We have Elasticsearch cluster of 3 nodes where 1 node is dedicated master and 2 nodes are dedicated data nodes. we have around 200GB of data on server and heavy _bulk and _search operations. The nodes configuration is
Master node :-
RAM - 8GB
Cores - 4
First Data node :-
RAM - 20GB
Cores - 8
Second Data node :-
RAM - 8GB
Cores - 4
We have 5 shards and 1 replica
for all nodes.
We are getting socket hang up
error very frequently for _bulk and _search both, we tried to make all nodes master eligible but it did not work.