Exit handler started before workflow was completed.

[IterableTrucks]

I run hundreds of workflows with exit handler on workflow's completion every day. There's a chance that one of them starts its exit handler before the last task of the workflow is completed. My argo workflow version is 3.4.11

[IterableTrucks]

I put workflow.status variable in the inputs of exit handler task. The ARGO_TEMPLATE environment of the task's pod shows workflow.status is Error. But in the UI, the Inputs/Outputs tab shows the variable is Succeeded

[agilgur5]

It's possible that the UI is delayed or inconsistent. Or there's some scenario the UI isn't handling correctly. It listens to an SSE Event Stream for changes.

If you refresh the UI does it show it as Error correctly?

[IterableTrucks]

No, the UI always shows the status as Succeeded

[agilgur5]

What does the completed Workflow CR in k8s say?
As in, are the UI and k8s consistent?

[IterableTrucks]

Phase status in UI and k8s workflow resource are consistent.

[agilgur5]

Oh, so that would suggest that the exit handler's workflow.status is the incorrect one, as it's the odd one out.

Was there an actual error in the Workflow? Is there anything in the Pod's logs?
Since it only occurs once every few hundred, that suggests a race condition somewhere. I might think that a transient error could perhaps be responsible?