Problem with quota for splunk-system-user- Why does size of dispatch folder not correspond to searches in Jobs menu?

[evelenke]

Hi Splunkers,

we had copied many searches to the search head node from indexer and now we have many errors like this: DispatchManager - Search not executed: reason="The maximum disk usage quota for this user has been reached.", usage=12513MB, quota=10000MB, user=splunk-system-user.
In Savedsearch menu all these failed searches are owned by nobody.
In Jobs menu there are many searches with status Done and size around 128 Mb, but the size of the dispatch folder does not correspond to the size of searches in Jobs menu (is it normal?).

We've increased quota up to 30 Gb, and this didn't help.

How can we resolve it or where to look at? Will reassignment to other user help?

[effem2]

This issue can be caused by a static blocksize in the filesystem.  Indications ist the regular occurance of 128M or 250/255/256M sizes across search-results. Splunk does read the blocksizes in full. 

Run "mount" on your cli to see what the mount options are for your filesystem.

In example if you use XFS and you see an option like allocsize=<a big number>, then just get rid of this option. 

Usually the default option are sufficient anyways.

[andygerberkp]

We had this problem on our monitoring console after an upgrade to 8.1.5. Root cause was the DMC Alert - Missing forwarders search - in our very large environment, it was running every 15 minutes and generating 250M of results. This search is set to keep results for 24 hours.  So that's 24G of results in a day.  Changed the settings to keep results for only one hour, and the solved the problem.

In any case, look at the "jobs" on the host and sort by size. You may see something similar - a frequently running job that is saving all results for too long.