1

I have weird requirement:

It is not possible to set up dedicated SOLR servers in my environment, they have to be embedded in JVMs running tomcat instances. It works perfect for single server - But I have multiple load balanced servers sharing the same database - so I need to keep those embedded SOLR instances in synch. Each local tomcat instance will be querying and updating local SOLR index, but I need to propagate updates to other nodes somehow. And of course, if one of nodes is not available at the time, it must get updates in batch after it is back online

It seems that standart replication is not a solution for me, as this requires single master and pushes lucene index files around. What alternatives I got?

Chris Martin
  • 30,334
  • 10
  • 78
  • 137
Konstantin Pribluda
  • 12,329
  • 1
  • 30
  • 35
  • the short answer is "no", solr does not support anything that is even close to master to master replication. The way you/I/everyone updates solr will populate incremental index updates, which will create same chunks for the underlying lucene index. Replication cycle will only trigger when the master server has a version update, and when this happends, slave will think it's index is corrupted (because it has the same chunk with different content) will download the complete index from master and continue. – kali Apr 29 '14 at 15:49
  • also this may or may not be a solution for you https://cwiki.apache.org/confluence/display/solr/SolrCloud – kali Apr 29 '14 at 16:16

0 Answers0