When we upgraded to 6.3.3 from 6.2.1 the license usage data has gone from <1% being squashed to almost 40%. On 6.2.1 I had set the squash_threshold value to 5000 from the default 2000 and we were successful with determining the data usage by host. After our upgrade (which included upgrading from SH pooling to SH clustering) this threshold seems to not make any difference - we are back to almost 40% of the information being squashed, without enough information about each host's license usage. I'm pretty sure the new value is being taken into account by splunkd because of the following:
$ splunk btool server list --debug | grep squash
/app/splunk-ix/splunk/etc/slave-apps/lds_all_indexer_base/local/server.conf squash_threshold = 5000
Did something change in 6.3.3 from 6.2.1 that would account for this behavior? Or is there something else I need to change to get this to work properly?
It looks like this is an issue might be the result of a bug (SPL-116968) that was corrected in Splunk version 6.3.5. This bug caused only the default squash_threshold value(2000) to be taken in consideration on the LicenseSlave.