Github user rmetzger commented on the pull request:
https://github.com/apache/flink/pull/948#issuecomment-131796852
Hi @ankurcha,
I've started a Mesos cluster on Google Compute Engine to try out your pull
request.
I've used this configuration:
```
flink.mesos.master: zk://127.0.0.1:2181/mesos
flink.uberjar.location: hdfs:///user/jclouds/flink-dist-0.10-SNAPSHOT.jar
flink.mesos.taskmanagers.mem: 512
flink.mesos.taskmanagers.cpu: 0.5
taskmanager.logging.level: INFO
streamingMode: streaming
jobmanager.web.port: 8081
webclient.port: 8080
```
But I'm getting this error
```
Exception in thread "main" java.lang.NullPointerException
at
org.apache.flink.mesos.scheduler.SchedulerUtils$class.createFrameworkInfoAndCredentials(SchedulerUtils.scala:255)
at
org.apache.flink.mesos.scheduler.FlinkScheduler$.createFrameworkInfoAndCredentials(FlinkScheduler.scala:31)
at
org.apache.flink.mesos.scheduler.FlinkScheduler$.main(FlinkScheduler.scala:183)
at
org.apache.flink.mesos.scheduler.FlinkScheduler.main(FlinkScheduler.scala)
```
I'll further investigate the issue.
*Why did you decide to start the JobManager alongside the Scheduler?*
For Flink on YARN, we are starting the JobManager in a separate container.
There is a lot of communication going on between the JobManager and
TaskManagers, also, we need to ensure that the TaskManagers are able to reach
the JM.
I think we can safely assume that containers can always communicate among
each other ... I'm not so sure about Mesos clients and cluster containers.
> The mesos scheduler is not HA and should be used with marathon or similar
service to ensure that there is always one instance running. This may be
addressed in future patches.
Would you start the mesos scheduler on the client machine or inside the
cluster, using a container?
Whats the typical deployment model for Mesos?
---
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.
---
