Shard rebalancing after node restart


I have a 4 node cluster. When I restart one of these nodes it seems to sync all the shards it originally had from the other nodes. This is inspite the fact that this node has most of the data already on disk.

How can I stop such massive data transfer from other nodes and instead ask the restarted node to take most of the data from its own disk, and then do sync for data that does not exist on it?

Upgrade to ES 1.6.0, there are improvements to help this (shard flushing) :slight_smile: