[squid-users] squid / swap usage.

Hey

I just have a query regarding squid's swap usage. We're running squid in
a fairly high load environment, and we're finding that we can't let
squid to utilise more than 20% of the available disk without it going
/really/ heavily into swap and killing the performance.

one box (for example) is an ultra 250 with an A1000 disk array (6 x
~36gb), 2gb memory running solaris 8. When squid initially starts, the
responses times are quick, but once it finishes rebuilding its store,
the responses times sky rocket and squid by then is consuming >1gb of
the box's swap. Is the answer simply that squid doesn't like disks that
large?

gory details: squid 2.4 stable 4.
configure opts: --enable-async-io, --enable-snmp, --enable-dlmalloc,
- --enable-poll, --enable-removal-policies="heap lru"
cache_mem 80
cache_dir aufs /cache/00 10000 16 256
cache_dir aufs /cache/01 10000 16 256
..etc
[priority_paging enabled, with 32768 FD's]

let me know if you need more info.
any advice is muchly appreciated.
Received on Wed Mar 13 2002 - 22:15:46 MST