[squid-users] Memory leak?

Hi there,

I'm running a Squid 2.5.STABLE3 Squid on Redhat 9 linux installed on a dual P4 2.4GHz, 3GB RAM and 6xSCSI 320 harddisks. The proxy is configured to work as transparent.
The problem is that after the proxy starts being hit, the memory consumption is going high as time passes until it starts using swap.
I have configured cache_mem to 192MB, and the squid process is using around 500MB of RAM but the rest is being eaten somehow. When I turn the squid process down (regulary), only the memory used by Squid (500MB) are being released while around 2.5 GB are left allocated, and the only way to deallocate them is to restart the server.
I have upgraded redhat to its latest packages, and tried different malloc libraries (GNU, DL,...) but it didnt help. I turned memory pools off, but that didnt help either. I applied the memory leak patches for squid 2.5.3 although I didnt use those functions but everything remained the same.
I'm using ReiserFS filesystem on the disks where the cache resides (notail, noatime,nodiratime).
Can somebody give me a hint on what could be going wrong here. Any hint is apprecated.
Compile options: --enable-cache-digests --with-aio --enable-snmp --enable-gnuregex --enable-removal-policies --enable-storeio=ufs,diskd --enable-linux-netfilter --disable-ident-lookups --enable-poll --enable-underscores --enable-xmalloc-statistics
ReiserFS: 3.6.25
Kernel: 2.4.20-20.9smp
glibc: 2.3.2-27.9

Hope somebody has an idea of whats going on as I'm running out of options here :)

regards,
Shpend Bakalli
Received on Mon Sep 01 2003 - 03:56:16 MDT