Support archived workflows logs in UI from archived logs

[alexec]

Summary

Currently, if you open an archived workflow, you are not able to see logs in the log viewer, even if the logs were archived to 'main.log'. We could fix this.

Use Cases

When would you use this?

---

Message from the maintainers:

Love this enhancement proposal? Give it a 👍. We prioritise the proposals with the most 👍.

[tczhao]

Hi @alexec , would you mind providing more details on this.
I couldn't get my head around what you meant by "Support archived workflows logs in UI from archived logs".
All I'm thinking right now is we can view the archived logs from archived workflows

[noam-codefresh]

If I understand correctly - the current UI does not make the request to get the logs (output artifacts) by the workflow's UID? because the server's endpoint for it exist.

[asfaltboy]

Does this relate to the Log UI "All" option not showing nothing after pods are garbage collected? Or, should that be considered a separate issue?

On a side note, what does this recommendation in the docs mean?:

⚠️ We do not recommend you rely on Argo Workflows to archive logs. Instead, use a conventional Kubernetes logging facility.

Are there some known issues? Also, does kubernetes provide any facility for collecting and returning logs for garbage collected pods, and then can argo workflows use this facility? I ask because viewing logs in the argo workflows UI is very convenient, and one of the main "selling points" of the application in my mind

[agilgur5]

Does this relate to the Log UI "All" option not showing nothing after pods are garbage collected? Or, should that be considered a separate issue?

Not quite. That is a Kubernetes limitation, not an Argo issue. k8s no longer has logs after the Pods are GC'd.
The archive logs feature is one way around that, as is using a conventional logging facility.

Also, does kubernetes provide any facility for collecting and returning logs for garbage collected pods, and then can argo workflows use this facility?

A pretty conventional cluster set-up is to have fluentd / fluent-bit as a DaemonSet collecting logs from all nodes and sending those to a log provider (e.g. Splunk, Logstash, etc).
That would be an interesting feature if Workflows could use it, as it is indeed convenient. As a workaround to that, you can add an external link to a query in your log provider. Not specific to this issue though.

EDIT: The archive logs docs are getting more specific improvements around recommendations in #12597

[agilgur5]

Not sure if #11121 ended up fixing this as well. cc @terrytangyuan

[terrytangyuan]

It's supported. Re-open if not.