| commit | 837cb8e459b9eefabe89ab17df0b7dafb5e3d631 | [log] [tgz] |
|---|---|---|
| author | Lorenz Brun <lorenz@monogon.tech> | Mon Dec 23 13:52:56 2024 +0100 |
| committer | Lorenz Brun <lorenz@monogon.tech> | Mon Dec 23 21:59:59 2024 +0000 |
| tree | 32337d84d4f32b0c2c523e2c5bd177f4acfe4808 | |
| parent | b6afed68fd1d2ee9b32d395b388d2db1338d0fa0 [diff] |
treewide: update Kubernetes to 1.32 Relatively easy change, one cadvisor fix is temporarily needed. The legacy log dir patch needed to be rebased, that's about it. I enabled single-process OOM killing again as that was the default for cgroupv1 and IMO the more sane behavior. Upstrem changelog at: https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG/CHANGELOG-1.32.md Change-Id: I537a6e37137d05efb6eec8635915e36fd8b37cbc Reviewed-on: https://review.monogon.dev/c/monogon/+/3721 Reviewed-by: Jan Schär <jan@monogon.tech> Tested-by: Jenkins CI
diff --git a/metropolis/node/kubernetes/kubelet.go b/metropolis/node/kubernetes/kubelet.go index 16fb520..7a75ecf 100644 --- a/metropolis/node/kubernetes/kubelet.go +++ b/metropolis/node/kubernetes/kubelet.go
@@ -121,6 +121,10 @@ MaxPods: 253, PodLogsDir: "/data/kubelet/logs", FeatureGates: extraFeatureGates.AsMap(), + // Only kill a single process on OOM instead of the whole container. + // This is generally the more sane behavior and was default under + // cgroups v1. + SingleProcessOOMKill: ptr.To(true), } }