[core] Deflakey gcs fault tolerance test in mac os #36471
Conversation
@can-anyscale @aslonnie Do we have plan to run mac tests inside container for better isolation? |
|
@jjyao I think mac os doesn't have container support. |
mac does not really have containers. container is mostly a linux thing. when you run containers on macs, it is mostly running on a linux virtual machine on that mac. there are experiments and research on porting container / linux namespace tech to osx or windows, but even that is stable, most players in the ecosystem do not use/build containers for them right now. |
…ault-tolerance-test
…ault-tolerance-test Signed-off-by: Yi Cheng <74173148+iycheng@users.noreply.github.com>
| @contextmanager | ||
| def _setup_redis(request): | ||
| with tempfile.TemporaryDirectory() as tmpdirname: | ||
| kill_all_redis_server() |
There was a problem hiding this comment.
minor, but may be it could be after yield? (so we kill it as a part of current conftest, not killing redis from the previous conftest)
There was a problem hiding this comment.
I think maybe we should do both? The reason I added it here is because maybe the program crashed and the kill has never been reached
## Why are these changes needed? The root cause is because the port already in use error. This PR fixed it by doing two things: - Cleanup and delete the old redis-server. Mac doesn't run code in container, so it might be the leak from last run. - Make sure ping success before progress. ## Related issue number Closes ray-project#31871 Signed-off-by: e428265 <arvind.chandramouli@lmco.com>
…ct#36471)" (ray-project#36835) This reverts commit e0655ff. Signed-off-by: e428265 <arvind.chandramouli@lmco.com>
Why are these changes needed?
The root cause is because the port already in use error. This PR fixed it by doing two things:
Related issue number
Closes #31871
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/under thecorresponding
.rstfile.