Assuming it as a helm-chart deployment*
On a cgroup2 enabled k8s cluster, in ingress controller pod the nginx process will use all the cores of the node and will not be restricted by container resource limits. This can cause OOM and ingress-controller pod will go to a CrashLoopBackOff.
As of now, workaround is to set the worker_processes in a configmap (nginx.conf).
Ref: https://github.com/kubernetes/ingress-nginx/issues/9665
Top comments (0)