recent issues with couchbase; eviction of php sessions and expired keys
i have two (1.8.2) cache servers in a cluster, with two vbuckets (one for app, one for php sessions) on separate ports, each using 8GB ram. and, i'm using moxi to communicate with the cluster. i recently had some issues with the cache servers where one was reporting as down / couldn't communicate because one of them had run out of RAM. i restarted the servers, and added a third to avoid any further issues. One week later, the cluster is responsive, but no cache keys are found, and those that are attempted to be set aren't written. no amount of restarting or waiting fixed the issue. i've since started over with a new cluster and all is well for now, but i suspect that the servers were full because php wasn't doing any garbage collection on the sessions, and they were not being expired by the server. i've done a bit of research, and found others that have similar problems, but can't find a clear workaround...
is the issue because this setting tells the server to never expire?
[r@couchbase01 bin]# ./cbstats localhost:11210 all|grep min_data_age
if i set this to 86400, using something like:
cbflushctl localhost:11210 set min_data_age 86400
will the items expire after one day? Is there a way to configure it per bucket? Do I have to set it on each server? Will this setting persist across server restarts?
Do you have any other suggestions on how to properly clean this unneeded info out to avoid any future complications? Is vacuuming the sqllite db still required?