Question

You maintain a Deployment that should run on nodes labeled disktype=ssd whenever possible, but the Pods must still be scheduled even if no SSD nodes are available. Your current spec uses requiredDuringSchedulingIgnoredDuringExecution with a nodeSelectorTerm matching disktype=ssd, and Pods now stay Pending when all SSD nodes are full. Which change to the Deployment's Pod spec achieves the desired behavior?

Accepted Answer

Replace the requiredDuringSchedulingIgnoredDuringExecution rule with preferredDuringSchedulingIgnoredDuringExecution, giving the disktype=ssd match a weight.

Answer

Add a nodeSelector of disktype=ssd alongside the existing required affinity rule.

Answer

Change the operator in the required rule from In to Exists so any node matches.

Answer

Add a toleration for a disktype=ssd taint so the Pods can schedule anywhere.