Clustering and replication

[Janne Peltonen]

On Tue, Jan 30, 2007 at 11:33:20AM +0200, Janne Peltonen wrote:
> >   And GFS 6.1 (current version) has some issues with large directories:
> > 
> > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=214239
> 
> This might be a problem, since we have some users that really have 10000
> messages in their INBOX.

I did some more research on this: correct me if I'm wrong, but it
appears that this GFS problem affects mostly cases where you have to
access all files on a large directory (say, ls) and/or stat the files.
Cyrus has its own cache and index; readdir()s (or whatever was the name
of that system call) should be rare, used mostly during reconstructs,
which are error recovery scenarios in any case.

But fstats seem to be more of a problem; a truss -c on a randomly
selected active imapd process gave 320 fstats / minute...


--Janne