okay...
i'd recommend to turn off the 'logging' option in /etc/vfstab for
/www-cachedir first.
then, configure squid to use 'aufs' in cache_dir type.
the poor performance you are expierencing can also be a result of
improperly
configured network interfaces. make sure you have set 100mbit full duplex
for
hme0 (i suppose you _are_ using 100mbit full-duplex).
'ndd -get /dev/hme adv_100_fdx_cap' should return 1
'ndd -get /dev/hme adv_autoneg_cap' should return 0
if autonegotiation is enabled and transceiver adv_10_hdx_cap returns 1, you
should set the parameters using
'ndd -set /dev/hme adv_autoneg_cap 0' and
'ndd -set /dev/hme adv_100_fdx_cap 1'.
you can make the changes permanent by adding 2 lines to the /etc/system
file:
set hme:hme_adv_autoneg_cap=0
set hme:hme_adv_100fdx_cap=1
hope this helps.
cheers
torsten
--------------------------------------------------------------------------
Thanks for your reply...
you really should spread your cache_dirs over different drives.
Yes I think we'll do it..
which sun machine are you using?
Ultra-60
which version of solaris do you use?
Solaris 8
which cache_dir type do you use (asyncufs, ufs)?
ufs
are there any special mount options for your cache filesystems in
/etc/vfstab?
/dev/md/dsk/d60 /dev/md/rdsk/d60 /www-cachedir ufs - yes
logging,noatime
Do you see something special ?
Thanks
-- Isabelle Moullet Groupe conseils et etudes Centre Informatique, UNIL Route de Chavannes 33 CH 1007 - Lausanne Tel: 41 21 692 22 23 Fax: 41 21 692 22 05 e-mail: Isabelle.Moullet@ci.unil.ch ************************************ HelpDesk du lundi au vendredi 8h30 - 17h00 non-stop tél. 021/692 22 11 mail helpdesk@ci.unil.ch ************************************Received on Mon Jan 21 2002 - 03:04:12 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:05:54 MST