Partial failure

[Alex_Davidovich]

I have a 3 nodes cluster with 1 primary and 2 replica shards.
I inserted a document and got this response:

IndexResponse[index=events_1513522587069,type=events,id=112,version=1,result=created,shards=
{"total":3,"successful":1,"failed":1,"failures":
[{"_index":"events_1513522587069","_shard":0,"_node":"87JEfX7tTYC9jIN3c-8hhQ","reason":
{"type":"node_disconnected_exception","reason":"[node-2][172.16.65.117:9300]
[indices:data/write/bulk[s][r]] disconnected"},"status":"INTERNAL_SERVER_ERROR","primary":false}]}]

I understand that only 1 shard was able to persist the document (the successful one...), 1 failed.
What about the third?
In addition, why 1 shard is written when didn't get ack from the rest of the nodes? I don't want to have this behavior.

[warkolm (Mark Walkom)]

What version are you on? How many shards and replicas in the index?

[Alex_Davidovich]

Version 5.4.
1 index with 1 primary shard and 2 replicas.
3 nodes in the cluster.