http.StatusInternalServerError)
+	}
+
+	fmt.Fprintf(w, host)
+}
+
 // serveRead writes our json encoded state
 func (s *State) serveRead(w http.ResponseWriter, r *http.Request) {
 	s.lock.Lock()
@@ -186,6 +202,7 @@ func main() {
 	http.HandleFunc("/read", state.serveRead)
 	http.HandleFunc("/write", state.serveWrite)
 	http.HandleFunc("/status", state.serveStatus)
+	http.HandleFunc("/client_ip", state.serveClientIP)
 
 	go log.Fatal(http.ListenAndServe(fmt.Sprintf("0.0.0.0:%d", *port), nil))
 

test/run_all.sh

@@ -23,4 +23,4 @@ if [ -n "$CIRCLECI" -o -n "$PARALLEL" ]; then
     RUNNER_ARGS="$RUNNER_ARGS -parallel"
 fi
 
-HOSTS="$HOSTS" "${DIR}/../tools/runner/runner" -timeout=300 $RUNNER_ARGS $TESTS
+HOSTS="$HOSTS" "${DIR}/../tools/runner/runner" -timeout=360 $RUNNER_ARGS $TESTS

test/tls/generate_certs.go

@@ -41,15 +41,15 @@ func main() {
 			log.Fatalf("The client key already exists.  Please remove it or specify a different key/cert.")
 		}
 
-		err = cert.GenerateCert(
-			[]string{""},
-			clientCertPath,
-			clientKeyPath,
-			caCertPath,
-			caKeyPath,
-			org,
-			bits,
-		)
+		err = cert.GenerateCert(&cert.Options{
+			Hosts:     []string{""},
+			CertFile:  clientCertPath,
+			KeyFile:   clientKeyPath,
+			CAFile:    caCertPath,
+			CAKeyFile: caKeyPath,
+			Org:       org,
+			Bits:      bits,
+		})
 		if err != nil {
 			log.Fatalf("Error generating client certificate: %s", err)
 		}
@@ -64,15 +64,15 @@ func main() {
 
 		if _, err := os.Stat(serverCertPath); os.IsNotExist(err) {
 			log.Printf("Creating server certificate: %s", serverCertPath)
-			err = cert.GenerateCert(
-				[]string{ip},
-				serverCertPath,
-				serverKeyPath,
-				caCertPath,
-				caKeyPath,
-				org,
-				bits,
-			)
+			err = cert.GenerateCert(&cert.Options{
+				Hosts:     []string{ip},
+				CertFile:  serverCertPath,
+				KeyFile:   serverKeyPath,
+				CAFile:    caCertPath,
+				CAKeyFile: caKeyPath,
+				Org:       org,
+				Bits:      bits,
+			})
 			if err != nil {
 				log.Fatalf("error generating server cert: %s", err)
 			}

testing/kubernetes/testing/actions.go

@@ -0,0 +1,30 @@
+// Modified from the original at k8s.io/client-go/testing to the bare minimum
+/*
+Copyright 2015 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package testing
+
+import (
+	"k8s.io/apimachinery/pkg/runtime/schema"
+)
+
+type Action interface {
+	GetNamespace() string
+	GetVerb() string
+	GetResource() schema.GroupVersionResource
+	GetSubresource() string
+	Matches(verb, resource string) bool
+}

testing/kubernetes/testing/fixture.go

@@ -0,0 +1,393 @@
+// Modified from the original at k8s.io/client-go/testing in line with
+// https://github.com/kubernetes/kubernetes/pull/68460
+/*
+Copyright 2015 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package testing
+
+import (
+	"fmt"
+	"sync"
+
+	"k8s.io/apimachinery/pkg/api/errors"
+	"k8s.io/apimachinery/pkg/api/meta"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+	"k8s.io/client-go/testing"
+)
+
+// ObjectTracker keeps track of objects. It is intended to be used to
+// fake calls to a server by returning objects based on their kind,
+// namespace and name.
+type ObjectTracker interface {
+	// Add adds an object to the tracker. If object being added
+	// is a list, its items are added separately.
+	Add(obj runtime.Object) error
+
+	// Get retrieves the object by its kind, namespace and name.
+	Get(gvr schema.GroupVersionResource, ns, name string) (runtime.Object, error)
+
+	// Create adds an object to the tracker in the specified namespace.
+	Create(gvr schema.GroupVersionResource, obj runtime.Object, ns string) error
+
+	// Update updates an existing object in the tracker in the specified namespace.
+	Update(gvr schema.GroupVersionResource, obj runtime.Object, ns string) error
+
+	// List retrieves all objects of a given kind in the given
+	// namespace. Only non-List kinds are accepted.
+	List(gvr schema.GroupVersionResource, gvk schema.GroupVersionKind, ns string) (runtime.Object, error)
+
+	// Delete deletes an existing object from the tracker. If object
+	// didn't exist in the tracker prior to deletion, Delete returns
+	// no error.
+	Delete(gvr schema.GroupVersionResource, ns, name string) error
+}
+
+// ObjectScheme abstracts the implementation of common operations on objects.
+type ObjectScheme interface {
+	runtime.ObjectCreater
+	runtime.ObjectTyper
+}
+
+// ObjectReaction returns a ReactionFunc that applies core.Action to
+// the given tracker.
+func ObjectReaction(tracker ObjectTracker) testing.ReactionFunc {
+	return func(action testing.Action) (bool, runtime.Object, error) {
+		ns := action.GetNamespace()
+		gvr := action.GetResource()
+
+		// Here and below we need to switch on implementation types,
+		// not on interfaces, as some interfaces are identical
+		// (e.g. UpdateAction and CreateAction), so if we use them,
+		// updates and creates end up matching the same case branch.
+		switch action := action.(type) {
+
+		case testing.ListActionImpl:
+			obj, err := tracker.List(gvr, action.GetKind(), ns)
+			return true, obj, err
+
+		case testing.GetActionImpl:
+			obj, err := tracker.Get(gvr, ns, action.GetName())
+			return true, obj, err
+
+		case testing.CreateActionImpl:
+			objMeta, err := meta.Accessor(action.GetObject())
+			if err != nil {
+				return true, nil, err
+			}
+			if action.GetSubresource() == "" {
+				err = tracker.Create(gvr, action.GetObject(), ns)
+			} else {
+				// TODO: Currently we're handling subresource creation as an update
+				// on the enclosing resource. This works for some subresources but
+				// might not be generic enough.
+				err = tracker.Update(gvr, action.GetObject(), ns)
+			}
+			if err != nil {
+				return true, nil, err
+			}
+			obj, err := tracker.Get(gvr, ns, objMeta.GetName())
+			return true, obj, err
+
+		case testing.UpdateActionImpl:
+			objMeta, err := meta.Accessor(action.GetObject())
+			if err != nil {
+				return true, nil, err
+			}
+			err = tracker.Update(gvr, action.GetObject(), ns)
+			if err != nil {
+				if errors.IsConflict(err) { // return the currently-stored version of the object
+					obj, _ := tracker.Get(gvr, ns, objMeta.GetName())
+					return true, obj, err
+				}
+				return true, nil, err
+			}
+			obj, err := tracker.Get(gvr, ns, objMeta.GetName())
+			return true, obj, err
+
+		case testing.DeleteActionImpl:
+			err := tracker.Delete(gvr, ns, action.GetName())
+			if err != nil {
+				return true, nil, err
+			}
+			return true, nil, nil
+
+		default:
+			return false, nil, fmt.Errorf("no reaction implemented for %s", action)
+		}
+	}
+}
+
+type tracker struct {
+	scheme  ObjectScheme
+	decoder runtime.Decoder
+	lock    sync.RWMutex
+	objects map[schema.GroupVersionResource][]runtime.Object
+	counter int
+}
+
+var _ ObjectTracker = &tracker{}
+
+// NewObjectTracker returns an ObjectTracker that can be used to keep track
+// of objects for the fake clientset. Mostly useful for unit tests.
+func NewObjectTracker(scheme ObjectScheme, decoder runtime.Decoder) ObjectTracker {
+	return &tracker{
+		scheme:  scheme,
+		decoder: decoder,
+		objects: make(map[schema.GroupVersionResource][]runtime.Object),
+	}
+}
+
+func (t *tracker) List(gvr schema.GroupVersionResource, gvk schema.GroupVersionKind, ns string) (runtime.Object, error) {
+	// Heuristic for list kind: original kind + List suffix. Might
+	// not always be true but this tracker has a pretty limited
+	// understanding of the actual API model.
+	listGVK := gvk
+	listGVK.Kind = listGVK.Kind + "List"
+	// GVK does have the concept of "internal version". The scheme recognizes
+	// the runtime.APIVersionInternal, but not the empty string.
+	if listGVK.Version == "" {
+		listGVK.Version = runtime.APIVersionInternal
+	}
+
+	list, err := t.scheme.New(listGVK)
+	if err != nil {
+		return nil, err
+	}
+
+	if !meta.IsListType(list) {
+		return nil, fmt.Errorf("%q is not a list type", listGVK.Kind)
+	}
+
+	t.lock.RLock()
+	defer t.lock.RUnlock()
+
+	objs, ok := t.objects[gvr]
+	if !ok {
+		return list, nil
+	}
+
+	matchingObjs, err := filterByNamespaceAndName(objs, ns, "")
+	if err != nil {
+		return nil, err
+	}
+	if err := meta.SetList(list, matchingObjs); err != nil {
+		return nil, err
+	}
+	return list.DeepCopyObject(), nil
+}
+
+func (t *tracker) Get(gvr schema.GroupVersionResource, ns, name string) (runtime.Object, error) {
+	errNotFound := errors.NewNotFound(gvr.GroupResource(), name)
+
+	t.lock.RLock()
+	defer t.lock.RUnlock()
+
+	objs, ok := t.objects[gvr]
+	if !ok {
+		return nil, errNotFound
+	}
+
+	matchingObjs, err := filterByNamespaceAndName(objs, ns, name)
+	if err != nil {
+		return nil, err
+	}
+	if len(matchingObjs) == 0 {
+		return nil, errNotFound
+	}
+	if len(matchingObjs) > 1 {
+		return nil, fmt.Errorf("more than one object matched gvr %s, ns: %q name: %q", gvr, ns, name)
+	}
+
+	// Only one object should match in the tracker if it works
+	// correctly, as Add/Update methods enforce kind/namespace/name
+	// uniqueness.
+	obj := matchingObjs[0].DeepCopyObject()
+	if status, ok := obj.(*metav1.Status); ok {
+		if status.Status != metav1.StatusSuccess {
+			return nil, &errors.StatusError{ErrStatus: *status}
+		}
+	}
+
+	return obj, nil
+}
+
+func (t *tracker) Add(obj runtime.Object) error {
+	if meta.IsListType(obj) {
+		return t.addList(obj, false)
+	}
+	objMeta, err := meta.Accessor(obj)
+	if err != nil {
+		return err
+	}
+	gvks, _, err := t.scheme.ObjectKinds(obj)
+	if err != nil {
+		return err
+	}
+	if len(gvks) == 0 {
+		return fmt.Errorf("no registered kinds for %v", obj)
+	}
+	for _, gvk := range gvks {
+		// NOTE: UnsafeGuessKindToResource is a heuristic and default match. The
+		// actual registration in apiserver can specify arbitrary route for a
+		// gvk. If a test uses such objects, it cannot preset the tracker with
+		// objects via Add(). Instead, it should trigger the Create() function
+		// of the tracker, where an arbitrary gvr can be specified.
+		gvr, _ := meta.UnsafeGuessKindToResource(gvk)
+		// Resource doesn't have the concept of "__internal" version, just set it to "".
+		if gvr.Version == runtime.APIVersionInternal {
+			gvr.Version = ""
+		}
+
+		err := t.add(gvr, obj, objMeta.GetNamespace(), false)
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func (t *tracker) Create(gvr schema.GroupVersionResource, obj runtime.Object, ns string) error {
+	return t.add(gvr, obj, ns, false)
+}
+
+func (t *tracker) Update(gvr schema.GroupVersionResource, obj runtime.Object, ns string) error {
+	return t.add(gvr, obj, ns, true)
+}
+
+func (t *tracker) add(gvr schema.GroupVersionResource, obj runtime.Object, ns string, replaceExisting bool) error {
+	t.lock.Lock()
+	defer t.lock.Unlock()
+
+	gr := gvr.GroupResource()
+
+	// To avoid the object from being accidentally modified by caller
+	// after it's been added to the tracker, we always store the deep
+	// copy.
+	obj = obj.DeepCopyObject()
+
+	newMeta, err := meta.Accessor(obj)
+	if err != nil {
+		return err
+	}
+
+	// Propagate namespace to the new object if hasn't already been set.
+	if len(newMeta.GetNamespace()) == 0 {
+		newMeta.SetNamespace(ns)
+	}
+
+	// Update version on new object
+	inputResourceVersion := newMeta.GetResourceVersion()
+	newMeta.SetResourceVersion(fmt.Sprintf("%d", t.counter))
+	t.counter++
+
+	if ns != newMeta.GetNamespace() {
+		msg := fmt.Sprintf("request namespace does not match object namespace, request: %q object: %q", ns, newMeta.GetNamespace())
+		return errors.NewBadRequest(msg)
+	}
+
+	for i, existingObj := range t.objects[gvr] {
+		oldMeta, err := meta.Accessor(existingObj)
+		if err != nil {
+			return err
+		}
+		if oldMeta.GetNamespace() == newMeta.GetNamespace() && oldMeta.GetName() == newMeta.GetName() {
+			if replaceExisting {
+				if oldMeta.GetResourceVersion() != inputResourceVersion {
+					return errors.NewConflict(gr, newMeta.GetName(), fmt.Errorf("Resource version mismatch"))
+				}
+				t.objects[gvr][i] = obj
+				return nil
+			}
+			return errors.NewAlreadyExists(gr, newMeta.GetName())
+		}
+	}
+
+	if replaceExisting {
+		// Tried to update but no matching object was found.
+		return errors.NewNotFound(gr, newMeta.GetName())
+	}
+
+	t.objects[gvr] = append(t.objects[gvr], obj)
+
+	return nil
+}
+
+func (t *tracker) addList(obj runtime.Object, replaceExisting bool) error {
+	list, err := meta.ExtractList(obj)
+	if err != nil {
+		return err
+	}
+	errs := runtime.DecodeList(list, t.decoder)
+	if len(errs) > 0 {
+		return errs[0]
+	}
+	for _, obj := range list {
+		if err := t.Add(obj); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func (t *tracker) Delete(gvr schema.GroupVersionResource, ns, name string) error {
+	t.lock.Lock()
+	defer t.lock.Unlock()
+
+	found := false
+
+	for i, existingObj := range t.objects[gvr] {
+		objMeta, err := meta.Accessor(existingObj)
+		if err != nil {
+			return err
+		}
+		if objMeta.GetNamespace() == ns && objMeta.GetName() == name {
+			t.objects[gvr] = append(t.objects[gvr][:i], t.objects[gvr][i+1:]...)
+			found = true
+			break
+		}
+	}
+
+	if found {
+		return nil
+	}
+
+	return errors.NewNotFound(gvr.GroupResource(), name)
+}
+
+// filterByNamespaceAndName returns all objects in the collection that
+// match provided namespace and name. Empty namespace matches
+// non-namespaced objects.
+func filterByNamespaceAndName(objs []runtime.Object, ns, name string) ([]runtime.Object, error) {
+	var res []runtime.Object
+
+	for _, obj := range objs {
+		acc, err := meta.Accessor(obj)
+		if err != nil {
+			return nil, err
+		}
+		if ns != "" && acc.GetNamespace() != ns {
+			continue
+		}
+		if name != "" && acc.GetName() != name {
+			continue
+		}
+		res = append(res, obj)
+	}
+
+	return res, nil
+}

vagrant-common.rb

@@ -16,7 +16,7 @@ def ansibleize(h)
 end
 
 def get_go_version_from_build_dockerfile()
-  go_regexp = /FROM golang:(\S*).*?/
+  go_regexp = /FROM golang:([\d\.]*).*?/
   dockerfile_path = File.expand_path(File.join(File.dirname(__FILE__), 'build', 'Dockerfile'))
   go_version = File.readlines(dockerfile_path).select { |line| line.match(go_regexp) }.first.match(go_regexp).captures.first
   if go_version.nil?

vendor/README.md

@@ -1,26 +0,0 @@
-
-# Managing Vendored Dependencies with `git submodule`
-
-These operations result in staged but uncommitted changes to your
-branch; you will need to commit them as normal. Execute them in the
-root of your checkout.
-
-## Adding a New Dependency
-
-    ~/weave$ git submodule add https://example.com/organisation/module vendor/example.com/organisation/module
-
-## Update All Dependencies to Upstream `master`
-
-    ~/weave$ git submodule foreach git pull origin master
-
-## Force a Specific Dependency to a Particular Commit
-
-	~/weave$ (cd vendor/example.com/organisation/module && git checkout <commit-ish>)
-	~/weave$ git add vendor/example.com/organisation/module
-
-## Remove a Dependency
-
-	~/weave$ git submodule deinit vendor/example.com/organisation/module
-    ~/weave$ git rm vendor/example.com/organisation/module
-	~/weave$ rm -rf .git/modules/vendor/example.con/organisation/module
-

vendor/github.com/Azure/go-ansiterm/LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2015 Microsoft Corporation
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

vendor/github.com/Azure/go-ansiterm/README.md

@@ -0,0 +1,12 @@
+# go-ansiterm
+
+This is a cross platform Ansi Terminal Emulation library.  It reads a stream of Ansi characters and produces the appropriate function calls.  The results of the function calls are platform dependent.
+
+For example the parser might receive "ESC, [, A" as a stream of three characters.  This is the code for Cursor Up (http://www.vt100.net/docs/vt510-rm/CUU).  The parser then calls the cursor up function (CUU()) on an event handler.  The event handler determines what platform specific work must be done to cause the cursor to move up one position.
+
+The parser (parser.go) is a partial implementation of this state machine (http://vt100.net/emu/vt500_parser.png).  There are also two event handler implementations, one for tests (test_event_handler.go) to validate that the expected events are being produced and called, the other is a Windows implementation (winterm/win_event_handler.go).
+
+See parser_test.go for examples exercising the state machine and generating appropriate function calls.
+
+-----
+This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/). For more information see the [Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/) or contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional questions or comments.

vendor/github.com/Azure/go-ansiterm/constants.go

@@ -0,0 +1,188 @@
+package ansiterm
+
+const LogEnv = "DEBUG_TERMINAL"
+
+// ANSI constants
+// References:
+// -- http://www.ecma-international.org/publications/standards/Ecma-048.htm
+// -- http://man7.org/linux/man-pages/man4/console_codes.4.html
+// -- http://manpages.ubuntu.com/manpages/intrepid/man4/console_codes.4.html
+// -- http://en.wikipedia.org/wiki/ANSI_escape_code
+// -- http://vt100.net/emu/dec_ansi_parser
+// -- http://vt100.net/emu/vt500_parser.svg
+// -- http://invisible-island.net/xterm/ctlseqs/ctlseqs.html
+// -- http://www.inwap.com/pdp10/ansicode.txt
+const (
+	// ECMA-48 Set Graphics Rendition
+	// Note:
+	// -- Constants leading with an underscore (e.g., _ANSI_xxx) are unsupported or reserved
+	// -- Fonts could possibly be supported via SetCurrentConsoleFontEx
+	// -- Windows does not expose the per-window cursor (i.e., caret) blink times
+	ANSI_SGR_RESET              = 0
+	ANSI_SGR_BOLD               = 1
+	ANSI_SGR_DIM                = 2
+	_ANSI_SGR_ITALIC            = 3
+	ANSI_SGR_UNDERLINE          = 4
+	_ANSI_SGR_BLINKSLOW         = 5
+	_ANSI_SGR_BLINKFAST         = 6
+	ANSI_SGR_REVERSE            = 7
+	_ANSI_SGR_INVISIBLE         = 8
+	_ANSI_SGR_LINETHROUGH       = 9
+	_ANSI_SGR_FONT_00           = 10
+	_ANSI_SGR_FONT_01           = 11
+	_ANSI_SGR_FONT_02           = 12
+	_ANSI_SGR_FONT_03           = 13
+	_ANSI_SGR_FONT_04           = 14
+	_ANSI_SGR_FONT_05           = 15
+	_ANSI_SGR_FONT_06           = 16
+	_ANSI_SGR_FONT_07           = 17
+	_ANSI_SGR_FONT_08           = 18
+	_ANSI_SGR_FONT_09           = 19
+	_ANSI_SGR_FONT_10           = 20
+	_ANSI_SGR_DOUBLEUNDERLINE   = 21
+	ANSI_SGR_BOLD_DIM_OFF       = 22
+	_ANSI_SGR_ITALIC_OFF        = 23
+	ANSI_SGR_UNDERLINE_OFF      = 24
+	_ANSI_SGR_BLINK_OFF         = 25
+	_ANSI_SGR_RESERVED_00       = 26
+	ANSI_SGR_REVERSE_OFF        = 27
+	_ANSI_SGR_INVISIBLE_OFF     = 28
+	_ANSI_SGR_LINETHROUGH_OFF   = 29
+	ANSI_SGR_FOREGROUND_BLACK   = 30
+	ANSI_SGR_FOREGROUND_RED     = 31
+	ANSI_SGR_FOREGROUND_GREEN   = 32
+	ANSI_SGR_FOREGROUND_YELLOW  = 33
+	ANSI_SGR_FOREGROUND_BLUE    = 34
+	ANSI_SGR_FOREGROUND_MAGENTA = 35
+	ANSI_SGR_FOREGROUND_CYAN    = 36
+	ANSI_SGR_FOREGROUND_WHITE   = 37
+	_ANSI_SGR_RESERVED_01       = 38
+	ANSI_SGR_FOREGROUND_DEFAULT = 39
+	ANSI_SGR_BACKGROUND_BLACK   = 40
+	ANSI_SGR_BACKGROUND_RED     = 41
+	ANSI_SGR_BACKGROUND_GREEN   = 42
+	ANSI_SGR_BACKGROUND_YELLOW  = 43
+	ANSI_SGR_BACKGROUND_BLUE    = 44
+	ANSI_SGR_BACKGROUND_MAGENTA = 45
+	ANSI_SGR_BACKGROUND_CYAN    = 46
+	ANSI_SGR_BACKGROUND_WHITE   = 47
+	_ANSI_SGR_RESERVED_02       = 48
+	ANSI_SGR_BACKGROUND_DEFAULT = 49
+	// 50 - 65: Unsupported
+
+	ANSI_MAX_CMD_LENGTH = 4096
+
+	MAX_INPUT_EVENTS = 128
+	DEFAULT_WIDTH    = 80
+	DEFAULT_HEIGHT   = 24
+
+	ANSI_BEL              = 0x07
+	ANSI_BACKSPACE        = 0x08
+	ANSI_TAB              = 0x09
+	ANSI_LINE_FEED        = 0x0A
+	ANSI_VERTICAL_TAB     = 0x0B
+	ANSI_FORM_FEED        = 0x0C
+	ANSI_CARRIAGE_RETURN  = 0x0D
+	ANSI_ESCAPE_PRIMARY   = 0x1B
+	ANSI_ESCAPE_SECONDARY = 0x5B
+	ANSI_OSC_STRING_ENTRY = 0x5D
+	ANSI_COMMAND_FIRST    = 0x40
+	ANSI_COMMAND_LAST     = 0x7E
+	DCS_ENTRY             = 0x90
+	CSI_ENTRY             = 0x9B
+	OSC_STRING            = 0x9D
+	ANSI_PARAMETER_SEP    = ";"
+	ANSI_CMD_G0           = '('
+	ANSI_CMD_G1           = ')'
+	ANSI_CMD_G2           = '*'
+	ANSI_CMD_G3           = '+'
+	ANSI_CMD_DECPNM       = '>'
+	ANSI_CMD_DECPAM       = '='
+	ANSI_CMD_OSC          = ']'
+	ANSI_CMD_STR_TERM     = '\\'
+
+	KEY_CONTROL_PARAM_2 = ";2"
+	KEY_CONTROL_PARAM_3 = ";3"
+	KEY_CONTROL_PARAM_4 = ";4"
+	KEY_CONTROL_PARAM_5 = ";5"
+	KEY_CONTROL_PARAM_6 = ";6"
+	KEY_CONTROL_PARAM_7 = ";7"
+	KEY_CONTROL_PARAM_8 = ";8"
+	KEY_ESC_CSI         = "\x1B["
+	KEY_ESC_N           = "\x1BN"
+	KEY_ESC_O           = "\x1BO"
+
+	FILL_CHARACTER = ' '
+)
+
+func getByteRange(start byte, end byte) []byte {
+	bytes := make([]byte, 0, 32)
+	for i := start; i <= end; i++ {
+		bytes = append(bytes, byte(i))
+	}
+
+	return bytes
+}
+
+var toGroundBytes = getToGroundBytes()
+var executors = getExecuteBytes()
+
+// SPACE		  20+A0 hex  Always and everywhere a blank space
+// Intermediate	  20-2F hex   !"#$%&'()*+,-./
+var intermeds = getByteRange(0x20, 0x2F)
+
+// Parameters	  30-3F hex  0123456789:;<=>?
+// CSI Parameters 30-39, 3B hex 0123456789;
+var csiParams = getByteRange(0x30, 0x3F)
+
+var csiCollectables = append(getByteRange(0x30, 0x39), getByteRange(0x3B, 0x3F)...)
+
+// Uppercase	  40-5F hex  @ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_
+var upperCase = getByteRange(0x40, 0x5F)
+
+// Lowercase	  60-7E hex  `abcdefghijlkmnopqrstuvwxyz{|}~
+var lowerCase = getByteRange(0x60, 0x7E)
+
+// Alphabetics	  40-7E hex  (all of upper and lower case)
+var alphabetics = append(upperCase, lowerCase...)
+
+var printables = getByteRange(0x20, 0x7F)
+
+var escapeIntermediateToGroundBytes = getByteRange(0x30, 0x7E)
+var escapeToGroundBytes = getEscapeToGroundBytes()
+
+// See http://www.vt100.net/emu/vt500_parser.png for description of the complex
+// byte ranges below
+
+func getEscapeToGroundBytes() []byte {
+	escapeToGroundBytes := getByteRange(0x30, 0x4F)
+	escapeToGroundBytes = append(escapeToGroundBytes, getByteRange(0x51, 0x57)...)
+	escapeToGroundBytes = append(escapeToGroundBytes, 0x59)
+	escapeToGroundBytes = append(escapeToGroundBytes, 0x5A)
+	escapeToGroundBytes = append(escapeToGroundBytes, 0x5C)
+	escapeToGroundBytes = append(escapeToGroundBytes, getByteRange(0x60, 0x7E)...)
+	return escapeToGroundBytes
+}
+
+func getExecuteBytes() []byte {
+	executeBytes := getByteRange(0x00, 0x17)
+	executeBytes = append(executeBytes, 0x19)
+	executeBytes = append(executeBytes, getByteRange(0x1C, 0x1F)...)
+	return executeBytes
+}
+
+func getToGroundBytes() []byte {
+	groundBytes := []byte{0x18}
+	groundBytes = append(groundBytes, 0x1A)
+	groundBytes = append(groundBytes, getByteRange(0x80, 0x8F)...)
+	groundBytes = append(groundBytes, getByteRange(0x91, 0x97)...)
+	groundBytes = append(groundBytes, 0x99)
+	groundBytes = append(groundBytes, 0x9A)
+	groundBytes = append(groundBytes, 0x9C)
+	return groundBytes
+}
+
+// Delete		     7F hex  Always and everywhere ignored
+// C1 Control	  80-9F hex  32 additional control characters
+// G1 Displayable A1-FE hex  94 additional displayable characters
+// Special		  A0+FF hex  Same as SPACE and DELETE

vendor/github.com/Azure/go-ansiterm/context.go

@@ -0,0 +1,7 @@
+package ansiterm
+
+type ansiContext struct {
+	currentChar byte
+	paramBuffer []byte
+	interBuffer []byte
+}

vendor/github.com/Azure/go-ansiterm/csi_entry_state.go

@@ -0,0 +1,49 @@
+package ansiterm
+
+type csiEntryState struct {
+	baseState
+}
+
+func (csiState csiEntryState) Handle(b byte) (s state, e error) {
+	csiState.parser.logf("CsiEntry::Handle %#x", b)
+
+	nextState, err := csiState.baseState.Handle(b)
+	if nextState != nil || err != nil {
+		return nextState, err
+	}
+
+	switch {
+	case sliceContains(alphabetics, b):
+		return csiState.parser.ground, nil
+	case sliceContains(csiCollectables, b):
+		return csiState.parser.csiParam, nil
+	case sliceContains(executors, b):
+		return csiState, csiState.parser.execute()
+	}
+
+	return csiState, nil
+}
+
+func (csiState csiEntryState) Transition(s state) error {
+	csiState.parser.logf("CsiEntry::Transition %s --> %s", csiState.Name(), s.Name())
+	csiState.baseState.Transition(s)
+
+	switch s {
+	case csiState.parser.ground:
+		return csiState.parser.csiDispatch()
+	case csiState.parser.csiParam:
+		switch {
+		case sliceContains(csiParams, csiState.parser.context.currentChar):
+			csiState.parser.collectParam()
+		case sliceContains(intermeds, csiState.parser.context.currentChar):
+			csiState.parser.collectInter()
+		}
+	}
+
+	return nil
+}
+
+func (csiState csiEntryState) Enter() error {
+	csiState.parser.clear()
+	return nil
+}

vendor/github.com/Azure/go-ansiterm/csi_param_state.go

@@ -0,0 +1,38 @@
+package ansiterm
+
+type csiParamState struct {
+	baseState
+}
+
+func (csiState csiParamState) Handle(b byte) (s state, e error) {
+	csiState.parser.logf("CsiParam::Handle %#x", b)
+
+	nextState, err := csiState.baseState.Handle(b)
+	if nextState != nil || err != nil {
+		return nextState, err
+	}
+
+	switch {
+	case sliceContains(alphabetics, b):
+		return csiState.parser.ground, nil
+	case sliceContains(csiCollectables, b):
+		csiState.parser.collectParam()
+		return csiState, nil
+	case sliceContains(executors, b):
+		return csiState, csiState.parser.execute()
+	}
+
+	return csiState, nil
+}
+
+func (csiState csiParamState) Transition(s state) error {
+	csiState.parser.logf("CsiParam::Transition %s --> %s", csiState.Name(), s.Name())
+	csiState.baseState.Transition(s)
+
+	switch s {
+	case csiState.parser.ground:
+		return csiState.parser.csiDispatch()
+	}
+
+	return nil
+}

vendor/github.com/Azure/go-ansiterm/escape_intermediate_state.go

@@ -0,0 +1,36 @@
+package ansiterm
+
+type escapeIntermediateState struct {
+	baseState
+}
+
+func (escState escapeIntermediateState) Handle(b byte) (s state, e error) {
+	escState.parser.logf("escapeIntermediateState::Handle %#x", b)
+	nextState, err := escState.baseState.Handle(b)
+	if nextState != nil || err != nil {
+		return nextState, err
+	}
+
+	switch {
+	case sliceContains(intermeds, b):
+		return escState, escState.parser.collectInter()
+	case sliceContains(executors, b):
+		return escState, escState.parser.execute()
+	case sliceContains(escapeIntermediateToGroundBytes, b):
+		return escState.parser.ground, nil
+	}
+
+	return escState, nil
+}
+
+func (escState escapeIntermediateState) Transition(s state) error {
+	escState.parser.logf("escapeIntermediateState::Transition %s --> %s", escState.Name(), s.Name())
+	escState.baseState.Transition(s)
+
+	switch s {
+	case escState.parser.ground:
+		return escState.parser.escDispatch()
+	}
+
+	return nil
+}

vendor/github.com/Azure/go-ansiterm/escape_state.go

@@ -0,0 +1,47 @@
+package ansiterm
+
+type escapeState struct {
+	baseState
+}
+
+func (escState escapeState) Handle(b byte) (s state, e error) {
+	escState.parser.logf("escapeState::Handle %#x", b)
+	nextState, err := escState.baseState.Handle(b)
+	if nextState != nil || err != nil {
+		return nextState, err
+	}
+
+	switch {
+	case b == ANSI_ESCAPE_SECONDARY:
+		return escState.parser.csiEntry, nil
+	case b == ANSI_OSC_STRING_ENTRY:
+		return escState.parser.oscString, nil
+	case sliceContains(executors, b):
+		return escState, escState.parser.execute()
+	case sliceContains(escapeToGroundBytes, b):
+		return escState.parser.ground, nil
+	case sliceContains(intermeds, b):
+		return escState.parser.escapeIntermediate, nil
+	}
+
+	return escState, nil
+}
+
+func (escState escapeState) Transition(s state) error {
+	escState.parser.logf("Escape::Transition %s --> %s", escState.Name(), s.Name())
+	escState.baseState.Transition(s)
+
+	switch s {
+	case escState.parser.ground:
+		return escState.parser.escDispatch()
+	case escState.parser.escapeIntermediate:
+		return escState.parser.collectInter()
+	}
+
+	return nil
+}
+
+func (escState escapeState) Enter() error {
+	escState.parser.clear()
+	return nil
+}

vendor/github.com/Azure/go-ansiterm/event_handler.go

@@ -0,0 +1,90 @@
+package ansiterm
+
+type AnsiEventHandler interface {
+	// Print
+	Print(b byte) error
+
+	// Execute C0 commands
+	Execute(b byte) error
+
+	// CUrsor Up
+	CUU(int) error
+
+	// CUrsor Down
+	CUD(int) error
+
+	// CUrsor Forward
+	CUF(int) error
+
+	// CUrsor Backward
+	CUB(int) error
+
+	// Cursor to Next Line
+	CNL(int) error
+
+	// Cursor to Previous Line
+	CPL(int) error
+
+	// Cursor Horizontal position Absolute
+	CHA(int) error
+
+	// Vertical line Position Absolute
+	VPA(int) error
+
+	// CUrsor Position
+	CUP(int, int) error
+
+	// Horizontal and Vertical Position (depends on PUM)
+	HVP(int, int) error
+
+	// Text Cursor Enable Mode
+	DECTCEM(bool) error
+
+	// Origin Mode
+	DECOM(bool) error
+
+	// 132 Column Mode
+	DECCOLM(bool) error
+
+	// Erase in Display
+	ED(int) error
+
+	// Erase in Line
+	EL(int) error
+
+	// Insert Line
+	IL(int) error
+
+	// Delete Line
+	DL(int) error
+
+	// Insert Character
+	ICH(int) error
+
+	// Delete Character
+	DCH(int) error
+
+	// Set Graphics Rendition
+	SGR([]int) error
+
+	// Pan Down
+	SU(int) error
+
+	// Pan Up
+	SD(int) error
+
+	// Device Attributes
+	DA([]string) error
+
+	// Set Top and Bottom Margins
+	DECSTBM(int, int) error
+
+	// Index
+	IND() error
+
+	// Reverse Index
+	RI() error
+
+	// Flush updates from previous commands
+	Flush() error
+}

vendor/github.com/Azure/go-ansiterm/ground_state.go

@@ -0,0 +1,24 @@
+package ansiterm
+
+type groundState struct {
+	baseState
+}
+
+func (gs groundState) Handle(b byte) (s state, e error) {
+	gs.parser.context.currentChar = b
+
+	nextState, err := gs.baseState.Handle(b)
+	if nextState != nil || err != nil {
+		return nextState, err
+	}
+
+	switch {
+	case sliceContains(printables, b):
+		return gs, gs.parser.print()
+
+	case sliceContains(executors, b):
+		return gs, gs.parser.execute()
+	}
+
+	return gs, nil
+}

vendor/github.com/Azure/go-ansiterm/osc_string_state.go

@@ -0,0 +1,31 @@
+package ansiterm
+
+type oscStringState struct {
+	baseState
+}
+
+func (oscState oscStringState) Handle(b byte) (s state, e error) {
+	oscState.parser.logf("OscString::Handle %#x", b)
+	nextState, err := oscState.baseState.Handle(b)
+	if nextState != nil || err != nil {
+		return nextState, err
+	}
+
+	switch {
+	case isOscStringTerminator(b):
+		return oscState.parser.ground, nil
+	}
+
+	return oscState, nil
+}
+
+// See below for OSC string terminators for linux
+// http://man7.org/linux/man-pages/man4/console_codes.4.html
+func isOscStringTerminator(b byte) bool {
+
+	if b == ANSI_BEL || b == 0x5C {
+		return true
+	}
+
+	return false
+}

vendor/github.com/Azure/go-ansiterm/parser.go

@@ -0,0 +1,151 @@
+package ansiterm
+
+import (
+	"errors"
+	"log"
+	"os"
+)
+
+type AnsiParser struct {
+	currState          state
+	eventHandler       AnsiEventHandler
+	context            *ansiContext
+	csiEntry           state
+	csiParam           state
+	dcsEntry           state
+	escape             state
+	escapeIntermediate state
+	error              state
+	ground             state
+	oscString          state
+	stateMap           []state
+
+	logf func(string, ...interface{})
+}
+
+type Option func(*AnsiParser)
+
+func WithLogf(f func(string, ...interface{})) Option {
+	return func(ap *AnsiParser) {
+		ap.logf = f
+	}
+}
+
+func CreateParser(initialState string, evtHandler AnsiEventHandler, opts ...Option) *AnsiParser {
+	ap := &AnsiParser{
+		eventHandler: evtHandler,
+		context:      &ansiContext{},
+	}
+	for _, o := range opts {
+		o(ap)
+	}
+
+	if isDebugEnv := os.Getenv(LogEnv); isDebugEnv == "1" {
+		logFile, _ := os.Create("ansiParser.log")
+		logger := log.New(logFile, "", log.LstdFlags)
+		if ap.logf != nil {
+			l := ap.logf
+			ap.logf = func(s string, v ...interface{}) {
+				l(s, v...)
+				logger.Printf(s, v...)
+			}
+		} else {
+			ap.logf = logger.Printf
+		}
+	}
+
+	if ap.logf == nil {
+		ap.logf = func(string, ...interface{}) {}
+	}
+
+	ap.csiEntry = csiEntryState{baseState{name: "CsiEntry", parser: ap}}
+	ap.csiParam = csiParamState{baseState{name: "CsiParam", parser: ap}}
+	ap.dcsEntry = dcsEntryState{baseState{name: "DcsEntry", parser: ap}}
+	ap.escape = escapeState{baseState{name: "Escape", parser: ap}}
+	ap.escapeIntermediate = escapeIntermediateState{baseState{name: "EscapeIntermediate", parser: ap}}
+	ap.error = errorState{baseState{name: "Error", parser: ap}}
+	ap.ground = groundState{baseState{name: "Ground", parser: ap}}
+	ap.oscString = oscStringState{baseState{name: "OscString", parser: ap}}
+
+	ap.stateMap = []state{
+		ap.csiEntry,
+		ap.csiParam,
+		ap.dcsEntry,
+		ap.escape,
+		ap.escapeIntermediate,
+		ap.error,
+		ap.ground,
+		ap.oscString,
+	}
+
+	ap.currState = getState(initialState, ap.stateMap)
+
+	ap.logf("CreateParser: parser %p", ap)
+	return ap
+}
+
+func getState(name string, states []state) state {
+	for _, el := range states {
+		if el.Name() == name {
+			return el
+		}
+	}
+
+	return nil
+}
+
+func (ap *AnsiParser) Parse(bytes []byte) (int, error) {
+	for i, b := range bytes {
+		if err := ap.handle(b); err != nil {
+			return i, err
+		}
+	}
+
+	return len(bytes), ap.eventHandler.Flush()
+}
+
+func (ap *AnsiParser) handle(b byte) error {
+	ap.context.currentChar = b
+	newState, err := ap.currState.Handle(b)
+	if err != nil {
+		return err
+	}
+
+	if newState == nil {
+		ap.logf("WARNING: newState is nil")
+		return errors.New("New state of 'nil' is invalid.")
+	}
+
+	if newState != ap.currState {
+		if err := ap.changeState(newState); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func (ap *AnsiParser) changeState(newState state) error {
+	ap.logf("ChangeState %s --> %s", ap.currState.Name(), newState.Name())
+
+	// Exit old state
+	if err := ap.currState.Exit(); err != nil {
+		ap.logf("Exit state '%s' failed with : '%v'", ap.currState.Name(), err)
+		return err
+	}
+
+	// Perform transition action
+	if err := ap.currState.Transition(newState); err != nil {
+		ap.logf("Transition from '%s' to '%s' failed with: '%v'", ap.currState.Name(), newState.Name, err)
+		return err
+	}
+
+	// Enter new state
+	if err := newState.Enter(); err != nil {
+		ap.logf("Enter state '%s' failed with: '%v'", newState.Name(), err)
+		return err
+	}
+
+	ap.currState = newState
+	return nil
+}

vendor/github.com/Azure/go-ansiterm/parser_action_helpers.go

@@ -0,0 +1,99 @@
+package ansiterm
+
+import (
+	"strconv"
+)
+
+func parseParams(bytes []byte) ([]string, error) {
+	paramBuff := make([]byte, 0, 0)
+	params := []string{}
+
+	for _, v := range bytes {
+		if v == ';' {
+			if len(paramBuff) > 0 {
+				// Completed parameter, append it to the list
+				s := string(paramBuff)
+				params = append(params, s)
+				paramBuff = make([]byte, 0, 0)
+			}
+		} else {
+			paramBuff = append(paramBuff, v)
+		}
+	}
+
+	// Last parameter may not be terminated with ';'
+	if len(paramBuff) > 0 {
+		s := string(paramBuff)
+		params = append(params, s)
+	}
+
+	return params, nil
+}
+
+func parseCmd(context ansiContext) (string, error) {
+	return string(context.currentChar), nil
+}
+
+func getInt(params []string, dflt int) int {
+	i := getInts(params, 1, dflt)[0]
+	return i
+}
+
+func getInts(params []string, minCount int, dflt int) []int {
+	ints := []int{}
+
+	for _, v := range params {
+		i, _ := strconv.Atoi(v)
+		// Zero is mapped to the default value in VT100.
+		if i == 0 {
+			i = dflt
+		}
+		ints = append(ints, i)
+	}
+
+	if len(ints) < minCount {
+		remaining := minCount - len(ints)
+		for i := 0; i < remaining; i++ {
+			ints = append(ints, dflt)
+		}
+	}
+
+	return ints
+}
+
+func (ap *AnsiParser) modeDispatch(param string, set bool) error {
+	switch param {
+	case "?3":
+		return ap.eventHandler.DECCOLM(set)
+	case "?6":
+		return ap.eventHandler.DECOM(set)
+	case "?25":
+		return ap.eventHandler.DECTCEM(set)
+	}
+	return nil
+}
+
+func (ap *AnsiParser) hDispatch(params []string) error {
+	if len(params) == 1 {
+		return ap.modeDispatch(params[0], true)
+	}
+
+	return nil
+}
+
+func (ap *AnsiParser) lDispatch(params []string) error {
+	if len(params) == 1 {
+		return ap.modeDispatch(params[0], false)
+	}
+
+	return nil
+}
+
+func getEraseParam(params []string) int {
+	param := getInt(params, 0)
+	if param < 0 || 3 < param {
+		param = 0
+	}
+
+	return param
+}