Re: scheduler core dumps ..

[Matti Aarnio <mea@nic.funet.fi>]

> Hello:
> 
> 	Has anyone been running into scheduler problems ..  Unexpectedly
> dying off ..  Also, I sent scheduler SIGUSR1 and it dumped core and
> gave me the following error ...
> 
> scheduler: reread configuration file: /local/share/mail/scheduler.conf
> memory corruption error detected, file malloc.c, line 291: corrupt
> block passed 
> to free()

	Uh, this was "easy" to locate and to fix.
	It was all in the  scheduler/readconfig.c:rereadconfig()
	routine, but your original problem is another matter.

	As I recall, the surprise deaths leave no cores ?
	I hope you have  "ulimit core unlimited" (or what your
	shell has..) so that it would not block core-dropping
	into  $POSTOFFICE/scheduler/ -directory.

	If you can tolerate it,  scheduler can be run with a verbose
	logging active on a debugger.
	Place breaks on   "exit()", "_exit()", and "abort()".
	Then start the  scheduler (on the debugger) with arg: -v

	Be prepared that it outputs A LOT...

	I have been "having fun" at revising a (finnish) user-guide
	to university UNIX systems, bare 200 pages of LaTeX...
	Time for feeding it to the offset is on Monday, so I might
	not yield a public dump of the current Zmailer until then.

> Cheers,
> 
> Marco

	/Matti Aarnio	<mea@nic.funet.fi>

PS:	Have you ever wanted to freeze the Zmailer so that it would
	not accept start while some critical system component is out ? 
	Next dump will have  freeze/thaw on  proto/zmailer.sh script :)
	(I have a cron-job watching after servers, and when it notices
	 a lack of Zmailer, it does "/etc/zmailer" to start it..)