Dccm troubles

Vernon Schryver vjs@calcite.rhyolite.com
Fri Sep 5 16:12:01 UTC 2003


> From: Dale Hege <fhege@sover.net>

> I have upgraded to 1.2.4 and compiled both libmilter and dcc with -g. So
> we'll see the next time I get a core. :)

Does that mean that 1.2.4 has not failed?

> Attached are the minutes leading up to one of my failures from
> sendmails milter (no core) 
>
> Sep  4 18:47:48 walleye dccm[5739]: 1.2.2 rejected messages to 718 targets
> and discarded messages to 0 targets among 1197 total since 09/04/03
> 18:39:19

That line from dccm is not in the log entries you sent. 
Have you set /var/dcc/dcc_conf to separate dccm and sendmail log
entries, or perhaps MAIL.NOTICE from MAIL.ERR entries?

> Looks like it gets busy and doesn't aways catch new connections. 
>
> Any ideas on how to make milter a little more verbose to try and find why
> its unhappy?

libmilter calls syslog (named "smi_log") with LOG_ERR before it gives up.
Those log entries should go somewhere.

You mentioned BSD/OS.  At least 4.1 and 4.2 have major bugs with mmap()
files.  Worst is that if you stress the system by trying work with
more data than the system has physical memory, the kernel will crash.
About half the time it seems to take a dump and reboot, but the other
half requires a reset.  Not quite as bad is that if you mmap() 100
MBytes, unmmap it, and then close the file or any other combination
starting with mapping and ending with trying to stop including using
fsync() and msync(), the system will laborious copy all of the data
once or twice.  While copying, the system is extremely slow even if
you have tolerable ulta-SCSI disks and multiple controllers. 
Realsoonnow I've got to take time to replace BSD/OS 4.2 with FreeBSD.


Vernon Schryver    vjs@rhyolite.com



More information about the DCC mailing list

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