Summary
The user is asking about cleanup actions for aborted tasks, considering options such as "onerror" for workflows, task-level cleanup with agents, and eager mode. They are also thinking about a signal-catching method for cleanup within containers, although this method is uncertain. The user plans to submit a documentation PR if their approach works, noting that terminating from the UI does not seem to have any effect.
josh210
It looks like this doesn't do anything when you terminate from the UI
josh210
Fair, so the on failure is called for aborts too?
kumare
that is by catching signals, but it is not guaranteed
josh210
I think there might actually be another way to do within-container cleanup. I'm trying something out today and I'll open up a documentation PR if it works out
kumare
Or use eager mode
kumare
If you want task level cleanup then agents is the only way
kumare
You can add onerror for workflow
josh210
Is there any way to have a task do some cleanup when it's being aborted?