Questions to Determine¶
- What PodDisruptionBudgets exist in the namespace and what are their configurations?
- How many replicas does the payment-service deployment currently have running?
- Is the PDB's
minAvailableormaxUnavailablesetting compatible with the current replica count? - Are there other nodes in the cluster with sufficient resources to schedule a new replica?
- Can the deployment be temporarily scaled up before draining?
- Is there a way to proceed with the drain without violating the PDB?
- What is the business impact of forcing the eviction with
--disable-eviction? - Who owns the PDB and was it intentionally set to block single-replica evictions?
- Are there pending cluster autoscaler actions that might add capacity?