Cluster.routing.allocation.same_shard.host: true not working as expected

Hi, I have a cluster with 29 physical servers running ElasticSearch 2.2.1. Three of the servers are dedicated master nodes and I am running 2 datanodes in each of the other servers (this is 52 datanodes in total). I am creating one index per day with 7 shards and 1 replica.

All the datanodes have the cluster.routing.allocation.same_shard.host: true parameter in the elasticsearch.yml and the network.host parameter is the same for each datanodes pair sharing the same physical host, but for some reason I see some shards with its primary and replica allocated on the same physical server.

|index_20180420|1|p|STARTED|248622278|35.7gb|XX.XXX.XXX.01|datanode29_2|
|index_20180420|1|r|STARTED|248622278|35.7gb|XX.XXX.XXX.01|datanode29_1|

Am i missing something here?

Of course I can fix this rerouting the shard to another server or using the shard allocation awareness, but I'd like to understand why is this happening.

Thanks

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