We get this error on one of our nodes: Caused by: org.apache.solr.common.SolrException: There is conflicting information about the leader of shard: shard2 our state says: http://server01:8983/solr/collection/ but zookeeper says: http://server02:8983/collection/
Then I noticed this in the log: ] o.a.s.c.c.ZkStateReader Load collection config from:/collections/collection 2016-02-09 00:09:56.763 INFO (qtp1037197792-12) [ ] o.a.s.c.c.ZkStateReader path=/collections/collection configName=collection specified config exists in ZooKeeper We have a clusterstate.json file left over from 4.X. I read this thread and the first comment or two suggested that clusterstate.json is now broken up and refactored into the collections' configuration: http://grokbase.com/t/lucene/solr-user/152v8bab2z/solr-cloud-does-not-start-with-many-collections So should we get rid of the clusterstate.json file or keep it? We have 4 Solr VMs in our devops environment. They have 2 CPUs and 4GB of RAM. There are about 7 collections shared between then, but all are negligible (like a few hundred kb each) except for one which is about 22GB. Thanks, Mike