On Thu, May 12, 2011 at 10:21:59AM -0400, Jeff Blaine wrote: > I'm pretty confused. > > We have an Apache-served RT 3.8.7 instance with Postgres > backend. It's using mod_perl. All 10 current httpd > processes are over *300MB* each already after a host > reboot 30 hours ago. > > We have a NON-RT system, with the same OS and package > versions, serving 10 times more people via mod_php + > Postgres (a SourceForge-like system). All current > httpd processes are ~50MB after being up for 5 days. > > What gives? Any ideas? > >
The memory footprint depends completely on the coding and resource needs of the software. RT caches a lot of data to improve performance. There are also some code areas that cause bad memory bloat if used. There is one issue with reading in the meta-data for every ticket found in a search when you select one of the search result tickets for display. The effect scales with the size of the result set. We have blown out our VM footprint to almost 2GB per httpd process and needed to restart httpd to regain memory and reasonable performance. I think that problem was fixed in 3.8.9 or 10 which is why we are eager to upgrade. Other than that our starting VIRT footprint for Apache/mod_perl is just over 300MB. Cheers, Ken
