Once you lose the zookeeper quorum, I have seen that the workers keep throwing exceptions that they are not able to connect to the zookeeper. But I haven't seen them die because of this. I have even seen them recover once the quorum is restored.
But if the worker process gets killed, they don't get relaunched because since the communication between nimbus and supervisor is broken. On Thu, Apr 24, 2014 at 4:32 PM, Gaurav Sehgal <[email protected]> wrote: > We have a cluster deployment of Storm and Zookeeper. Where storm is > running on 2 machines and zookeeper is running on 3 machines. > > Now if the zookeeper followers are not available, though the leader is > available storm keeps throwing an exception that it is not able to connect > to zookeeper; and ultimately dies. > > Has anyone encountered this issue? > > Regards, > Gaurav > >
