Bryan Boreham
517a2e0e0a
Fix nonzero exit code for a pre-existing plugin on launch;
...
do this purely for the 'plugin running' case, not 'restarting' etc.
2016-01-12 17:06:38 +00:00
Bryan Boreham
bb8c7c599c
Revert "don't launch anything when plugin is running"
...
This reverts commit 3ade796e8d
2016-01-12 16:51:49 +00:00
Bryan Boreham
0d0fc39236
More review feedback
2016-01-12 15:35:34 +00:00
Bryan Boreham
e468399616
Extend the timeout for dead containers to 30 seconds,
...
but remove them if they are destroyed, to reduce starvation
when users are creating and destroying containers quickly.
2016-01-12 14:38:44 +00:00
Bryan Boreham
b2b293a8ee
Added 'ContainerDestroyed()' callback from docker client
2016-01-12 14:38:44 +00:00
Bryan Boreham
2f885a7906
Reclaim all addresses on restart
2016-01-12 14:38:44 +00:00
Bryan Boreham
7985e43e68
Add 'weave restart' command
2016-01-12 14:38:44 +00:00
Bryan Boreham
f84b933d7c
Leave container allocations for a time after death
...
so they will be reclaimed if the container is restarted.
Still cancel pending operations immediately a container dies
2016-01-12 14:38:44 +00:00
Bryan Boreham
4672e9ec1f
Generalise ticker so we can use it for more than paxos,
...
and call tryPendingOps() from it so we eventually re-try operations
2016-01-12 14:38:44 +00:00
Adam Harrison
a5ddba03f7
Merge branch '1.4'
2016-01-12 14:28:16 +00:00
Adam Harrison
302a5a4efc
Merge pull request #1884 from weaveworks/issues/1882-check-after-subscribe
...
Fix race condition in weavewait
2016-01-12 14:28:03 +00:00
Bryan Boreham
3b29e8e6f2
Check for existing interfaces and routes after subscribing to receive new ones
2016-01-12 12:27:55 +00:00
Bryan Boreham
a1dff70ba2
Review feedback
2016-01-11 11:55:11 +00:00
Matthias Radestock
9336cc4121
Merge branch '1.4'
2016-01-11 11:49:49 +00:00
Matthias Radestock
12671be260
Merge branch 'latest_release_doc_updates' into 1.4
2016-01-11 11:49:22 +00:00
Matthias Radestock
d2ad5ddb01
oops
2016-01-11 11:47:22 +00:00
Tom Wilkie
73df093e48
Merge pull request #1864 from weaveworks/shrink-travis-usage
...
Shrink travis usage
2016-01-11 11:27:02 +00:00
Bryan Boreham
9a4adccec6
Remove inconsistent information about restart policy
2016-01-11 11:20:40 +00:00
Matthias Radestock
abea21245d
rename Makefile target
...
since travis is no longer using it
2016-01-11 10:33:33 +00:00
Bryan Boreham
f90390974a
Add page explaining IP addresses, networks and routes
2016-01-11 10:30:37 +00:00
Matthias Radestock
0b4f1b5af0
cosmetic: whitespace
2016-01-11 00:42:14 +00:00
Matthias Radestock
9933a4a368
refactor: tweak overlay & bridge initialisation sequence
...
for clarity
2016-01-11 00:13:29 +00:00
Matthias Radestock
c3f2150b5b
cosmetic: whitespace
2016-01-11 00:03:37 +00:00
Matthias Radestock
46566e2159
refactor: extract sleeve crypto initialisation
2016-01-10 23:52:46 +00:00
Matthias Radestock
1d8d098450
for Travis, this no longer is a go project
...
Travis insists that one picks a project language. There is no go left
in the steps executed by travis. Instead we have Python, Ruby and
shell. Let's pick the first...
Picking Python makes travis unhappy with "Can not perform a '--user'
install. User site-packages are not visible in this
virtualenv". Dropping the '--user' flag fixes that.
2016-01-09 14:21:23 +00:00
Matthias Radestock
1038cbd1a3
shrink Travis usage
...
There is no point running builds, tests, lints on Travis, since we
already do that on Circle.
The only thing we aren't doing (and can't do) on Circle is doc
publishing, so now Travis does just that.
2016-01-09 14:01:27 +00:00
Matthias Radestock
21a2cdd579
Merge branch '1.4'
2016-01-08 18:14:56 +00:00
Matthias Radestock
f598ae297b
Merge branch 'latest_release_doc_updates' into 1.4
2016-01-08 18:14:23 +00:00
Bryan Boreham
4b0fe502f0
Document the --no-multicast-route flag
2016-01-08 17:12:20 +00:00
Matthias Radestock
d9b9ce6bcd
simplify gitignore
2016-01-08 15:23:37 +00:00
Matthias Radestock
4fffa6fa43
fix typo
2016-01-08 15:11:21 +00:00
Tom Wilkie
c0cd9dffc2
Merge pull request #1862 from weaveworks/1651-pause-differently
...
Fix race condition in ipam TestHTTPCancel
2016-01-08 14:07:06 +00:00
Bryan Boreham
f91d7bb7e4
Rewrite TestHTTPCancel to avoid fiddling inside Allocator
2016-01-08 13:37:15 +00:00
Matthias Radestock
3e6f31c8e2
add missing dependencies and netgo checks
2016-01-08 13:27:37 +00:00
Matthias Radestock
19b53cdeb1
refactor: get rid of superfluous rule
2016-01-08 12:06:38 +00:00
Matthias Radestock
c6cc10169f
refactor: put all dependencies in one place
2016-01-08 12:05:48 +00:00
Bryan Boreham
987d89632d
Merge pull request #1847 from weaveworks/1613-weaveutil
...
extract script utilities into single binary; LGTM
2016-01-07 18:53:25 +00:00
Bryan Boreham
8804f88c93
Merge pull request #1856 from weaveworks/no-gossip-sender-gc
...
get rid of GossipSender gc; LTGM
2016-01-07 17:29:54 +00:00
Bryan Boreham
3c654bc036
Merge pull request #1561 from weaveworks/proxy-retry-attach
...
Proxy retry attaching containers; fixes #1556
2016-01-07 17:27:19 +00:00
Paul Bellamy
8ddbb3d8af
[proxy] Keep retrying container attaches until they succeed.
2016-01-07 17:09:34 +00:00
Tom Wilkie
422e9b58cd
Merge pull request #1857 from weaveworks/issues/864-gce-retry
...
Stop masking failures in gce hosts destroy
2016-01-07 14:53:02 +00:00
Matthias Radestock
64a017c972
Merge branch '1.4'
2016-01-07 14:36:08 +00:00
Matthias Radestock
e4295a2da9
don't attach rewrite_etc_hosts helper container to weave
...
...since that is a) inefficent and b) can block when IPAM hasn't had
consensus.
We can only end up in this situation when DOCKER_HOST is pointing at
the weave proxy and we are running `weave run|attach` with
`--local`. So it's very much an edge case.
Fixes #1858 .
2016-01-07 14:32:52 +00:00
Matthias Radestock
216a0fe4fc
refactor: WEAVE_CIDR=none is redundant when --net=host
2016-01-07 14:29:51 +00:00
Bryan Boreham
604f5c599a
Check if there are any machines to destroy, and stop masking failures
2016-01-07 13:40:34 +00:00
Matthias Radestock
9aa4313d37
refactor
2016-01-06 17:38:40 +00:00
Matthias Radestock
969dc4e527
get rid of GossipSender gc
...
With the per-broadcast-source GossipSenders gone, we can get rid of
GossipSender gc...
Instead of maintaining a Connection->GossipSender map in the
GossipChannels - which we then need to gc when connections get closed
- we keep gossip_channel_name->GossipSender maps in LocalConnections.
Access to the map is guarded by a separate lock, rather than re-using
the LocalConnection lock. This reduces lock contention and risk of
deadlock.
In the GossipSenders, we replace the Stop() method with a signalling
channel, i.e. senders stop when that channel gets closed. Handily, the
LocalConnection.finished channel fits that description. So when a
LocalConnection's run loop terminates, any existing and future
associated GossipSenders will end up terminating too.
2016-01-06 17:38:40 +00:00
Bryan Boreham
bbd0b15792
Merge pull request #1855 from weaveworks/gossip-broadcast-stall-3
...
don't stall gossip broadcasting when there are blocked connections; LGTM
2016-01-06 16:43:37 +00:00
Matthias Radestock
99b7fbbc71
optimisation: stop GossipSender on send error
2016-01-05 17:18:41 +00:00
Matthias Radestock
486eb740b5
optimisation: abort GossipSender.deliver when stopped
2016-01-05 17:18:41 +00:00
Bryan Boreham