Run shows as `failed` while step shows as `running`, in truth already failed

[ritazagoni]

On both the history page and when inspecting, a run is showing as failed while the step is showing as running.

Meanwhile the step has already failed with RuntimeError: TypeError: Cannot read properties of undefined (reading '0')

This is a workflow that was triggered by another workflow.

v2.7.19

I have reproduced this locally on main:

• [ ] Yes
• [ x] No

[taylordowns2000]

Could add some sort of trigger that sets steps to failed|lost|crashed|skipped (whatever the right decision is here!) when the run has finished and the channel is closed

[stuartc]

@taylordowns2000 yeah we've been thinking the same; but we need to be really careful about how we do it - we're essentially changing data to something we don't know for certain. So some considerations about what the worker might have been doing or how it might have failed and only apply this policy after a period of time where it would be theoretically impossible for it to change.

[taylordowns2000]

Yeah I think we're on the same page. The ideal is "100% honest and as helpful as possible" - where honesty is more important than helpfulness. So "lost" would be just fine for all these as that's our shorthand for "I don't know what happened."

As in: "that step is :lost because I never got information about whether it succeeded or failed."

[taylordowns2000]

@christad92 , why is this something we wont fix?

[christad92]

I added the tag based on the chat with the Stu when we were reviewing bugs. We had placed in the "Later" Section while dropping the initial comment.

Removing the tag now and adding