After moving a shard related files do not disappear from the disk

I have a big index with ~16Gb shards. Recently I moved one of these big shards to less disk constrained node. The storage in the new node was reduced the expected size, and the shard was exchanged by a smaller one. But the disk storage from the parting shard never was recovered. Elastic search version is 2.4

Diggin in elastic search directories I found this

elasticsearch/clustername/nodes/0/indices/indexname$ du -hcx .
8.0K	./_state
8.0K	./1/_state                       <--- shard 0 does not live here
119M	./1/translog
15G	./1/index
15G	./1
8.0K	./2/_state
114M	./2/translog                  
16G	./2/index                       
16G	./2
8.0K	./4/_state                        <--- shard 4 does not live here
13M	./4/translog         
16G	./4/index
16G	./4
8.0K	./3/_state
421M	./3/translog
16G	./3/index
16G	./3
62G	.
62G	total

looking to the shards listings:

curl -s 'http://localhost:9200/_cat/shards?h=n,i,s,p,sto' | grep the_index | sort
node-01 the_index        1 p   14.7gb 
node-01 the_index        2 p   15.8gb 
node-01 the_index        3 r   15.4gb 
node-02 the_index        0 r   16.2gb 
node-02 the_index        2 r   15.7gb 
node-02 the_index        3 p   13.1gb 
node-02 the_index        4 p   18.1gb 
node-03 the_index        0 p   15.1gb 
node-03 the_index        1 r   14.3gb 
node-03 the_index        4 r   18.2gb 

So in node 01 wich is the one I'm looking there is no shard 4 of the index but the files still there.

Is there a way to ask the cluster to do cleanup of these directories ?

I ought to remove them when it doesn't need them any more. The usual reason for the files to stick around for a while is that you have a scroll open for the shard that hasn't yet timed out. They ought to vanish once the scroll times out. I'm not sure of a good way to see the count of open contexts now that most of the shard has moved off of the node though.

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