Github user tillrohrmann commented on the pull request:
https://github.com/apache/flink/pull/1202#issuecomment-145598003
The PR looks good. I think we cannot do much about it being a temporary
workaround for the limitation that our `ExecutionGraph` is not serializable.
Once we have the resources this should be addressed, though.
I've got some minor comments. The main point is the blocking call in the
`notifyLeaderAddress` of the `JobManagerArchiverRetriever`. This should be
fixed imo.
Moreover, I think I've found a race condition. The `RuntimeMonitorHandler`
checks whether the current leader is its own or a remote one. If it's the local
JM, then the respective RequestHandler is called. Within the `RequestHandler`,
such as `CurrentJobsOverviewHandler` the `JobManagerArchiverRetriever` is
called again. If now the leader has changed to be a remote JM, then we'll have
a problem. This could be fixed as well.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
