Fix the cgroup 2 process attaching problem #677
Conversation
kbfu
requested review from
ispeakc0de,
ksatchit and
uditgaurav
as code owners
kbfu
changed the title
|
Hello @kbfu, thank you for your contribution through the pull request. I would like to inquire about the specific cluster environment and container runtime where you have conducted your tests. We encountered an issue when running it on a GKE cluster with containerd and cgroupv2. Here's the error we observed: could not get cgroup manager --- at /litmus-go/chaoslib/litmus/stress-chaos/helper/stress-helper.go:134 (prepareStressChaos) --- Caused by: Error in getting groupPath,nsenter: unrecognized option: C BusyBox v1.35.0 (2022-08-01 15:14:44 UTC) multi-call binary. Usage: nsenter [OPTIONS] [PROG ARGS] -t PID Target process to get namespaces from -m[FILE] Enter mount namespace -u[FILE] Enter UTS namespace (hostname etc) -i[FILE] Enter System V IPC namespace -n[FILE] Enter network namespace -p[FILE] Enter pid namespace -U[FILE] Enter user namespace -S UID Set uid in entered namespace -G GID Set gid in entered namespace --preserve-credentials Don't touch uids or gids -r[DIR] Set root directory -w[DIR] Set working directory -F Don't fork before exec'ing PROG |
|
Hi @uditgaurav , I rebuilt the image and replaced the base image from alpine to debian. I believe nsenter command from busybox was outdated. |
|
@kbfu, Thanks for your response. I'm wondering if we can integrate this capability within the Alpine-based image itself, as this would help in maintaining a smaller image size. The corresponding version of util-linux package in Alpine is For your reference, the experimental Dockerfile is located here - litmus-go Dockerfile. It uses the base image |
|
Hi @kbfu, I've created a test experiment image using the same Alpine-based image, which includes package Current Output: Previous Output: It works well with
Tagging @ispeakc0de, for any suggestions for additional tests or use cases for |
There was a problem hiding this comment.
LGTM 🙌
Note: We need to update the base image here with updated util-linux (using apk add --no-cache util-linux=2.38-r1).
| //groupPath, err := cgroupsv2.PidGroupPath(t.Pid) | ||
| //if err != nil { | ||
| // return nil, cerrors.Error{ErrorCode: cerrors.ErrorTypeHelper, Source: t.Source, Target: fmt.Sprintf("{podName: %s, namespace: %s, container: %s}", t.Name, t.Namespace, t.TargetContainer), Reason: fmt.Sprintf("fail to get pid group path: %s", err.Error())} | ||
| //} | ||
| // | ||
| //cgroup2, err := cgroupsv2.LoadManager("/sys/fs/cgroup", groupPath) | ||
| //if err != nil { | ||
| // return nil, cerrors.Error{ErrorCode: cerrors.ErrorTypeHelper, Source: t.Source, Target: fmt.Sprintf("{podName: %s, namespace: %s, container: %s}", t.Name, t.Namespace, t.TargetContainer), Reason: fmt.Sprintf("fail to load the cgroup: %s", err.Error())} | ||
| //} | ||
| //return cgroup2, nil |
There was a problem hiding this comment.
Can we remove the commented code?
What this PR does / why we need it:
Fix the problem when attaching the process to another cgroup when using cgroup 2.
Which issue this PR fixes (optional, in
fixes #<issue number>(, fixes #<issue_number>, ...)format, will close that issue when PR gets merged): fixes #Fixed this issue.
litmuschaos/litmus#3902
Special notes for your reviewer:
Checklist:
breaking-changestagrequires-upgradetag