Posted by bennos on February 3, 2010 at 4:36pm
Hi.
Have cloned the mercury(varnish, apache, memcache, apc) install on a dedicated Server with Core i7 Cpu, Raid1 and 8 GB RAM.
All is functional und load test are very good, but here is something that makes me unhappy
Mem: 8053832k total, 7228636k used, 825196k free, 25940k buffers
Swap: 1959920k total, 0k used, 1959920k free, 4222012k cached
The Swap is not used, but there are over 4 GB that are cached.
It seems that the 4GB are kernel caching.
the swapiness is set to 0 via
vm.swappiness=0 in etc/sysctl.conf
Sometimes the cache is going down, but can not found an event in the logfiles.
Tipps?
Comments
Might be normal
My first impression is that is normal. My current Varnish is purring right along using 6.2G of RAM on a 12G machine. As long as the memory usage doesn't appear to be constantly growing out of control, there shouldn't be much cause for alarm. What about it makes you think there's a problem?
More information would be helpful - especially your varnishd config (from conf.d or wherever your distro keeps it).
-Greg
Have installed munin and
Have installed munin and looked in the stats. this is no real problem.
The setting with swapiness = 0 does not deactivate the swap caching, because it is a kernel module. Only modifying the kernel deactivate it. In the current state there are no problems with the config.
I made some load tests. Scaling from 5 to 100 concurrent user and the server has no problem.