Slow/locking UFS/VFS in 6.1?

[Philip Murray]

Hi,

I just upgraded a webserver (serving a pretty busy site) from a Dual  
Xeon 2.4Ghz with a 3Ware 7850 8-port (8x Seagate 120GB IDE drives in  
RAID5) running FreeBSD 4.11 to a Dual Opteron 248 with a 3Ware  
9550SX-8LP (8x Hitachi SATA2 250GB drives in a RAID5) running FreeBSD  
6.1-Stable.

The site runs considerably slower on the faster machine running 6.1,  
things seem to be blocking on IO a lot more. For instance a page  
could load very quickly from Apache, and if you reload straight away  
it can take up to 20 seconds for the page to load again.

When this happens it seems a lot of Apache processes get stuck in the  
UFS state (in top).

Is there a way to debug what's going on? or work around it. It should  
be much much faster which the faster disks and raid controller. Is  
this lock contention in VFS?

Thanks!

Phil


_______________________________________________
freebsd-stable@... mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscribe@..."

[Ivan Voras-2]

Philip Murray wrote:

> When this happens it seems a lot of Apache processes get stuck in the
> UFS state (in top).

I've seen this when there's a huge directory involved, such as created
by a runaway web cache function, PHP sessions, or "PHP accelerator" that
creates a lot of small cache files. If so, you'll probably need to
delete that directory completely and make sure it doesn't happen again
(e.g. use a multi-stage directory tree instead of a big one).

You might have luck with increasing dirhash size (vfs.ufs.dirhash_maxmem).

_______________________________________________
freebsd-stable@... mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscribe@..."