Fu Wei

I think the #6150 feature can report the detail if you enable the trace level log. Anyway, it is better to handle the case in CRI protocol, like https://github.com/kubernetes/enhancements/pull/1338. Maybe...

please use your real name to sign off. Thanks

@Burning1020 could you mind to add the integration test case with failpoint? Thanks

No sure I understand it correctly. As far as I know, each group has those own tag naming format. IMO, it's hard to force use suffix to mark it immutable,...

The issue is caused by RunPodSandbox rollback. I am trying to fix it in https://github.com/containerd/containerd/pull/5885/files.

> **containerd 1.5.13** + debian 11 + k8s 1.24.2 - same issue Hi, we are working on this and will update the status after merge. Thanks

hi, @mimmus it depends on the kernel behavior. In v3.10, kernel will notify subscriber (containerd) at first before kill [1](https://github.com/torvalds/linux/blob/8bb495e3f02401ee6f76d1b1d77f3ac9f079e376/mm/memcontrol.c#L2218) And if the oom_kill has been disabled or the kernel...

@mimmus Basically, the containerd will report the OOM_EVENT if there is any out of memory. But it is related to kernel. I think you can check the memory cgroup setting,...

> Perhaps these info aren't really applicable to my use case. > > I think that a consolidated OS like CentOS already has OOM killer enabled. > > I'm trying...

Thanks for the information. Will take a look and reproduce it.