Greetings,

Hello friends, my name is Marcos Cavallari Conrado and I�m an ISP 
administrator in Brazil. We implemented a server to work as a 
transparent-PROXY with SQUID and Linux Red Hat 7.3 as its OS and 
Linux-2.4.20 as kernel.

The hardware we use exclusively for proxy-cache is:
DeLL Power Edge 1400 XEON 1.6 (Hyper Threading, so the kernel is compiled 
SMP)
Seagate 10K rpm . 30 GB with reiserfs (notail and noatime) for the cache
1 GB . RAM

This server has 1800+ clients.

I compiled SQUID-2.5-STABLE3 using:
./configure --enable-linux-netfilter --disable-ident-lookups 
--enable-removal-policies=heap,lru --enable-delay-pools 
--enable-cache-digests --enable-pool --enable-truncate 
--enable-stacktraces --enable-xmalloc-statistics --enable-underscores 
-with-pthreads --enable-storeio=aufs --enable-async-io=128 
--enable-kill-parent-hack --enable-time-hack 
--enable-default-err-language=Portuguese --with-aufs-threads=10

My resumed squid.conf:
----------------------
cache_mem  48 MB
cache_replacement_policy heap LFUDA
memory_replacement_policy heap LFUDA
maximum_object_size 150096 KB
minimum_object_size 0 KB
cache_access_log /var/squid/logs/access.log
cache_log none
cache_store_log none
half_closed_clients off
cache_swap_high 95
cache_swap_low 90
cache_dir aufs /squid/cache1 10000 25 256
cache_dir aufs /squid/cache2 10000 25 256
cache_dir aufs /squid/cache3 10000 25 256
buffered_logs on
------------------------


The performance is OK, but I�m getting several .squidaio_queue_request: 
WARNING - Queue congestion. and several system halts.
Could anyone more experienced than me with squid give me any advices about 
what can I do to avoid these problems and improve stability/performance 
or if my hardware is not enough for this type of service. 
I would like to thank you all in advance for reading this email and spending your 
valuable time with my doubt.
Congratulations for SQUID team for the development of such a master piece 
of software.


All yours -> Marcos Cavallari Conrado (Belo Horizonte - MG / Brazil)

Reply via email to