runc icon indicating copy to clipboard operation
runc copied to clipboard
verified publisher icon opencontainers

flaky test: TestProcessEmptyCaps

Open cyphar opened this issue 2 years ago • 2 comments

I've started seeing this error in CI more often recently (in GHA CI runs). I suspect there's a race with the kernel creating the cgroup dentries. Maybe we should have some retry code? Idk.

=== RUN   TestProcessEmptyCaps
time="2023-12-07T04:19:25Z" level=warning msg="unable to get oom kill count" error="openat2 /sys/fs/cgroup/memory/test/integration/memory.oom_control: no such file or directory"
    exec_test.go:299: unexpected error: unable to start container process: unable to apply cgroup configuration: openat2 /sys/fs/cgroup/cpuset/test/integration/cpuset.mems: no such file or directory
--- FAIL: TestProcessEmptyCaps (0.13s)

cyphar avatar Dec 07 '23 06:12 cyphar

Yes, I am seeing something like this, suspected a kernel race, and suggested a retry (#3351). Maybe we should do it :)

See also #3540.

kolyshkin avatar Oct 12 '24 01:10 kolyshkin

This is mostly about cpuset.cpus and cpuset.mems for cgroup v1. We have an ugly mechanism of copying values from the parent cgroup during Apply, which no one has courage to remove. At least we don't have it for cgroup v2.

Anyway, here's my try: https://github.com/opencontainers/runc/pull/4442

kolyshkin avatar Oct 12 '24 19:10 kolyshkin