SUMMARY: sendmail 8.8.3 error message (bug)

From: Marina Daniels (Marina.Daniels@ccd.tas.gov.au)
Date: Wed Nov 20 1996 - 23:29:00 CST


Thanks very much to:

rsk@itw.com (Rich Kulawiec)
Claus Assmann <ca@informatik.uni-kiel.de> at Internet
Iv n Villalobos Hern ndez <denp@acnet.net> at Internet

*** ORIGINAL QUESTION ****

Does anybody know what this means? It's a Solaris 2.5 system running sendmail
8.8.3
This is from /var/log/syslog

Nov 20 10:02:12 condor sendmail[17000]: proc_list_probe: lost pid 10116
Nov 20 10:02:12 condor sendmail[17000]: proc_list_probe: lost pid 10271
Nov 20 10:02:12 condor sendmail[17000]: proc_list_probe: lost pid 10479

*** ANSWERS ****

I forgot to mention sendmail was still running correctly even though reporting
this problem. I've installed it on 2 other solaris 2.5 machines but this is the
only one reporting this problem.

********

Yes. It's a bug in 8.8.3 I found this in comp.mail.sendmail yesterday --
haven't had a chance to install and test it yet.

>From CThomas@ucla.edu Tue Nov 19 13:50:35 EST 1996
>Article: 38158 of comp.mail.sendmail
>From: CThomas@ucla.edu (Chris Thomas)
>Newsgroups: comp.mail.sendmail
>Subject: Re: Problem with sendmail 8.8.3? (Lost child)
>Date: Tue, 19 Nov 1996 07:59:16 -0800
>Organization: UCLA Office of Academic Computing
>Lines: 27
>Message-ID: <CThomas-1911960759170001@ts30-12.wla.ts.ucla.edu>
>References: <CThomas-1811960735260001@csthome.oac.ucla.edu>
>NNTP-Posting-Host: ts30-12.wla.ts.ucla.edu
>
>In article <CThomas-1811960735260001@csthome.oac.ucla.edu>,
>CThomas@ucla.edu (Chris Thomas) wrote:
>
>> When trying to run 8.8.3, I get a large number of the following in syslog:
>>
>> Nov 18 07:21:29 lambda sendmail[15724]: NOQUEUE: SYSERR(root): SMTP-MAIL:
>> lost child: No child processes
>>
>
>I received the following (indirectly) from Eric.:
>
>: >My best guess is a couple of lines I put in to prevent this sort
>: >of problem if there had been an ETRN command. Take a look in
>: >srvrsmtp.c, routine runinchild, arround lines 1283 and again around
>: >line 1313. You'll see a couple of calls to setsignal() guarded
>: >by an #ifdef SIGCHLD. Try commenting out both of them.
>
>Commenting out both calls gets rid of the msgs. I am running the result
>successfully on one production machine. Note that this is not an official
>patch, or a "fix" for the problem - it merely bypasses it so you can try
>8.8.3.
>
>--
>Chris Thomas
>Manager, Systems Programming
>UCLA Office of Academic Computing
>CThomas@ucla.edu

*******

Seems to be related to this:

        Improvements to MaxDaemonChildren code. If you think you have too
                many children, probe the ones you have to verify that they
                are still around. Suggested by Jared Mauch of CICnet, Inc.
                Also, do this probe before growing the vector of children
                pids; this previously caused the vector to grow indefinitely
                due to a race condition. Problem reported by Kyle Jones of
                UUNET.

Thanks again!!
Marina



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:11:16 CDT