Expected Behavior

After a TaskRun is timed out or cancelled, I should still be able to view the logs of the failed or cancelled TaskRun.

Actual Behavior

It appears that the pods for TaskRuns are being deleted due to a change implemented in #2365. In the func, failTaskRun in taskrun.go, the pod associated with the TaskRun is being deleted as noted here.

Assuming this is not an expected feature, I would suggest checking for the failure reason before deleting the TaskRun pod:

if reason != v1beta1.TaskRunReasonTimedOut && reason != v1beta1.TaskRunReasonCancelled {
		err := c.KubeClientSet.CoreV1().Pods(tr.Namespace).Delete(tr.Status.PodName, &metav1.DeleteOptions{})

		if err != nil && !k8serrors.IsNotFound(err) {
			logger.Infof("Failed to terminate pod: %v", err)
			return err
		}
}

If this is expected, it would help to document this behavior for TaskRuns as well as recommended best practices for preserving logs.

Steps to Reproduce the Problem

1. Create a TaskRun with a timeout such that the timeout will be violated or cancel a running TaskRun
2. Run kubectl get pods to find if the pod of the TaskRun has been deleted

Additional Info

• Kubernetes version:

N/A

• Tekton Pipeline version:

Using what is in latest as of v0.15.0.