@DavidTurner I did one experiment. I removed one node (node1) and start indexing. Now all traffic is moved to node2 according to thread_pool But indexing rate is same on both node (node2 and node3).
Some weird thing happening.
Is the distribution of primary and replica shards even?
@Christian_Dahlqvist I dont have replica shards. And all primary shards distributed equally on both the server. And shards are not an even . Also i am using heavy bulk update.
Also one experiment i did. I reduced my bulk queue size from 10000 to 1000. Which caused rejection of document if queue size exceeded from 1000. I was expecting it will be route to other node whose bulk queue size is 10k.
Isn't it strange ?
No, it gets queued where the primary shard resides for shard.
But my all bulk request getting queued on one node only. And in my shard allocation, it shows my primary shards is well distributed . So according to this i guess some request should be equally queued on all nodes.
Because i am receiving bulk request for all indexes.
Please could you share the full outputs of the following commands from your cluster, all taken at about the same time? I know we've seen some of these before, but maybe not all at the same time so they won't match up.
GET /_cat/shards
GET /_cat/thread_pool
GET /_tasks?detailed
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.