Hi Bobby,

Thank you very much for all the trouble taken. I was killing all the worker
processes using Ctrl-C.

I wasn't running any of the processes (workers, nimbus, etc.) under
supervision.
That was the reason causing above mentioned behavior.

Now, I am running all of these processes using `supervisord` and kill them
using
`kill` command. Now the replay happens correctly. I read about running the
processes
under supervision in [1].

Thanks again for your kind help.

[1] http://www.michael-noll.com/tutorials/running-multi-node-storm-cluster/


Regards,
Thilina

On Sat, Aug 8, 2015 at 12:17 AM, Bobby Evans <[email protected]>
wrote:

> I have tried several times to reproduce this and I am not able to.  Do
> which server are you shooting and restarting?  Are you killing the
> supervisor and all of the worker processes at the same time?  Are you
> shooting one worker process?  What exactly are you doing?  I have tried
> both of these several times using the latest from branch-0.9.x and I have
> not been able to see it happen.
> - Bobby
>
>
>      On Thursday, August 6, 2015 11:33 PM, Thilina Rathnayake <
> [email protected]> wrote:
>
>
>  Hi Bobby,
>
> Thank you very much for the reply.
>
> I have created a git repository with my code in [1].
>
> `RepeatWordsSpout` in [2] is just keeps repeating the same set of words in
> different batches.
>
> In `RWEmitter`[3],  I use a hash map to keep track of the batches emitted
> and remove a batch
> from the map after the batch successfully gets processed.
>
> Other code is just boiler plate.
>
>
> [1] https://github.com/thilinarmtb/trident-demo
> [2]
>
> https://github.com/thilinarmtb/trident-demo/blob/master/src/main/java/trident/demo/spouts/RepeatWordsSpout.java
> [3]
>
> https://github.com/thilinarmtb/trident-demo/blob/master/src/main/java/trident/demo/spouts/RWEmitter.java
>
>
>
> Regards,
> Thilina
>
> On Thu, Aug 6, 2015 at 4:48 PM, Thilina Rathnayake <[email protected]
> >
> wrote:
>
> > Hi All,
> >
> > I posted this question in user's mailing list as well. I am posting here
> > because I
> > am not sure whether this is a bug in trident or not.
> >
> > I am playing with a trident topology having a single spout. The topology
> > simply
> > counts the words emitted and stores it in redis [1]. I run this topology
> > on 4 worker
> > nodes.
> >
> > I deliberately shutdown one worker node and restart it and monitor the
> > batch
> > replay (I am logging the transaction id of the batches being emitted).
> >
> > I see the last batch which didn't complete before the shutdown gets
> > re-emitted
> > again and again without succeeding. Any idea why this happens?
> > (I created the spout implementing `ITridentSpout`.)
> >
> > If I don't crash the server, the topology runs normally and finishes.
> >
> > Thanks in advance for any explanation about this.
> >
> > [1] http://redis.io/
> >
> > Regards,
> > Thilina
> >
>
>
>
>

Reply via email to