G'day folks,
Recently, I had the "pleasure" of trying to reconfigure three sibling
proxies for a multi-site organisation. The boxes had been built by
another firm, using Solaris 7 on Ultra 10s. Each machine was supplied
with only 256MB of RAM and 3x18GB external drives (plus one internal).
The squid version is only 2.3STABLE4.
The external drives were built into a single large stripe (~50GB) and
squid had the following configuration entries:
cache_mem was commented out - defaulted to 8MB?
cache_dir ufs /usr/local/squid/cache 50000 16 256
As you can imagine, after a while these machines ran like absolute
(legless?) dogs.
What I've done so far is:
Brought the machines to 1GB RAM
Reset cache_mem and cache_dir as follows:
cache_mem 300
cache_dir ufs /usr/local/squid/cache 45000 128 256
Obviously, the "cache-size" had to be dropped or the filesystem would
have eventually flooded. Secondly, with the large cache size, I felt
it much better to have a "broader" tree allowing for smaller individual
directories.
My other thought is to break up the stripe and specifically define three
separate cache_dirs. The systems owners have also asked if I'd be
willing to upgrade to 2.4STABLE1 (obviously, the answer there is yes).
I'd like to know the list members' opinions on what I've done and on the
proposed unstriping.
Thanks muchly for any advice.
Ciao.
Received on Fri Jun 15 2001 - 03:21:48 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:00:45 MST