Bryan Boreham

Is it possible the old one is still running? Given the symptoms I suspect rebooting the node is a good thing to try.

Under Kubernetes, Weave Net is 100% unaware of Docker. Both your errors indicate something else is blocking it; e.g.: FATA: 2019/03/13 13:05:52.327108 Failed to bind metrics server: listen tcp :6781:...

I would suspect the process that is reporting it can’t bind 6781 is not the one that already has it open, reported by netstat. CrashLoopBackoff means it’s dead, exited after...

I wonder if your install was actually running Docker the first time? If, say, it was running cri-o and you go through the steps to shut down Docker and re-add...

> Maybe weave-npc isn't letting go of it's port during the docker restart? That's impossible - the kernel cleans up ownership when the process exits. You have shown in your...

Same thing - there is another instance of Weave Net running and locking the file. Please track down what is running that other instance on your computer.

Thanks for opening the issue. Could you please re-upload the heap profile as a binary rather than text file, since it downloads as 29 bytes for me. You may need...

Looks like something is locked up; can you supply the goroutine profile too please.

Thanks for the PR! Would you consider adding a test, e.g. extending https://github.com/weaveworks/weave/blob/master/test/830_cni_plugin_test.sh (the integration tests at CircleCI always fail for 3rd-party branches, but I've pushed your branch to this...

> I am not sure if it makes sense to configure also `checkAlive` In a Kubernetes context it won't make any difference - that parameter is only used when Weave...