Ability to easily collect all the relevant information when submitting bug report

[antoninbas]

There should be a CLI command that takes care of collecting all the relevant information and can package them in a tarball or upload them somewhere:

• Antrea logs
• Antrea version, Kubernetes version, container runtime information...
• Dump of OVS flows

@frapposelli suggested leveraging https://github.com/vmware-tanzu/crash-diagnostics if possible.

[antoninbas]

This was discussed at the 12/11/2019 Antrea community meeting. We agreed that we should first define what should be collected before writing a tool that collects it in an automated fashion. A few things that were brought-up:

• Antrea logs
• kubelet logs
• kube-proxy logs / config maps
• OVS logs, OVS flows dump, iptables rules dump
• kube-apiserver logs
• which container runtime is used?

Could any sensitive information be exposed by the logs above? Please feel free to comment if you think anything relevant should be added to this list.
@salv-orlando @McCodeman

[weiqiangt]

@antoninbas, beyond you have mentioned, I think we could also execute all commands once and records all results.

I'm actually implementing one on https://github.com/weiqiangt/antrea/tree/antctl-bundle during the implementation of the framework.
For now, it dumps:

• iptables
• ip links
• flow entries
  After dumping them, these data would be packed into a tar.

[McCodeman]

Additionally, we will want to capture uname -a or equivalent and status of ovs kernel module (not sure if the OVS logs above report errors if the kernel module not loaded correctly). We could also offer to optionally dump the k8s network policies.

[McCodeman]

Also, status of IP forwarding in kernel

[antoninbas]

Now that we have the antctl get supportbundle command added by @weiqiangt in #694, I think we can close this issue.

It is not documented yet, but there is an open issue for documentation (#765).