Commit Graph
32 Commits
Author SHA1 Message Date
Bryan Boreham 27848992b1 Merge branch '2.5' 2019-05-07 13:32:55 +00:00
Bryan Boreham 8a4fc9c02a Fix peer-list fuzz test to re-fetch list when it may be nil 2019-05-03 17:25:36 +00:00
Bryan Boreham eb7ba691ee Fetch storedPeerList after read from Kubernetes
This avoids a race condition where the peer has added itself to the
list and the annotation since we read storedPeerList.
2019-04-02 09:11:27 +00:00
Bryan Boreham 15b9b7416a Re-read peers ConfigMap from Kubernetes at start of every reclaim
A previous reclaim may have left the map set to nil.
2019-03-26 17:25:34 +00:00
Bryan Boreham 5024270232 Avoid panic on nil map
It can be set to nil after a failed update; check rather than crashing.
2019-03-26 17:20:24 +00:00
Bryan Boreham bd24dd7dae Make test harness and comments match Kubernetes behaviour on conflict error
Kubernetes always returns nil on error; somehow I had the wrong idea
here.
2019-03-26 17:19:22 +00:00
Bryan Boreham e1e4a7887a Merge branch '2.5' 2019-01-18 11:20:25 +00:00
Bryan Boreham 27178b1c79 Check and report the error return from getKubePeers() 2019-01-15 11:15:36 +00:00
Murali Reddy d2a9a51296 follow project conventions in ordering the imports 2018-12-04 10:58:43 +05:30
Murali Reddy 81b17469f4 prevent kubernetes node connecting to self by excluding the node IP from the list of the peers passed to weaver
Fixes #3398
2018-11-21 11:47:30 +05:30
Murali Reddy c48c746791 replace peerlist with current set of Kubernetes nodes 2018-10-31 18:06:48 +05:30
Murali Reddy f334575e69 perform weave connect replace on node delete with empty peer list 2018-10-30 11:54:59 +05:30
Murali Reddy ea423755cc review comment: removing unnessary verbose code 2018-10-30 10:56:05 +05:30
Murali Reddy 4a40380d0f addressing review comments 2018-10-24 13:45:35 +05:30
Murali Reddy df467be26e watch for node delete events and call reclaimRemovedPeers 2018-10-24 13:45:35 +05:30
Murali Reddy 59c1abe9b9 addressing review comments 2018-10-04 14:08:56 +05:30
Murali Reddy 7a959df648 handles an edge case where peer claimed to own the action (to reclaim) but no longer
exists hence lock persists foever, fix makes a peer own the reclaim when nonexistent node found

Fixes #3386
2018-09-27 16:00:44 +05:30
Bryan Boreham 008eeb03ad Improve comments and error checking 2018-09-12 14:11:09 +00:00
Bryan Boreham 41c48dc678 Update test to call reclaimPeer() 2018-09-12 14:11:09 +00:00
Bryan Boreham 09be35dee0 Fetch stored peer list every time round the retry loop
This ensures we take account of other peers' updates and closes a race
condition whereby another peer's changes can be overwritten.
2018-09-12 14:11:09 +00:00
Bryan Boreham c3ad641d94 Refactor: extract reclaimPeer() function 2018-09-12 14:11:09 +00:00
Bryan Boreham bacb3952a3 Fetch Kubernetes node list every time round reclaim loop
If we retain the Kubernetes list for a long time we might see a peer
appear in the Weave list and conclude it has been deleted. We must
always fetch the Kubernetes list after fetching the Weave list, during
deletion.
2018-09-12 14:10:59 +00:00
Bryan Boreham 70257a6c19 Add test for kube-utils peer list
- In parallel on several goroutines, randomly add and remove peers.
 - Maintain the test's own idea of current state.
 - Check that the test's state and kube-utils' state are consistent.
2018-09-12 12:37:16 +00:00
Bryan Boreham bfe9f00e55 Fake kubernetes API which implements optimistic locking
Mostly copied from kubernetes/client-go files
2018-09-12 12:37:16 +00:00
Bryan Boreham 591be2d08f Refactor: Introduce an interface so weave client can be mocked 2018-09-12 12:37:16 +00:00
Bryan Boreham c3ce052cc5 Refactor: use Kubernetes.Interface instead of *Clientset
So we can call with a mock for testing
2018-09-12 12:35:34 +00:00
Bryan Boreham 7dcd4bdf47 Refactor: extract peerlist fns to new file 2018-09-12 12:35:34 +00:00
Martynas Pumputis 78e00f9a3a Merge pull request #3397 from weaveworks/issues/3394-delete-old-me
Reclaim IPs when node name is re-used
2018-09-12 13:57:12 +03:00
Bryan Boreham 2ff0bbe150 Reclaim IPs when node name is re-used
Change the map used to reclaim peers so it is indexed by PeerName
which should be unique over time, rather than NodeName which can be
re-used.  Create an auxilliary set of NodeNames to help.

This way, if we can see that a different peer was previously running
with the same node name, it can be included for reclaim.
2018-09-10 15:10:51 +00:00
Martynas Pumputis acc24f1237 Do not exclude k8s node without any IP addr in reclaim
Otherwise, such node will be rmpeer'd which can lead to the IPAM
inconsistencies "Received update for IP range I own at <..>".
2018-09-06 14:49:20 +03:00
Martynas Pumputis 25c26519ed Merge remote-tracking branch 'origin/2.3' 2018-06-12 14:47:42 +02:00
Bryan Boreham 2a04df1049 Rename "kube-peers" to "kube-utils"
It's now doing things like network status that are outside of the
scope of "peers".
2018-06-11 11:22:07 +00:00