Github user tillrohrmann commented on the issue:
https://github.com/apache/flink/pull/4697
I guess some of the problems like taking a savepoint could still be served
under `/jobs/:jobid/savepoints/:savepointid` because this resource will be
removed upon first request anyway.
I'm not sure how well providing better error messages will work in the end,
because right now, we don't distinguish between completed and running jobs on
the handler side. What it currently does is trigger `takeSavepoint` on the
`Dispatcher`. The `Dispatcher` will fail with a `FlinkJobNotFoundException` if
the job is no longer executed/non-existent. In order to see whether there was a
job being executed, the handler then had to access some other information
Monitoring a job is indeed a bit tricky with splitting the URLs.
Theoretically, this is already the case, since we only store a limited number
of `ArchivedExecutionGraphs` on the `MemoryArchivist`. Thus, if you have many
completing jobs, it might be the case that you don't see it when querying
`/jobs/:jobid`.
Nevertheless, the advantages of keeping it like it is are evident. I'll
change the method to `PATCH`. Let's see how far we get with that.
---
