Looks like a new bug in clients ...
Carter Bullard
carter at qosient.com
Fri Aug 17 17:10:08 EDT 2007
Argus does a stat on the output filename, periodically, to see if it
needs
to close the moved file's file descriptor, and create a new one. I
think
your internal clock is getting screwed and so 'check filename' timestamp
isn't getting exceeded, so it doesn't check and keeps writing out to the
old file. The file is growing somewhere in the filesystem, but the
link to
it is gone, so, ....
I'm working on the leak, but the timestamp issue is still something
that could be a mystery still.
Carter
On Aug 17, 2007, at 3:42 PM, Peter Van Epp wrote:
> The latest clients.rc.48 seems to have broken log file switching
> somehow. The archived log file looks to end early (this should have
> been the
> noon switch):
>
> 07-08-17 11:40:58 e tcp 142.58.61.205.1971 -
> > 209.85.171.166.80 2 3 1034
> 1677 CON
> 07-08-17 11:43:12 eIs tcp 151.21.221.110.64696 -
> > 206.12.16.87.41417 2 0
> 134 0 REQ
> 07-08-17 11:43:12 e tcp 192.75.243.138.4645 -
> > 209.85.167.99.80 2 2 1125
> 443 CON
>
> and it didn't actually create a new com_argus when it did the
> switch (which
> it thinks it did at noon from the archive log):
>
> + gzip com_argus.2007.08.17.11.00.00.0
> + TIMESTAMP=2007.08.17.12.00.54
> 2007.08.17.12.00.54 compression completed
> + mv com_argus.2007.08.17.11.00.00.0.gz /archive/argus3/
> com_argus.archive/2007/08/17
> 2007.08.17.12.00.54 argusarchive exiting successfully
>
> But an lsof of the ra task indicates no output file present ... Ah
> illumination just struck as I was typing this. It is trying to use the
> file that was moved and compressed instead of creating a new
> com_argus:
>
> stat(/Volumes/Teradactyl/argus/argus/com_argus.
> 2007.08.17.11.00.00.0): No such file or directory
>
> this is now /Teradactyl/argus/argus/com_argus.
> 2007.08.17.11.00.00.0.gz
> the archived file, ra hasn't recognized it needs to recreate
> /var/log/argus/com_argus for some reason and is trying to write to
> the moved
> archive file.
> This may be a threads issue, since a test just now caused the file
> to be recreated correctly when I moved it. I'll recompile clients
> without
> threads and see if that helps.
>
> lsof | grep ra3
> ra3 1388 vanepp cwd VDIR 14,10 2856
> 122 /Volumes/Teradactyl/argus/argus
> ra3 1388 vanepp txt VREG 14,2 1289728
> 1159587 /usr/local/bin/ra3
> ra3 1388 vanepp txt VREG 14,2 1797576
> 1315128 /usr/lib/dyld
> ra3 1388 vanepp txt VREG 14,2 4402196
> 1636735 /usr/lib/libSystem.B.dylib
> ra3 1388 vanepp 0r VCHR 3,2 0t0
> 127907460 /dev/null
> ra3 1388 vanepp 1w VREG 14,10 649543026
> 50850 /Volumes/Teradactyl/argus/argus.logs/com_ra3.log
> ra3 1388 vanepp 2w VREG 14,10 649543026
> 50850 /Volumes/Teradactyl/argus/argus.logs/com_ra3.log
> ra3 1388 vanepp 3r PSXSHM 0x07c29104
> 4096 obj=0x07dd17f8
> ra3 1388 vanepp 4u IPv4 0x0829ea94 0t0
> TCP test4.ucs.sfu.ca:51828->hcids.ucs.sfu.ca:rmonitor (ESTABLISHED)
> ra3 1388 vanepp 5u
> VREG stat(/Volumes/Teradactyl/argus/
> argus/com_argus.2007.08.17.11.00.00.0): No such file or directory
>
> Peter Van Epp / Operations and Technical Support
> Simon Fraser University, Burnaby, B.C. Canada
>
More information about the argus
mailing list