Eaelier, I wrote:
>Over the past week, our mail server has been regularly reporting
>the following error:
>
>Aug 4 06:08:28 eplrx7 sendmail[110]: NOQUEUE: SYSERR: daemon: cannot fork: No m ore processes
>Aug 4 06:08:30 eplrx7 vmunix: proc: table is full
>
>In addition to these errors, we get lot of these in /var/log/syslog:
>
>Aug 3 23:56:25 eplrx7 sendmail[3124]: AA02982: locked
>Aug 3 23:56:25 eplrx7 sendmail[3124]: AA03061: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA02902: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA02758: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA02928: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA02982: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA03061: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA03169: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA03247: locked
>Aug 4 00:56:25 eplrx7 sendmail[3375]: AA03374: locked
>
>We're unable to figure out what's doing this, but every evening, the
>system becomes unavailable for login and has to be rebooted every
>morning, due to the process table being full.
>
>Earlier in the week, a non-local mailing list (one member of
>which is here) was looping, feeding error messages back to itself.
>The mailing list owner has apparently fixed the problem, since
>my local user is no longer getting the error messages. Nonetheless,
>it would appear that someone's sendmail queue somewhere still has
>a batch of the old error messages which it's still trying to
>deliver at 56 minutes after the hour.
>
>I've added 'OX3' to sendmail.cf, in an effort to defend
>myself. This is supposed to refuse incoming smtp connections if the system
>load level is 3 or higher
>
>With it being Friday, I hate the prospect of continued system failures
>over the weekend, so if anyone has any ideas/suggestions on dealing
>with this, I'd appreciate hearing them.
>
Thanks to:
jeffw@triple-i.com (Jeff Wasilko)
Christopher A. Stewart" <allan@mazama.com>
tl@gam0.phy.anl.gov (Torben Lauritsen)
Alan Thew <Alan.Thew@liverpool.ac.uk>
All recommended treating the symptoms by upping the MAXUSERS parameter
in the kernel config and building a new kernel. This took care of the
problem. Presumably, this allowed the pig (all the queued error messages
from the bad mailing list) to get through the python. All seems well now.
-- Tim Evans | E.I. du Pont de Nemours & Co. tkevans@eplrx7.es.dupont.com | Experimental Station (302) 695-9353/8638 (FAX) | P.O. Box 80357 EVANSTK AT A1 AT ESVAX | Wilmington, Delaware 19880-0357
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:31 CDT