kubectl delete: Command Reference for CI/CD

Remove resources by name, label, or the manifest you applied.

kubectl delete powers teardown jobs and preview-environment cleanup. This reference lists the flags that control scope and termination timing, plus a CI pattern that scopes deletes to a namespace so a stray --all cannot wipe the wrong place.

Common flags and usage

• -f <manifest>: delete exactly what a manifest set created
• -l, --selector: delete every object matching a label
• --all: delete all objects of a type in the namespace (scope with -n!)
• --wait=false: return without blocking on graceful termination
• --grace-period=0 --force: skip graceful termination for a stuck pod
• --timeout: cap how long delete blocks before failing

Example

# Tear down a per-build preview namespace
kubectl delete namespace ci-${BUILD_ID} --timeout=120s

# Targeted teardown of just this build's objects
kubectl delete -f k8s/ -l build=${BUILD_ID} --wait=true

In CI

Always scope teardown to a dedicated namespace and delete that namespace rather than using --all without -n, which deletes everything in the default namespace. Pass --timeout so a finalizer that hangs a delete fails the step instead of stalling until the job-level timeout.

Key takeaways

• Scope teardown to a namespace; --all without -n hits the default namespace.
• A delete that hangs is usually a finalizer; cap it with --timeout.
• --grace-period=0 --force is a last resort for known-stuck fixtures only.

Related guides