Shards not balanced across four nodes

[tjyang]

Hi

• I ran into 3 shards un-assigned issue so I ran following elasticsearch command to disable the replica number from 1 to 0.

[root@syslog01 ~]# cat disable_replica.bash
curl -XPUT ‘http://syslog01.test.com:9200/_settings’ -d ‘{“index” : {“number_of_replicas” : 0}}’
[root@syslog01 ~]#

• Now my elasticsearch turned back to green(from yellow).
• This problem is that 0-1-2-3 shards are not balanced across 4 nodes. See following.

• Question 1 is why the shards are not balanced out across when replica setting is “0” ?
• Question 2 is what is the elasticsearch command to spread "0"s to other nodes ?

[jan (Jan Doberstein)]

• answer 1: that can be configured in ES - depending on the Version that is a little different.
• answer 2: it will automatically distributed to all nodes that are part of the cluster and hold indices.

[tjyang]

@jan

Thanks for the lighting response to my question.

• I am using elasticsearch supported by graylog “Graylog v2.4.3+2c41897”.

[root@syslog01 ~]# rpm -qa |grep elastic
elasticsearch-5.6.5-1.noarch
[root@syslog01 ~]#

• “it will automatically distributed to all nodes that are part of the cluster and hold indices.”
  • Does this mean I can just wait it out then “0” and “3” shards will be distributed eventually ?
  • Is there a command to speed it up ?

[jan (Jan Doberstein)]

you can get a balance with some some commands to the elasticsearch index. The important would be cluster.routing.allocation.same_shard.host and then you need to force a rebalance. The following doc can help you.

https://www.elastic.co/guide/en/elasticsearch/reference/current/shards-allocation.html#_shard_balancing_heuristics