Troubleshooting Pod ContainerCreating

Troubleshooting / Guide 81 of 103 / Beginner

• Kubernetes1.36.4
• Cluster4 nodes
• Runtimecontainerd 2.2.6
• CNICalico v3.32.1
• TimeAbout 16 min

Event text is the kubelet's own. The retry interval and the (x6 over 20s) counter will differ on your cluster; the message format is what to learn.

This guide includes

Use this when a Pod is scheduled and still will not start. This matters because ContainerCreating means the scheduler has finished and the kubelet has not, so the answer sits in the kubelet's events rather than anywhere the Pod's own phase will tell you.

• creating a Pod that mounts a Secret which does not exist, and watching it never leave ContainerCreating
• reading the kubelet's event, which names both the reason and the missing object
• seeing why .status.phase disagrees with the STATUS column, and which of the two to trust
• comparing three Pods - a missing Secret, a missing ConfigMap, and one that is not the same reason at all
• creating the Secret and watching the retry succeed with no further intervention

Before you start

• guide 80 - the other half of this split, where the Pod never reaches a node at all.

1. 1

   A Pod that never leaves ContainerCreating

2. 2

   What the kubelet is complaining about

3. 3

   Why the phase field will not tell you

4. 4

   The same reason, and the one that is not

5. 5

   Fix it, and watch the retry succeed

Official sources

• Debug Pods
• Pod lifecycle