[squid-users] Fwd: 2 instances - Fixed Issue

2007-10-05 Thread Frank Ruiz
Greetings,

Hopefully, I can save someone a few hours of heartache.

Summary:
If you plan to use squid on solaris 10 x86, and configure to use
/dev/poll, make sure its patched.

Thanks everyone for reading the spam below if you did.



-- Forwarded message --
From: Frank Ruiz [EMAIL PROTECTED]
Date: Oct 5, 2007 12:18 PM
Subject: 2 instances
To: squid-users squid-users@squid-cache.org


Also,

Just killed all connections to squid process. CPU still pegged at 15%
and decreasing

Here is what the process is doing:

ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 1
accept(8, 0xFD7FFFDFFB90, 0xFD7FFFDFFB7C, SOV_DEFAULT) = 12
getsockname(12, 0xFD7FFFDFFB80, 0xFD7FFFDFFB7C, SOV_DEFAULT) = 0
fcntl(12, F_GETFL)  = 130
fcntl(12, F_SETFD, 0x0083)  = 0
fcntl(12, F_GETFL)  = 130
fcntl(12, F_SETFL, FWRITE|FNONBLOCK)= 0
accept(8, 0xFD7FFFDFFB90, 0xFD7FFFDFFB7C, SOV_DEFAULT) Err#11 EAGAIN
write(3, \f\0\0\0\0\b\0\0\f\0\0\0.., 16)  = 16
ioctl(3, DP_POLL, 0x005036B0)   = 1
read(12,  G E T   / f i l e / 1 1.., 4095)= 51
write(3, \f\0\0\0\0\b\0\0\f\0\0\0.., 32)  = 32
ioctl(3, DP_POLL, 0x005036B0)   = 1
write(12,  H T T P / 1 . 0   2 0 0.., 1736)   = 1736
close(12)   = 0
write(3, \f\0\0\0\0\b\0\0, 8) = 8
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 0
ioctl(3, DP_POLL, 0x005036B0)   = 1



I have 2 squid instances.

Both are taking the same amount of connections, and both are
connecting to the same exact pool of origins via a lb.

 2344 root   13G   13G cpu1 00   4:03:10  25% squid/1
 2096 root   13G   13G sleep   310   4:47:22 9.2% squid/1


After about 14 hours of runtime, the instances hit 25% utilization,
and then never seem to restabilize.


This is all I see in my cache.log for the instance that is maxed out:

2007/10/05 10:47:25| clientReadRequest: FD 107 (82.38.189.46:5430)
Invalid Request
2007/10/05 11:05:13| WARNING: unparseable HTTP header field
{Accept-CharsetGET /pict/320155568274_1.jpg HTTP/1.1}
2007/10/05 11:06:55| parseHttpRequest: Unsupported method 'Connection:'
2007/10/05 11:06:55| clientReadRequest: FD 1808 (84.71.71.234:35312)
Invalid Request
2007/10/05 11:06:55| parseHttpRequest: Unsupported method 'Connection:'
2007/10/05 11:06:55| clientReadRequest: FD 1007 (12.25.108.29:63647)
Invalid Request
2007/10/05 11:22:01| parseHttpRequest: Unsupported method 'Connection:'
2007/10/05 11:22:01| clientReadRequest: FD 1612 (81.104.41.63:48329)
Invalid Request
2007/10/05 11:22:01| parseHttpRequest: Unsupported method 'Connection:'
2007/10/05 11:22:01| clientReadRequest: FD 1685 (74.236.38.154:50482)
Invalid Request
2007/10/05 11:22:06| parseHttpRequest: Unsupported method 'Connection:'
2007/10/05 11:22:06| clientReadRequest: FD 6278 (83.112.151.249:53849)
Invalid Request


The box is a 2 cpu dual core, so each squid instance maxes out at 25% cpu.
The are strictly in memory cache (no disk), and they each have 9G of
RAM per instance.

Can someone give me an idea of what is happening?

Thanks.
**

Greetings,

Also, here is some truss output to show what the process is doing.

accept(9, 0xFD7FFFDFFAF0, 0xFD7FFFDFFADC, SOV_DEFAULT) Err#11 EAGAIN
write(5635,  G E T   / f i l e / 2 7.., 842)  = 842
write(5636,  G E T   / f i l e / 2 5.., 654)  = 654
write(5637,  G E T   / f i l e / 2 3.., 475)  = 475
write(5638,  H T T P / 1 . 

Re: [squid-users] Fwd: 2 instances - Fixed Issue

2007-10-05 Thread Adrian Chadd
On Fri, Oct 05, 2007, Frank Ruiz wrote:
 Greetings,
 
 Hopefully, I can save someone a few hours of heartache.
 
 Summary:
 If you plan to use squid on solaris 10 x86, and configure to use
 /dev/poll, make sure its patched.

:P Which patch did you apply?



Adrian


-- 
- Xenion - http://www.xenion.com.au/ - VPS Hosting - Commercial Squid Support -
- $25/pm entry-level bandwidth-capped VPSes available in WA -