Replace Launcher from v1.Job to v1.Pod

[carmark]

We need to find the real job's status, not active number, since pending pod is a active job.

[gaocegege]

SGTM

[terrytangyuan]

Seems fine. cc @rongou to double check as well

[rongou]

I think the restart policy is useful for recovering from transient errors. Are you planning to implement similar logic in the controller?

[terrytangyuan]

Also, downstream apps may have been consuming from launcher job so we may need to be careful regarding compatibility. We probably want to make a release before merging the changes.

[carmark]

I think the restart policy is useful for recovering from transient errors. Are you planning to implement similar logic in the controller?

@rongou Yes. We can implement it in controller if user want it. However, we saw many users would check the error and relaunch it.

Also, downstream apps may have been consuming from launcher job so we may need to be careful regarding compatibility. We probably want to make a release before merging the changes.

@terrytangyuan This chance should not impact the API, we may do the same things in Controller or consider whether it is necessary. What do u think?

[terrytangyuan]

@carmark I am fine with this change as long as the API keeps the same and restart policy is implemented (there are users who are already relying on this I think).

[carmark]

@terrytangyuan shall we add this issue for v1 graduation?

[terrytangyuan]

Sounds good to me. Link #217