3

I have setup solrcloud cluster with 2 shard 2 replica configuration. But when I start my shard nodes and only one replica node for each shard, the shard leader nodes continuously prints a message: waiting until we see more replicas up,total 3 found 2 timeoutin 175000, the wait time for other replicas to come up here is around 3 min.

My question is how to reduce this wait time, to speedup development ?

Nikana Reklawyks
  • 3,233
  • 3
  • 33
  • 49

1 Answers1

7

In solr.xml, add a cores attribute of leaderVoteWait=0. It defaults to 180000 (3 minutes).

This is simply to protect against starting the cluster with an old node - you don't want it to become the leader before other nodes get to participate in the election - so we wait until the known replicas are up to ensure they can participate in the election.

If you are careful to not start the cluster in a way that an out of data node comes up before up to date nodes, it's not necessary.

  • Mark
Mark
  • 131
  • 2
  • Is leaderVoteWait only used on start-up (seems so?) or does it apply post-startup? This seems like a setting that should be per-collection or global (and thus belongs in zookeeper), not per-node. And finally, can this leaderVoteWait setting be used to force an old restored replica (from backup) to replace technically newer replicas on other nodes that you start up after the leaderVoteWait period? Even though the "version" (timestamp) and "generation" are greater? – David Smiley Jan 31 '14 at 20:51