Re: [squid-users] squid hangs when "Disk I/O overloading"

From: forgetful tan <forgetful_tan@dont-contact.us>
Date: Tue, 19 Apr 2005 15:48:38 +0800

hi
i used strace to check the squid processes.
11897 ? S 0:00 /home/squid/sbin/squid -D
11899 ? R 276:25 (squid) -D
11903 ? S 0:00 (unlinkd)
the 11897 block at wait4(-1, ..)
the 11899 shows nothing
the 11903 block at read(0, ..)

coz i'm using linux 2.6.10, redhat 9 in these boxes, ldd shows that
squid is to load the /lib/tls/libpthread.so.0
which should be NPTL enabled. maybe that's the problem .. any one have
an idea to check out what's happening ?

thx

Elsen Marc wrote:

>
>
>
>>hi
>>
>>I'm using squid 2.5 STABLE9 in linux. I compiled squid with
>>32 async io
>>threads. these days, my squid stop working with the following message
>>shown in cache.log.
>>
>>2005/04/18 10:37:01| squidaio_queue_request: WARNING - Disk
>>I/O overloading
>>2005/04/18 10:37:01| squidaio_queue_request: Queue Length:
>>current=3651,
>>high=3651, low=321, duration=160
>>2005/04/18 10:37:16| ctx: exit level 0
>>2005/04/18 10:37:16| ctx: enter level 0:
>>'http://xxxxxxxxx/05/0207/03/1BV6VGSS001618FL_2.html'
>>2005/04/18 10:37:16| squidaio_queue_request: WARNING - Disk
>>I/O overloading
>>2005/04/18 10:37:16| squidaio_queue_request: Queue Length:
>>current=3920,
>>high=3920, low=321, duration=175
>>2005/04/18 10:37:27| ctx: exit level 0
>>2005/04/18 10:37:27| ctx: enter level 0:
>>'http://xxxxxxxxx/05/0417/17/1HIAJ1TA00051CC8_4.html'
>>2005/04/18 10:37:27| squidaio_queue_request: Async request
>>queue growing
>>uncontrollably!
>>2005/04/18 10:37:27| squidaio_queue_request: Syncing pending I/O
>>operations.. (blocking)
>>
>>I then run squid -k shutdown twice. The first time, squid
>>complain that
>>2005/04/18 10:37:27| Killing RunCache, pid 10657
>>2005/04/18 10:37:27| kill 10657: (1) Operation not permitted
>>
>>
>
> Make sure to run this command as root, or as the effective user
>squid runs as, when this command is given.
>
>
>
>>and the second time, it restart instead of shutdown.
>>
>>
>
> What was in cache.log when this happened ?
>
>
>
>>Have any one got an idea what's wrong with my squid ?
>>
>>
>>
>
> - What's your avg. http reqs/sec during normal load ?
> - Check whether disk/io is a bottleneck , using OS commands.
>
> M.
>
>
>
>
Received on Tue Apr 19 2005 - 01:49:24 MDT

This archive was generated by hypermail pre-2.1.9 : Sun May 01 2005 - 12:00:04 MDT