Change the RestartPolicy processing logic in ReplicationController and NodeController #6273
Conversation
ReplicationController allows pod RestartPolicy to be Always, OnFailure and Never. NodeController should consider the RestartPolicy when deleting pods. add some tests for replication controller Testing restart policy
|
Thanks for your pull request. It looks like this may be your first contribution to a Google open source project, in which case you'll need to sign a Contributor License Agreement (CLA) at https://cla.developers.google.com/. If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Check the information on your CLA or see this help article on setting the email on your git commits. Once you've done that, please reply here to let us know. If you signed the CLA as a corporation, please let us know the company's name. |
| if api.RestartPolicyNever == pod.Spec.RestartPolicy { | ||
| continue | ||
| } else if api.RestartPolicyOnFailure == pod.Spec.RestartPolicy { | ||
| if api.PodSucceeded == pod.Status.Phase { |
There was a problem hiding this comment.
What was the rationale behind this change?
There was a problem hiding this comment.
Thanks for your reply.
Please see this use case:
We create a pod called myPod which is used to process a text file and its RestartPolicy is OnFailure. If the pod's status is PodSucceeded, we think the processing is complete.
Now we assume that a node has been crashed, the NodeController may delete all the pods (include myPod) on this node. When the node reboot, myPod will be run twice.
There was a problem hiding this comment.
The right way to do this would be to add a job controller API to the system. #1624
There is a GSoC proposal for this. In the meantime, we're trying to make it simpler and more efficient to add a new controller to the system.
|
Thank you for the PR. Please follow up on #1624. |
We think ReplicationController should allow pod's RestartPolicy to be OnFailure and Never(it is just allowed to be Always now). We also think NodeController should consider the RestartPolicy when deleting pods.