OOM-killed core-dump-handler pods

[yichang-lin]

We're using core-dump-handler chart version v9.0.0 and running it on AWS EKS v1.31. Although we have increased request_mem and limit_mem to 256Mi, we are still having core-dump-handler pods get OOM-killed. Is this related to the new EKS version or the new core-dump-handler version? Thanks!

[No9]

Hi @yichang-lin
Have you looked at the kubelet logs to understand the eviction events in more detail?

[yichang-lin]

@No9 We haven't checked the logs but the issue was resolved by rolling back the version to v8.10.0.
Here's the detailed log from the OOM-killed pod:

Process core-dump-agent (pid: 954575, oom_score: 132118, oom_score_adj: 998) triggered an OOM kill on itself. The process had reached 65536 pages in size.

[No9]

Thanks very much - looking at the diff between 9.0.0 and 8.10.0
v8.10.0...v9.0.0
The main changes in the agent was for deployment parameters, a change in the inotify api and a bump in dependencies.
A candidate here is the tokio-scheduler dependency bump.
I'll downgrade the changes and publish 9.1