Cluster pods remain PENDING for too long

Issue descriptionIssue description

• Attempting to start new pods in a Managed Service for Kubernetes cluster results in these pods remaining PENDING for a long period without switching to RUNNING.
• New pods get frozen as PENDING.
• One or more pods in a Managed Service for Kubernetes cluster fail to switch to RUNNING.

SolutionSolution

Pods can get stuck in the PENDING status if the processes within the init container of one or more pods fail to complete successfully. For example, the kubelet daemon in such pods may continuously crash. In this case, the kubelet immediately restarts due to the restartPolicy: Always policy.

For more details about init containers, see this Kubernetes overview article.

If the issue persistsIf the issue persists

For more detailed diagnostics of kubelet issues in newly created pods, create a support ticket. Provide the following information in your ticket:

1. ID of the Managed Service for Kubernetes cluster in question.
2. Logs with kubelet/kernel/system syslog tags from one or more Managed Service for Kubernetes cluster nodes experiencing issues when attempting to start new pods.

You can also use our script to collect Managed Service for Kubernetes diagnostic logs automatically.