Workaround the race condition in runc.#643
Conversation
|
Thank you for your PR @jankaluza Has this been handled on runc side yet @kolyshkin ? |
|
Hm, it seems this workaround did not work here in the CI tests. The good thing is this is not just an issue with my system - it can be reproduced also here. I'll add more debug output and also report that to runc. |
|
Actually, the update I've pushed was wrong. Let's see what happens now with the latest one... |
The tests started failing randomly for me after upgrade to Fedora 43. I've found out that sometimes `runc start` fails with "cannot start a container that has stopped" despite the fact that `runc state` says it is in `created` state. It is fixed by running `runc state`. This commit workarounds this issue until it is properly investigated and fixes so the tests pass. Signed-off-by: Jan Kaluza <jkaluza@redhat.com>
This comment was marked as outdated.
This comment was marked as outdated.
|
Looking at runc git log, one possible change that can theoretically result in a bug like this is opencontainers/runc@d9f2a24a but it's there since 1.2.0.
Looking into what |
|
My bad, somehow I missed this. Will try to repro |
|
I'm closing this PR. I think there is no easy workaround for this probably. It just happens randomly. I think we can just live with that and wait for runc fix Conmon is not that heavily developed, so we can survive some time with broken tests on F-43. |
3 participants
The tests started failing randomly for me after upgrade to Fedora 43. I've found out that sometimes
runc startfails with "cannot start a container that has stopped" despite the fact thatrunc statesays it is increatedstate. It is fixed by runningrunc state.This commit workarounds this issue until it is properly investigated and fixes so the tests pass.