What you expected to happen?
weaver detects/claims pre-existing IPs attached to the weave bridge on node via
|
func findExistingAddresses(dockerCli *weavedocker.Client, bridgeName string) (addrs []ipam.PreClaim, err error) { |
What happened?
weaver stopped using the host process namespace (via hotsPID) as of #3876, which was needed for detecting devices via the peer ids / process ids here:
|
netDevs, err := weavenet.GetNetDevsByVethPeerIds(pid, peerIDs) |
.
Specifically, given the change of #3876, the following command only detects processes within the PID namespace of the weave-net pod:
|
pids, err := common.AllPids("/proc") |
How to reproduce it?
- Move the boltdb file so that the persisted data gets cleared
- Rotate the
weave-net pod in the kube-system namespace
- Rotate any pod once the
weave-net pod comes back up
- Notice how the IP assigned to the pod has already been assigned to another pod on the cluster
Anything else we need to know?
This is specifically for weave running on k8s.
Versions:
weave version 2.8.1
What you expected to happen?
weaverdetects/claims pre-existing IPs attached to theweavebridge on node viaweave/prog/weaver/reclaim.go
Line 26 in 8c84763
What happened?
weaverstopped using the host process namespace (via hotsPID) as of #3876, which was needed for detecting devices via the peer ids / process ids here:weave/prog/weaver/reclaim.go
Line 84 in 8c84763
Specifically, given the change of #3876, the following command only detects processes within the PID namespace of the
weave-netpod:weave/prog/weaver/reclaim.go
Line 79 in 8c84763
How to reproduce it?
weave-netpod in thekube-systemnamespaceweave-netpod comes back upAnything else we need to know?
This is specifically for
weaverunning onk8s.Versions:
weave version 2.8.1