Leader not elected after after node goes down in SOLR

1
votes

I am using SOLR cloud with Tomcat. When one of my leader nodes is down(e.g. Shard 1); its replica is not elected as leader, but is still up. I am using a stand-alone zookeeper(well, 3 of them). When I restart the failed node, the node is not visible, and still no leader for Shard 1 is elected. The replica is still up and leader-less. This remains so even if I restart zookeeper and all nodes.

I am using SOLR 4.7.0 The setup I have is: 1 tomcat instance per shard. I have 2 shards and two replica shards. The replicas are on separate servers. I use an external zookeeper instance to manage all four nodes/cores. I have also tested with 3 external zk instances, but had the same result.

Can anyone tell me where to look if this is a configuration issue or sounds familiar?

1
tomcatsolrconfigurationcloudapache-zookeeper
Are you using the embedded zookeeper or external zookeeper instance? – buddy86
I am using an external zookeeper instance for this infrastructure. – Ayoub
The only way I could get a leader elected again is by deleting the index and restarting both zookeeper and all tomcat servers(luckily I am in the dev and test phases right now) – Ayoub
I have downgraded from SOLR 4.7.0 to SOLR 4.2.1, and election is working properly again. – Ayoub

1 Answers

0
votes

I have downgraded from SOLR 4.7.0 to SOLR 4.2.1, and election/node restart is working properly again. My guess is this problem was due an incompatibility between the newest SOLR 4.7 version and the very old zookeeper version that I was using.