Aller au contenu

KubePodCrashLooping#

Fires when a pod fails to start.

Troubleshooting hints#

  • Get on the cluster and run kubectl -n <namespace> describe pod <pod>
    • the name of the failed pod is in the description
    • the namespace of the failed pod is in the alert tags
  • The reason why it failed should appear in the Events: section at the end.

The reason for an OOM, see resource constraints such as too little memory, insufficient CPU required at start-up, etc.

Known cases#

Quay#

Quay may require container images not available on the node it is scheduled on (and if it is down, it cannot retrieve them from itself). All its images where saved on the bastion:

ssh bst.ocb-corp.caascad.com

[cloud@bst ~]$ ls
init-config-quay_v1.3.0.tar  quay-builder_qui-gon.tar  quay.tar  redis_6.0.9.tar  spilo-14_2.1-p7.tar

Copy them with scp and load and tag them with docker on the relevant node ip.

scp quay-builder_qui-gon.tar 10.0.45.173:/home/cloud/quay-builder_qui-gon.tar
ssh 10.0.45.173 "sudo docker load --input quay-builder_qui-gon.tar"
docker tag ...