Github user mxm commented on the pull request:
https://github.com/apache/flink/pull/681#issuecomment-104707707
I've updated the pull request.
- The `JobID` is now propagated from the `ExecutionEnvironment` to the
`JobGraph`. That's why a setter for setting the `JobID` at the `JobGraph` was
necessary. It is ensured that the `JobID` can never be set to null. Job graphs
which are instantiated manually, still generate a random `JobID` which does not
depend on the session.
- A session timeout has been introduced that specifies the life span of a
session. A timeout of zero implies immediate termination of the session after
the job has been completed. Otherwise the `ExecutionGraph` is kept for the
specified number of seconds.
- The methods `startNewSession()` and `setSessionTimeout(timeout)` have
been added to the `ExecutionEnvironment`.
- The method `endSession()` has been added to the `Client` which is invoked
through the `RemoteExecutor` or the `ContextEnvironment`.
- The `LocalExecutor` has been modified to respect the session, i.e. it
does not shut down the Flink cluster it creates after execution when a session
timeout >0 has been specified. In case a new session is invoked through the
`LocalEnvironment`, a new LocalExecutor is created.
- The `finalize()` and a shutdown hook try to end the session when the
`ExecutionEnvironment` gets garbage collected or the JVM is shut down.
I'm missing integration tests for the `RemoteEnvironment` and
`LocalEnvironment` but there is not much to test at the moment. The core
session functionality is tested in the `JobManagerITCase`.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
