Closed sabidib closed 7 years ago
Curious if updating a coordinator in a suspended state makes sense.
Yes! This is actually a really important usecase. A deploy shouldn't skip a suspended coord, nor should it resume it.
I gave this a test spin with a local instance of oozie and switched a coordinator between the minimal coordinator and the full coordinator successfully. The diffs made sense as well. It also handled "updating" to the same coordinator xml showing an empty diff.
It also handled "updating" to the same coordinator xml showing an empty diff.
Presumably this was with identical config data?
Presumably this was with identical config data?
Yes, the configuration was the same.
Any outstanding comments left?
This adds the ability to update a coordinator's definition without killing and recreating the coordinator.
According to the docs a coordinator's definition can only be updated if the following are not changed:
If any of these are changed, then there should be a hard failure on the request.
I couldn't find anything specific on what coordinator states allow for an update so I assumed we would only want to update coordinators that are
active
orrunning
orsuspended
.Curious if updating a coordinator in a
suspended
state makes sense. @kmtaylor-github @cfournie Thoughts?