Closed mookerji closed 5 years ago
Nice patch! Yes, please add a test case and I'll merge it ASAP
Yves
@ybastide thanks for the quick reply, will do!
I have a question around application configuration for the termination timeout, which I'll note in the PR.
Hi, this PR is relevant to my interests, and would be interested in seeing this merged. Is there any sort of timeline for this, or anything outstanding? Thanks
Hi @nstott; it should be merged in a few days, unless @jbbarth has objections
No objection, go ahead @ybastide, and thanks @nstott ! :-)
Hello!
I've run into this issue with using simpleflow activities to supervise an external subprocess (i.e., a long-running bash or Python executable). I see that you have some of this behavior captured in
simpleflow.execute
: do you use these in actual workflows? Separately, would make sense to modify the theActivityPoller
'sspawn
code to reap the entire process tree at the worker PID.Here's a modified version of
examples/basic.py
that reproduces the issue.And some commands to start:
You'll see that that on retry, the forked subprocess from first activity execution is still printing its PID once the retried execution starts:
I have a patch available that fixes this, which I'll put up and link to this issue. If there's a more idiomatic way to use the simpleflow API to achieve this, I'm happy to use that too, although it seems like this is probably undesirable behavior.
Edit: patch linked. If it seems ok, let me know and I can add test coverage.