openstack team mailing list archive

[Robert van Leeuwen <Robert.vanLeeuwen@xxxxxxxxxxxxx>]

> XFS keeps some useful runtime statistics
> on /proc/fs/xfs/stat that can be maybe used 
> to test this theory:  http://xfs.org/index.php/Runtime_Stats 
Thx, I did not know about this.
Directory lookups is currently at 3.817.426.894 and going up by 500/sec
ig_missed is at 3.051.577.459 and going up by 200/sec
This probably points to a cache that's to small.

>Does anyone knows how big is the XFS inode and directory caches?
They are "part" of the slabcache and can be seen by using slabtop

You can tune the amount of memory used indirectly by setting the vfs_cache_pressure.
This will decide at what rate the kernel tries to reclaim entry's. 
Default is 100 (described as "fair" rate), setting it to 0 will never reclaim (and the oom killer will probably have some fun).
See: http://www.kernel.org/doc/Documentation/sysctl/vm.txt

Cheers,

Robert van Leeuwen