Cluster Deletion Hanging Issues and Resolution

Products

• Tanzu Kubernetes Runtime

Issue/Introduction

Clusters can experience hanging issues during deletion from a CAPI (Cluster API) perspective. This can manifest as delays or failures in the complete removal of cluster resources, including the cluster itself and associated namespaces. A key aspect to understand is that the cluster's operational state, such as etcd health, is ignored during the deletion process.

Cause

Primary reasons for clusters hanging during deletion are:

• Paused Cluster Objects: Cluster objects being in a paused state will prevent CAPI from reconciling and processing deletion requests. This pause can halt the progression of the deletion workflow.
• Unhealthy Supervisor: An unhealthy supervisor, often due to an expired CAPI certificate, can disrupt the deletion process. When the certificate expires, the cert-manager component, crucial for certificate management, may need a restart to allow the deletion to proceed.

Resolution

Identify the specific cause (paused objects or unhealthy supervisor) and take appropriate action, such as unpausing objects or restarting cert-manager.

• Workload Cluster stuck in Deleting state due to "paused: true" variable in Cluster Manifest
• vSphere Supervisor System Pod Certificate Expiry due to Cert-Manager Issues