On the storm UI are you seeing your topology getting the required number of
workers and is the topology emitting any tuples?

Also Storm UI should tell you if any of the supervisors were started and if
they were on which machine and you can ssh to that machine to look at
supervisor logs.

But the general behavior is that once workers start up they will report
their heartbeats into ZK, and Nimbus will pull this data from ZK to confirm
which worker is up and which is down.

Hope this helps!

Ankit


On Fri, Aug 15, 2014 at 10:15 PM, Yaron Rosenbaum <[email protected]
> wrote:

> That's the thing - I don't know.
> Mesos is supposed to start the workers, so either they don't start, or
> they start - but they can't communicate with Nimbus.
> Or do they need to communicate with Zookeeper directly?
>
> This is a slightly more complicated setup than usual.
>
> (Y)
>
> On Aug 15, 2014, at 8:07 PM, Ankit Toshniwal <[email protected]>
> wrote:
>
> Are the workers starting fine? I have seen this in the past when the
> workers were not reporting their heartbeats leading to this message in
> nimbus.log
>
> Ankit
>
>
> On Fri, Aug 15, 2014 at 9:18 AM, Yaron Rosenbaum <[email protected]
> > wrote:
>
>> Hi
>>
>> I've set up a Docker based Mesos setup, including a mesos-storm framework
>> installed (plus chronos, marathon, ..).
>> You can find it here: https://github.com/yaronr/docker-mesos
>>
>> It works fine, storm UI is accessible.
>> I am able to deploy topologies, and storm-ui reflects these deployments
>> correctly.
>> However, from the storm logs, it appears that the executors are not able
>> to communicate with nimbus. Or they don't exist (mesos is responsible for
>> spawning 'supervisors' or executors).
>> 2014-08-15 16:03:11 b.s.d.nimbus [INFO] Executor
>> wordcount-1-1408118461:[3 3] not alive
>>
>> I'm suspecting this has something to do with some ports / links that I
>> have to define, but I have nothing to go on but this wild guess.
>>
>> Any help would be appreciated!
>>
>> (Y)
>>
>>
>
>

Reply via email to