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. > > > >
