https://pantsbuild.org/ logo
s

sparse-lifeguard-95737

07/19/2022, 7:07 PM
if you set a timeout for a
python_test
target, does that timeout start ticking only when the
pytest
process starts running? or does the timeout start when the process is queued for execution? we sometimes see tests time out in CI that run in well under their configured timeout individually
h

high-yak-85899

07/19/2022, 7:07 PM
I've experienced the latter
I've also seen the behavior you're describing and I attribute it to the machine executing a bunch of tests being under much higher load than running one test by itself
👍 3
s

sparse-lifeguard-95737

07/19/2022, 7:08 PM
I’m wondering if the structural problem from https://github.com/pantsbuild/pants/issues/16188 applies to more operations than the remote-cache reads
yeah until now I had the same thought, but I just saw a test take 3x as long running in CI as it does locally so I thought I should double-check 😬
w

witty-crayon-22786

07/19/2022, 7:40 PM
only once it starts executing
that location puts the timeout very under scheduling, but also after the sandbox for the process has been created. basically, as late as possible: right when the subprocess has been spawned.
I’m wondering if the structural problem from https://github.com/pantsbuild/pants/issues/16188 applies to more operations than the remote-cache reads
almost certainly! not here though i don’t think. when you see differences in CI vs locally, it might be due to resource contention, such as when running lots of tests concurrently (as @high-yak-85899said)
👍 1
4 Views