Thank you Leonardo
in this URL you can see the graphics: http://smb.cancun.net.mx/squid/
I noted also the LOTS of Open FD in cache .log
I Included at the website a cache_log.txt file,
this is a extract of the log after
a "squid -k restart" http://smb.cancun.net.mx/squid/cache_log.txt
there are a lot of lines that said something like
2009/08/03 21:23:11| 65536 entries written so
far.
2009/08/03 21:23:11| 131072 entries written so
far.
2009/08/03 21:23:11| 196608 entries written so
far.
2009/08/03 21:23:11| 262144 entries written so
far.
2009/08/03 21:23:11| 327680 entries written so
far.
2009/08/03 21:23:11| 393216 entries written so far.
then the restart:
2009/08/03 21:23:15| Finished. Wrote 3482566
entries.
2009/08/03 21:23:15| Took 4.48 seconds
(777704.28
entries/sec).
CPU Usage: 3969.994 seconds = 2319.688 user +
1650.306
sys
Maximum Resident Size: 0
KB
Page faults with physical i/o:
301
Memory usage for squid via
mallinfo():
total space in arena: 121088
KB
Ordinary blocks: 120691 KB 5287
blks
Small blocks: 0 KB 1
blks
Holding blocks: 343612 KB 1504
blks
Free Small blocks: 0
KB
Free Ordinary blocks: 396
KB
Total in use: 464303 KB
383%
Total free: 396 KB
0%
2009/08/03 21:23:15| aioSync: flushing pending
I/O
operations
2009/08/03 21:23:15| aioSync:
done
2009/08/03 21:23:15| aioSync: flushing pending
I/O
operations
2009/08/03 21:23:15| aioSync:
done
2009/08/03 21:23:16| aioSync: flushing pending
I/O
operations
2009/08/03 21:23:16| aioSync:
done
2009/08/03 21:23:16| aioSync: flushing pending
I/O
operations
2009/08/03 21:23:16| aioSync: done
and finally, alot of lines like this
2009/08/03 21:23:16| Open FD UNSTARTED 10
/var/spool/squid/06/77/000677EE
2009/08/03 21:23:16| Open FD UNSTARTED 18
/var/spool/squid/04/81/00048198
2009/08/03 21:23:16| Open FD UNSTARTED 19
/var/spool/squid/02/B6/0002B610
2009/08/03 21:23:16| Open FD UNSTARTED 21
/var/spool/squid/09/54/00095456
2009/08/03 21:23:16| Open FD UNSTARTED 23
/var/spool/squid1/04/EB/0004EB5D
2009/08/03 21:23:16| Open FD UNSTARTED 24
/var/spool/squid1/12/CB/0012CBDC
2009/08/03 21:23:16| Open FD UNSTARTED 26
/var/spool/squid/00/FB/001CFBC8
2009/08/03 21:23:16| Open FD UNSTARTED 27
/var/spool/squid1/07/2C/00072C10
2009/08/03 21:23:16| Open FD UNSTARTED 28
/var/spool/squid1/16/24/001624E2
2009/08/03 21:23:16| Open FD UNSTARTED 29
/var/spool/squid/16/AF/0016AF91
2009/08/03 21:23:16| Open FD UNSTARTED 31 /var/spool/squid1/02/F8/001EF81C
this are lost file descriptors from the las restart yesterday
Thank you Leonardo, I dont know if is something
related to the object cleaning process
Im using HEAP GDFS for memory and cache replacement policy
Juan Bou
At 04:39 03/08/2009, Leonardo Rodrigues wrote:
juanb@xxxxxxxxxxxxx escreveu:
Hello
I have been running 3 squid servers with
squid-3.0.STABLE15-2, std rpm on a Linux Fedora 10,
I have been running those servers as a
transparent cache and the traffic to them is redirected by some
Linux and mikrotik routers,
Everything goes fine, but day to day we are
losing file descriptors, here is a weekly and monthly graphic,
sometimes a graphic says more than 100 words
if anyone had the same experience and have a solution please let me know
there are 2 graphics from different machines,
running exactly the same versions of everything, doing the same
If I wait more days the machine will crash but
a simple restart make the thing work again, we restart the squid
every 2 or 3 weeks to be safe, but I think it is not the solution
images didnt get attached to the list ...
please post them in some http server
(imageshack.us for example) and send the URLs.
there was a known bug for file descriptors
some time ago, but that should have been fixed in 3.0 STABLE11.
at the time i opened the bugtraq first
explain the bug and giving some further
details, the easy way of identify those lost
filedescriptors was asking squid to shutdown.
On shutdown, i could see something like this on cache.log:
i can also see LOTS of OpenFD messages during shutdown:
2008/07/23 23:46:51| Open FD 98 /var/squid/05/EB/0005EBB7
2008/07/23 23:46:51| Open FD 99 /var/squid/02/3D/00023D8C
2008/07/23 23:46:51| Open FD 100 /var/squid/05/EB/0005EBB7
2008/07/23 23:46:51| Open FD 101 /var/squid/04/74/00047432
2008/07/23 23:46:51| Open FD 102 /var/squid/05/EB/0005EBB7
2008/07/23 23:46:51| Open FD 103 /var/squid/05/EB/0005EBB7
and last .... shutting down and restarting squid will really close all FDs and
bring things back to normal. despite of what seems to be a bug, squid shuts
cleanly.
2008/07/23 23:46:51| Open FD 159 /var/squid/02/4D/00024D8B
2008/07/23 23:46:51| Open FD 223 /var/squid/07/12/000712CA
2008/07/23 23:46:51| Open FD 288 /var/squid/02/9B/00029B57
2008/07/23 23:46:51| Squid Cache (Version 3.0.STABLE8): Exiting normally.
and those lost of 'Open FD' entries really
prove that file descriptors were being lost by squid.
can you please check the bugzilla entry and
try to reproduce some of the steps took during
that time ? Watching filedescriptors with
cachemgr.cgi and lsof as well as during squid
shutdown can give you a hint of what's really happening.
the bugzilla entry for the bug supposely fixed on 3.0 STABLE11
http://www.squid-cache.org/bugs/show_bug.cgi?id=2424
i cannot confirm the bug still exists
because i stop using 3.0 and go to 2.7 during that bug time
--
Atenciosamente / Sincerily,
Leonardo Rodrigues
Solutti Tecnologia
http://www.solutti.com.br
Minha armadilha de SPAM, NÃO mandem email
gertrudes@xxxxxxxxxxxxxx
My SPAMTRAP, do not email it
