version 1.1.20 of the DCC

Daniel V Klein dvk@lonewolf.com
Fri Jan 24 17:38:42 UTC 2003


Well, having read this thread I figured I'd look at my newly upgraded 1.1.20
dccm and dccd.  Here's what I see, running BSDI Unix and Sendmail 8.12.4. 
The 137:32 of CPU time is since 01:30, so although it hasn't been using
the whole CPU the whole time, it's been trying to :-)

Sorry, but I'm backing out to 1.1.16 again - but LMK if there is anything you
need from my logs!

Load averages:  1.51,  1.74,  1.75                                     12:24:33
82 processes:  4 running, 76 sleeping, 1 stopped, 1 zombie
CPU states:  8.6% user, 78.6% nice, 12.8% system,  0.0% interrupt,  0.0% idle
Memory: Real: 138M/168M Virt: 250M/682M Free: 46M

  PID USERNAME PRI NICE  SIZE   RES STATE   TIME   WCPU    CPU COMMAND
 4744 dcc       97    4 2632K 4008K run/0 137:32 91.50% 91.50% dccm
29545 www        2    0 4224K 2524K sleep   0:02  0.05%  0.05% httpd
11297 www        2    0 3824K 2516K sleep   0:01  0.05%  0.05% httpd
17595 www       18    0 1140K 1276K sleep   0:00  0.00%  0.00% tcsh
  359 root      18  -12  336K 1184K sleep   8:00  0.00%  0.00% ntpd
   20 root      10    0  752K 1556K sleep   0:57  0.00%  0.00% mount_mfs
    1 root      10    0  136K  980K sleep   0:20  0.00%  0.00% init
  362 root      10    0   72K  912K sleep   0:00  0.00%  0.00% <nfsiod>

> Vernon Schryver wrote:
> >> From: "Sven" <sven@dmv.com>
> >
> >>> Version 1.1.20 of the DCC source is in
> >
> >>     ... As of this morning, I noticed [from the 'top' command] that
> >> the dccm was using 99% CPU utilization (luckily these are multi-proc
> >> machines) and error messages were appearing in the log files ranging
> >> from: milter_read(dcc): cmd read returned 0, expecting 5
> >> to
> >> Milter (dcc): error connecting to filter: Connection refused by
> >> /var/run/dcc/dccm
> >> to
> >> Milter (dcc): timeout before data read
> >> Milter (dcc): to error state
> >> ...
> >
> > Those messages look like parts of sendmail log messages complaining
> > about a milter filter not responding.
> 
> I went ahead and kill -9 the processes and restarted (start-dccm). Within a
> couple minutes same message.
> 
> > Unless a dccm thread went into an infinite loop, I would expect this
> > version to use fewer CPU cycles, because HTML tags are not being
> > shoved through MD5.
> 
> Which would be great! Version 1.17 dccm uses anywhere from 0.10% to 2.5% CPU
> depending on the particular machine. Version 1.20 uses 95-99% .....
> 
> > Were there any system log messages from dccm?
> 
> I have dccm logging to local2 but no error messages. Only the aforementioned
> error messages in maillog
> 
> > Is it possible to find a sample of a message that makes dccm very
> > slow, perhaps by looking at the sendmail queue IDs that should be
> > part of
> > those syslog Milter complaints?
> 
> That might be possible, but keep in mind that this is happening on all 6
> machines within the same time frame. I will see if I can trap one when I
> revert one machine back to 1.20 again.
> 
> Sven
> 
> 
> _______________________________________________
> DCC mailing list      DCC@rhyolite.com
> http://www.rhyolite.com/mailman/listinfo/dcc



More information about the DCC mailing list

Contact vjs@rhyolite.com by mail or use the form.