Yes we are killing it by kill -9.but we are devloping the new product.we have to test that product weather this product working on clusture or not.we are very new on storm.it might be posible that we dont know how to test node failures.can u pls guide me on this. Is there any issue in testing by kill- 9 to test node failure cases? On 28 Jun 2014 18:37, "Naresh Kosgi" <[email protected]> wrote:
> How are killing the worker? If you trying restart the topology with less > or more workers, you have to stop the topology and restart it with more or > less workers. It is not a good idea to just kill one worker (i'm guessing > your using kill -9 to kill one of the workers java process) > > > On Sat, Jun 28, 2014 at 8:51 AM, M.Tarkeshwar Rao <[email protected]> > wrote: > >> Hi all, >> >> I am facing a issue in a deployment of a topology.we are using trident. >> Actually our topology working fine with one worker. when I set the number >> of worker 2 it is working.but when I kill one of the worker process it is >> giving issue. >> > Sometimes it is working but sometimes the communication between two >> workers getting hanged. I analyzed the issue and found that when the killed >> worker again alive on new slot (port) it is working fine. >> But when this killed worker alive on same port comunication between >> workers getting hanged.Can u please help me out to resolve it >> >> We are using scheduler where we sheduled master-batch-controlller, >> co-ordinator and emitor on single machine and in single worker.remaining >> bolts on 2nd worker. >> >> Regards >> Tarkesh >> > >
