I have created ES-lab with 3master,2client and 2 datanode. All request is routed through client node.
I have found during indexing if one of my datanode goes down, my entire cluster will get unavailable for 1-2 minutes.
I believe this is because of primary shard belongs to indexing index was also unavailable with the same node.
If I am right then can we make any improvement to minimize such downtime ?