Hi again, a follow-up on this: I ended up fixing it by uploading a new version of clusterstate.json to Zookeeper with the missing hash ranges set (they were easily deducible since they were sorted by shard name). I still don't know what the correct solution to handle index corruption (where all replicas of a shard needs to be repaired) while still keeping the cloud available for search, would be?
Thanks, Rikke On Aug 22, 2013, at 21:27 , Rikke Willer <r...@dtic.dtu.dk<mailto:r...@dtic.dtu.dk>> wrote: Hi, I have a Solr cloud set up with 12 shards with 2 replicas each, divided on 6 servers (each server hosting 4 cores). Solr version is 4.3.1. Due to memory errors on one machine, 3 of its 4 indexes became corrupted. I unloaded the cores, repaired the indexes with the Lucene CheckIndex tool, and added the cores again. Afterwards the Solr cloud hash range has been set to null for the shards with corrupt indexes. Could anybody point me to why this has occured, and more importantly, how to set the range on the shards again? Thank you. Best, Rikke
