On Tue, Jan 22, 2002 at 03:05:55PM -0600, Steve Lord wrote:
> On Tue, 2002-01-22 at 15:04, Austin Gonyou wrote:
> > Scratch that...it just dropped back down about an hour ago. I wonder why
> > that would be? I mounted a loopback device around that time. Could that
> > have anything to do with it?
> About the only thing which would shrink it is a memory shortage, one
> of the things this can cause is a shrink of the inode cache.
It's a known problem in 2.4 that it keeps too many dentries/inodes in memory
around. It shouldn't keep more inodes than dentries for example, but
it does after some stress. It's one of the things that didn't get fixed in t
he VM rewrite. Fortunately it doesn't have too many bad effects except for
some wasted memory and longer lookup times for inodes due to overlong hash