Forcemerge operation failing to include second digit

[danielkasen]

Elasticsearch version (bin/elasticsearch --version): 6.5.1

Description of the problem including expected versus actual behavior:

When executing a forcemerge operation it would appear that the check to see if it should proceed to the max_num_segments works, but the actual FM operation only includes the first digit. For example if you were to issue the cmd:

POST nginx-2018.12.05.14.00-1/_forcemerge?max_num_segments=20

It would forcemerge each shard to 2 segments opposed to the defined 20. Obviously this causes a large amount of resource utilization. I have tested this against several indices and all show the same result.

Now if the index already had only 20 segments then this would return immediately. So I believe it is in the execution step.

Steps to reproduce:

1. Create an index
2. Get it to create 30 segments per shard
3. Try to forcemerge the index down to 25 segments and it should actually reduce it to 2 per shard

[danielkasen]

Actually I think I am misreading these numbers. It would appear that it always tries to force merge down to 1. Did we remove the option to pass a number to this in the 6.5 release and not document it? This is causing a serious strain on our cluster.

[elasticmachine]

Pinging @elastic/es-search

[elasticsearchmachine]

Pinging @elastic/es-distributed (Team:Distributed)

[javanna]

It would appear that it always tries to force merge down to 1. Did we remove the option to pass a number to this in the 6.5 release and not document it? This is causing a serious strain on our cluster.

Yes, that is the case, sorry for the very late reply, see #31689 .