Re: [squid-users] Returning to 200req/sec - mission failed

From: Andriy Korud <akorud@dont-contact.us>
Date: Thu, 19 Feb 2004 15:37:43 +0200

No, nothing unusual in log.

We have apporox 2500 clients, however they are NAT'ed before Squid into 128
public addresses, so Squid see ~100 very-heavy-browsing clients.
Maybe problem is here?

Andriy

> Do you see any error messages in the cache.log?
>
> How many clients are accessing the Squid?
>
> Peter
>
> >>> Andriy Korud <akorud@polynet.lviv.ua> 19-02-2004 14:14:36 >>>
> Hi,
> I'm just trying to build large Linux/Squid server and the problem is:
> Srever: Intel Xeon 2.8/HT, 1G RAM, Linux kernel 2.6.
>
> squid compile options:
> ./configure --enable-async-io --enable-truncate
> --disable-ident-lookups
> --enable-linux-netfilter --enable-poll --enable-cache-digests
> --disable-htcp
> --enable-snmp --disable-wccp --disable-referer-log
> --disable-useragent-log
> --disable-icmp --disable-delay-pools --enable-underscores
>
> Squid.conf (without acl):
>
> cache_mem 150 MB
> memory_pools on
> memory_pools_limit 100 MB
>
> cache_dir aufs /cache1/squid 20480 38 256
> cache_dir aufs /cache2/squid 20480 38 256
>
> #cache_access_log /var/log/squid/access.log
> cache_access_log none
> cache_log /var/log/squid/cache.log
> cache_store_log none
> half_closed_clients off
>
> Storage on dedicated 15k SCSI disks with ReiserFS (noatime, notail).
>
> And what:
> at approx 78req/s (transfer ~4Mbit) I already have 90% CPU usage (and
> disk
> activity is very low). And this is only half-load of our network :-(
>
> Maybe anybody has some ideas?
>
> Andriy
>
>
>
>
>
>
Received on Thu Feb 19 2004 - 06:37:51 MST

This archive was generated by hypermail pre-2.1.9 : Mon Mar 01 2004 - 12:00:02 MST