Hello! My cluster was upgraded to v1.33.1-gke.1386000 from 1.32.4-gke.1353001 in the rapid channel this weekend. I know rapid channel is not part of the GKE SLA, but I wanted to report the issue and see if anyone else was having similar issues.
Several of the more intense applications in my cluster are now crashing, reporting “Too many open files”, although the applications have not changed in some time.
I also have the same problem, in ingress. We need help to solve this case as soon as possible.
I’ve moved my cluster off of a release channel, disabled auto-upgrades on the node pool, and then set the version for the node pool to 1.32.4-gke.1353001
There was an issue with an underlying changed made in the defaults for containerd 2.0 which is used in GKE 1.33 and later.
Check out https://github.com/GoogleCloudPlatform/k8s-node-tools/tree/master/containerd-nofile-infinity for a fix.