create .github/ISSUES.md for this repo

[chrislovecnm]

Would be nice to have a list of info to include in a new issue. @marccarre provided this list for k8s, and I added a couple

1. weave or weave-kube version
2. full weave and weave-npc logs
3. full iptables rules
4. full dmesg logs
5. full lsb_release -a output
6. uname -a
7. kubernetes verion

What other debugging information is helpful??

[chrislovecnm]

Also with weave-kube I do not know how to run the status commands with weave.

[faraazkhan]

You can run the weave commands as follows @chrislovecnm

kubectl -n kube-system exec -it weave-net-6af9j -- /home/weave/weave --local report > weave_report.json

Where weave-net-6af9j is the name of the pod you are targeting.

[scottvidmar]

I think we need to include more asks around basic information that proves out where the issue likely resides, thus providing context to weave engineers:

• If there are actual network errors (TCP handshakes aren't completing, etc...), provide the capture data illustrating the issue.
• Context around what communication is being attempted, and what the machine is attempting to connect to.
• Verify connectivity from machine, and verify connectivity from another machine in the same subnet.
• Full description of what's being seen, including which direction(s) the communication is seeing issues (ingress/egress/bi-directional).
• System metric information around saturation and utilization. If physical machines, include information about resources.

I could be missing something from here, and iperf UDP test data could be another proof that an issue is actually being seen. I typically take the approach of (1) prove the problem exists, (2) gather context around the problem, (3) understand behavior and drill to issue. This is an oversimplification of the USE method curated by Brendan Gregg (http://brendangregg.com/usemethod.html) and has provided me with success.

[brb]

Thanks for the suggestions. Such list could reduce the number of round-trips while debugging an issue.

[bboreham]

We added a template in #2873; it doesn't cover anything like what was discussed above, but it's a start