Search code examples
solrluceneload-balancingdistributionsharding

Solr 3.6 Distribution dynamically adding shards


I am looking at using distribution and sharding with Solr 3.6 vs Solr 4+ (SolrCloud)

I can see that 3.6 can have multiple shards set up, ideally each shard would rest on a different box. On a large scale once the boxes start to run low on memory I would like to add new shards to the index. From what I have seen this cannot be done/ isn't documented.

Does this require a full re index of the data?

Can a 3 shard indexed be re-indexed into a 4 shard instance?

Can queries still be invoked on the index during a re-index?

What are the space overheads required to re index?

The schema.xml (field names and types) would not be changed, just a new shard location added.


Solution

  • Self Answer- From what I have seen it would be best to stop filling the 3 shards and fill only the new 4th shard with data. Then update the shards parameter to include the new shard to search.