Github user ankurcha commented on the pull request:
https://github.com/apache/flink/pull/948#issuecomment-136236719
@rmetzger I have finally got some time to work on this again. Let me
address your question one by one:
> Why did you decide to start the JobManager alongside the Scheduler?
This is basically a "easy" first step way of getting things running the way
it was done in a whole bunch of projects. The easiest way to run a "single"
master + multiple worker application is to make the scheduler run the master
process and have another meta-framework such as marathon submit the whole
framework as a task to the mesos server. In the lack of marathon or aurora etc,
mesos-submit ( an app that ships with mesos) can be used to submit the
scheduler as a task. This means the job manager + scheduler would be running in
the mesos cluster submitted as an app (just like in YARN).
My eventual goal is to make the scheduler support a completely standalone
mode of operation but that requires coordination in order to assure that only
one scheduler instance exists at a time - this may have some hooks that can be
a part of the HA job manager initiative.
> Tests
I am working on some docker and vagrant based scripts that can make the
"setup" part of the tests more palatable.
---
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.
---
