Distributed Solr: different number of results each time

[ffriend]

Hi all,

I'm using faceted search in SolrCloud to get some statistics on my corpus. I'm using query very similar to:

http://master-node:8983/solr/select?q=field1:value1&distrib=true&facet=true&facet.field=field1&facet.limit=50000&rows=0&fq=field2:value2&fq=field3:value3&fq=somemore:filters

However, in production with high number of documents, shared between ~5 machines, master node returns different number of results very frequently.

I expected it is because of small timeout for shards, so I added following params to the search request handler in solr/conf/solrconfig.xml:

    <int name="shard-socket-timeout">10000</int>
    <int name="shard-connection-timeout">10000</int>

but it hadn't helped.

Any ideas of what may be the reason for such a behavior?

[ffriend]

It seems like the error was caused by wrong list of shard URLs, kept in ZooKeeper. One possible workaround is to specify list of shards manually with

    shards=slave-node1,slave-node2,slave-node3,...

(see SolrCluster documentation for details)