some questions about job sync

[ultraji]

some questions:

q1:

https://github.com/kubeflow/tf-operator/blob/7e6953133c3d95c6f3103348351b999e4e84f955/pkg/controller.v1/tensorflow/controller.go#L321

why don't need to delete expectations here?

q2:

https://github.com/kubeflow/tf-operator/blob/7e6953133c3d95c6f3103348351b999e4e84f955/pkg/controller.v1/tensorflow/controller.go#L251

https://github.com/kubeflow/tf-operator/blob/7e6953133c3d95c6f3103348351b999e4e84f955/pkg/controller.v1/tensorflow/controller.go#L317

same code block in a sync loop?

[Jeffwan]

@ultraji

Q1: This is at different level. when pod/services are removed, expectations are deleted but not here at the job level.
example: https://github.com/kubeflow/common/blob/2f3f636f16ef4cedb12543a96ac1412da98bbca5/pkg/controller.v1/common/pod.go#L211

The logic you attached indicates the job is deleted. That means normal path has been went through. Does it make sense?

Q2:

processNextWorkItem normally just validates the key. It passes the key to syncTFJob. Since syncTFJob needs to retrieve the job object for a deep copy, it has to handle the error one more time. However, it may not be reached because it has been handled earlier.

[Jeffwan]

/kind question

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.