On 4/16/2015 7:08 AM, Adir Ben Ami wrote: > I am using Solr 4.10.0 with tomcat and embedded Zookeeper. > I use SolrCloud in my system. > > Each Shard machine try to reach/connect with other cluster machines in order > to index the document ,it just checks if it is still the leader. > I don't use replication so why does it has to check who is the leader? > How can I bypass this constraint and make my solrcloud not use > ClusterStateUpdater.checkIfIamStillLeader when i am indexing?
You might not need that functionality, but Solr must address the general case, which includes multiple replicas for each shard,where one of them will be leader. I hope this is a test installation ... running in production without fault tolerance is a bad idea. Using the embedded zookeeper in production is another bad idea, for the same reason - fault tolerance. You can file an issue in Jira for a configuration mode where the leader check is disabled. I would oppose having that happen automatically ... another replica could be added to the cloud at any time. Thanks, Shawn