Re: Interrupting controlled shutdown breaks Kafka cluster

2014-11-10 Thread Solon Gordon
Thanks, Neha. I tried the same test with 0.8.2-beta and am happy to report I've been unable to reproduce the bad behavior. I'll follow up if this changes. On Sun, Nov 9, 2014 at 9:30 PM, Neha Narkhede wrote: > We fixed a couple issues related to automatic leader balancing and > controlled shutdo

Re: Interrupting controlled shutdown breaks Kafka cluster

2014-11-09 Thread Guozhang Wang
Solon, You may be hitting this: KAFKA-1305 . You can try with 0.8.2-beta version and see if this issue is re-producible. Guozhang On Fri, Nov 7, 2014 at 11:52 AM, Solon Gordon wrote: > We're using 0.8.1.1 with auto.leader.rebalance.enable=true.

Re: Interrupting controlled shutdown breaks Kafka cluster

2014-11-09 Thread Neha Narkhede
We fixed a couple issues related to automatic leader balancing and controlled shutdown. Would you mind trying out 0.8.2-beta? On Fri, Nov 7, 2014 at 11:52 AM, Solon Gordon wrote: > We're using 0.8.1.1 with auto.leader.rebalance.enable=true. > > On Fri, Nov 7, 2014 at 2:35 PM, Guozhang Wang wrot

Re: Interrupting controlled shutdown breaks Kafka cluster

2014-11-07 Thread Solon Gordon
We're using 0.8.1.1 with auto.leader.rebalance.enable=true. On Fri, Nov 7, 2014 at 2:35 PM, Guozhang Wang wrote: > Solon, > > Which version of Kafka are you running and are you enabling auto leader > rebalance at the same time? > > Guozhang > > On Fri, Nov 7, 2014 at 8:41 AM, Solon Gordon wrote

Re: Interrupting controlled shutdown breaks Kafka cluster

2014-11-07 Thread Guozhang Wang
Solon, Which version of Kafka are you running and are you enabling auto leader rebalance at the same time? Guozhang On Fri, Nov 7, 2014 at 8:41 AM, Solon Gordon wrote: > Hi all, > > My team has observed that if a broker process is killed in the middle of > the controlled shutdown procedure, th

Interrupting controlled shutdown breaks Kafka cluster

2014-11-07 Thread Solon Gordon
Hi all, My team has observed that if a broker process is killed in the middle of the controlled shutdown procedure, the remaining brokers start spewing errors and do not properly rebalance leadership. The cluster cannot recover without major manual intervention. Here is how to reproduce the probl