Unequal shard allocaiton

i have 8 data node elk and metricbeat configured with shard 2 replica 1. most of the times all of the shard is maximum spread in 3 node . in order to max performance i trying to create index with 5 shard replica 1 and doing a reindex . but i see all primary , replica created in just 3 nodes instead of using all nodes . 1st node has 4 shard , 2nd node 5 and 3rd node 1.

question is how can i ensure all my primary to be in different node so that my performance is better?
out of 8 nodes 4 nodes have disk size 50% of remaining nodes . i.e 4 nodes have 1.5 tb and others have 3tb . is there problem on performance and unequal allocation

What is the output of GET /_cat/nodes?v&h=id,ip,du,dup,v,r,m?

id ip du dup v r
PoPi 908.8gb 90.18 6.5.2 mdi
XX1O 689.8gb 68.45 6.5.2 mdi
OUcj 928.9gb 92.17 6.5.2 mdi
bNap 2.1gb 0.72 6.5.2 i
6aBT 67gb 67.11 6.5.2 i
oQ96 1.9tb 99.09 6.5.2 mdi
z-Mm 948.4gb 94.11 6.5.2 mdi
6QD- 2.9gb 0.97 6.5.2 i
wHZ5 1.8tb 92.04 6.5.2 mdi
dq5d 1.8tb 91.96 6.5.2 mdi
ixvx 804.6gb 79.85 6.5.2 mdi

What is the shard count per node? Do you have shards of very different sizes? How large is the largest shard?

we have 15 metric indexes each 550-600gb with 2 shards 1 replica and 6 rollup indexes less than 110gb same shard/replica set up

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.