[Fwd: [squid-users]Memory leak problem on epoll i/o squid on IA64]

Forwarded to squid-dev where this discussion belongs.

-----Vidarebefordrat meddelande-----

From: MUTHUKUMAR KANDASAMY <kmuthu_gct@hotamil.com>
To: squid-users@squid-cache.org
Cc: Henrik Nordstrom <hno@squid-cache.org>, David Nicklay <dnicklay@web.turner.com>, Adam Aube <aaube@firstindependent.net>
Subject: [squid-users]Memory leak problem on epoll i/o squid on IA64
Date: 04 Sep 2003 16:01:19 +0530

Hello all ,

I am currently working in a development works in epoll I/O on squid for satisfying the requests of 2000 in IA64.I have passed only 300 req/sec in IA64.My squid is using more than 1.9 GB outof 2 GB.I have tried a lot in the squid to overcome the leaks / problem.I have decreased the swap size from 1.9 GB to 256 MB(Adams view).
But no use .I have use the ElectricFence tool [ Electric Fence 2.0.5] to find the probelms regarding to the memory allocations in squid.

Configure options:
=============
Squid Cache: Version 3.0-PRE3-20030818
configure options: '--prefix=/usr/local/squid' '--enable-epoll' '--disable-poll' '--disable-select' '--disable-kqueue' '--enable-storeio=null,aufs,ufs' '--with-file-descriptors=16384' '--enable-async-io=16' '--with-pthreads' 'CC='

In the compilation of squid ,I used cache_mem 1200 MB cache_dir null /dev/null ,half_closed_clients off ,server_persistent_connection off and other normal options.

kernel parameters:<kernel-2.4.20 Redhat-7.2 IA64>
=============
net.ipv4.ip_forward = 1
net.ipv4.conf.default.rp_filter = 1
kernel.sysrq = 0
kernel.core_uses_pid = 1
fs.file-max = 16384
net.ipv4.ipfrag_low_thresh = 900000
net.ipv4.ipfrag_high_thresh = 1000000
net.ipv4.ipfrag_time = 45
net.ipv4.tcp_rmem = 2000000 2250000 2500000
net.ipv4.tcp_wmem = 1000000 1250000 1500000
net.ipv4.neigh.default.gc_thresh1 = 1024
net.ipv4.neigh.default.gc_thresh2 = 4096
net.ipv4.neigh.default.gc_thresh3 = 8192
net.core.rmem_max = 1500000
net.core.rmem_default = 1500000
net.core.wmem_max = 1000000
net.core.wmem_default = 1000000

Compilation
=========
     # ./squid -z
Electric Fence 2.0.5 Copyright (C) 1987-1998 Bruce Perens.
squid(13049): unaligned access to 0x2000000000cdbfec, ip=0x20000000003a9600
squid(13049): unaligned access to 0x2000000000cdbff4, ip=0x20000000003a9610
squid(13049): unaligned access to 0x2000000000cdbfcc, ip=0x20000000003a9611
squid(13049): unaligned access to 0x2000000000ce3fe4, ip=0x20000000003aa100

     # ./squid
Electric Fence 2.0.5 Copyright (C) 1987-1998 Bruce Perens.
squid(13049): unaligned access to 0x2000000000cdbfec, ip=0x20000000003a9600
squid(13049): unaligned access to 0x2000000000cdbff4, ip=0x20000000003a9610
squid(13049): unaligned access to 0x2000000000cdbfcc, ip=0x20000000003a9611
squid(13049): unaligned access to 0x2000000000ce3fe4, ip=0x20000000003aa100

So what is the problem regarding to this in squid.Is there any way to improve the squid to satisfy more requests in IA64.Please tell the suggesstion to me to do further.

Or any other tool is used to check more.I am using the sys_epoll enabled kernel-2.4.20 on IA64.I think there is no problem with kernel.(or)any other parameters must be tuned to get a good result above then 300 requests.

What is the nessesity that the squid with epoll accessing more than 1.9GB for 300 requets

**********And using the poll I/O in that sys_epoll enabled kernel also consuming the 1.8 GB for 300 requests.******************

So what is the problem regarding to this issue.

Thanks in advance to get the solution.

<Muthukumar>
Received on Thu Sep 04 2003 - 04:59:50 MDT