[comp.mail.sendmail] 554 System file missing

taylor@cs.glasgow.ac.uk (Jem Taylor) (12/11/88)

Can anyone tell me what the problem is which causes failures like these ?
We get one or two a day out of several hundreds of messages; the sender
is a 4.3BSD MIPS/1000, running as a Yellow Pages server, while the receiver
is a SUN 3 running SunOS 3.5, also a Yellow Pages server.

----- Begin Forwarded Message -----


|From MAILER-DAEMON Fri Dec  2 12:50:56 1988
Date: Fri, 2 Dec 88 12:50:52 GMT
Message-Id: <69.8812021250@crete.cs.glasgow.ac.uk>
Subject: Returned mail: System file missing
To: pp

   ----- Transcript of session follows -----
554 hawaii.ether... 554 System file missing
554 PAT@LEVA.ICF.LEEDS.AC.UK... System file missing
		
   ----- Unsent message follows -----
|From: Mr Paul Philbrow <pp>
Date: Fri, 2 Dec 88 11:55:37 GMT
Message-Id: <15731.8812021155@crete.cs.glasgow.ac.uk>
To: PAT@LEVA.ICF.LEEDS.AC.UK
Subject: ....

etc.	
	


----- End Forwarded Message -----


Confused -Jem Taylor.

Andrew.Findlay@brunel.ac.uk (Andrew Findlay) (12/15/88)

| We get one or two a day out of several hundreds of messages; the sender
| is a 4.3BSD MIPS/1000, running as a Yellow Pages server, while the receiver
| is a SUN 3 running SunOS 3.5, also a Yellow Pages server.

| Subject: Returned mail: System file missing
| 
| 554 hawaii.ether... 554 System file missing

Looking through the Sendmail source (5.52 with IDA in my case) I can
only see two cases where EX_OSFILE is referenced. One occurs if
sendmail.cf cannot be opened, and the other if the smtp service cannot
be found from services.byname.

Is it possible that your services database is sometimes faulty?

Andrew

---------------------------------------------------------------------
|  From Andrew Findlay at Brunel University, Uxbridge, UB8 3PH, UK  |
|  Andrew.Findlay@brunel.ac.uk          phone: +44 895 74000 x2512  |
---------------------------------------------------------------------

lear@NET.BIO.NET (Eliot Lear) (12/15/88)

Here are two possible answers:

If the site is running sendmail, it could mean that there is a problem
with opening their sendmail.cf, or that sendmail cannot get the
service number using getservbyname.

There can be more... search for EX_OSFILE in the source if you have
it..
-- 
Eliot Lear
[lear@net.bio.net]

lindberg@cs.chalmers.se (Gunnar Lindberg) (12/19/88)

In <9016.8812102223@oba.cs.glasgow.ac.uk> taylor@cs.glasgow.ac.uk
(Jem Taylor) asks about:

       ----- Transcript of session follows -----
    554 hawaii.ether... 554 System file missing
    554 PAT@LEVA.ICF.LEEDS.AC.UK... System file missing

on a 4.3BSD MIPS/1000 system run as an YP server.

Andrew.Findlay@brunel.ac.uk (Andrew Findlay) brought up:

    Looking through the Sendmail source (5.52 with IDA in my case) I can
    only see two cases where EX_OSFILE is referenced. One occurs if
    sendmail.cf cannot be opened, and the other if the smtp service
    cannot be found from services.byname.

Here's my 2 cents: We run a few VAXen with MtXinu 4.3BSD and one
annoying thing was "getservbyname()" that simply didn't make it,
sometimes...

We called MtXinu and there showed to be an error in "ndbm.c",
which made "ypserv" loose track of next entry in "services.byname"
(which is the only map that is scanned instead of being directly
accessed) *if* somebody else had called in the meantime (yes, a
real Heisenbug :-).

The error does not seem to exist in SunOS 3.* so you could possibly
test by using "/usr/etc/yp/ypset" to point to either Sun or MIPS as
server and then do several "ypcat services" in parallell and compare
the output.

	Gunnar Lindberg
	Dept of Comp Sc, Chalmers Univ of Tech

nims@casbah.acns.nwu.edu (Christopher Nims) (02/13/91)

We're running sendmail-5.65+IDA-1.3.5 on a uVax II running Mt.Xinu 4.3 BSD.
This machine is being used as a mailhost for many worstations on campus and
recently many users are getting messages bounced back with the following
info in the header:

...
451 makeconnection: server "smtp" unknown: Address already in use
554 casbah.acns.nwu.edu (tcpld)... 554 System file missing
554 <galanter@casbah.acns.nwu.edu>... System file missing
...

One problem is that we can't find a deterministic method of reproducing the
problem.  If a message bounces, quit often sending it a second time will
get it through.

One more piece of information...when running sendmail in address test mode,
no matter what sendmail.cf I try (I grabbed a couple from a couple other
machines just to test them) it gives the following information:

accuvax 128 # /usr/lib/sendmail -bt
ADDRESS TEST MODE
Enter <ruleset> <address>
[Note: No initial ruleset 3 call]
> 0 nims@acns.nwu.edu
rewrite: ruleset  0   input: nims @ acns . nwu . edu
rewrite: ruleset  3   input: nims @ acns . nwu . edu
rewrite: ruleset  8   input: nims @ acns . nwu . edu
rewrite: ruleset  8 returns: nims @ acns . nwu . edu
rewrite: ruleset  6   input: nims < @ acns . nwu . edu >
rewrite: ruleset  6 returns: nims < @ accuvax . nwu . edu >
rewrite: ruleset  3 returns: nims < @ accuvax . nwu . edu >
rewrite: ruleset  6   input: nims < @ accuvax . nwu . edu >
rewrite: ruleset  6 returns: nims < @ accuvax . nwu . edu >
rewrite: ruleset  0 returns: $# local $: nims
>

Even though it gives the Note about ruleset 3, it still seems to work.


Any advice would be helpfull.
Thanks.
===============================================================================
Christopher D. Nims 					Chris_Nims@nwu.edu
Distributes Systems Services
Academic Computing and Networking Services: 		Northwestern University

k2@bl.physik.tu-muenchen.de (Klaus Steinberger) (02/13/91)

nims@casbah.acns.nwu.edu (Christopher Nims) writes:

>We're running sendmail-5.65+IDA-1.3.5 on a uVax II running Mt.Xinu 4.3 BSD.
>This machine is being used as a mailhost for many worstations on campus and
>recently many users are getting messages bounced back with the following
>info in the header:

>...
>451 makeconnection: server "smtp" unknown: Address already in use
>554 casbah.acns.nwu.edu (tcpld)... 554 System file missing
>554 <galanter@casbah.acns.nwu.edu>... System file missing
>...

>One problem is that we can't find a deterministic method of reproducing the
>problem.  If a message bounces, quit often sending it a second time will
>get it through.


I've suggest you are running YP (NIS) on your machine?
I've seen this problem on my systems also. The problem comes from
ypserv, which can't serve simultaneous accesse correctly in
many implementations. We found this bug on our PCS/Cadmus
system as well as on a CDC box (really a MIPS ). Both systems
have now a patched ypserv installed, which clearly resolves the problem.

There is a simple way to determine if YP is the problem.

Write a simple program, which querys for a service in an endless loop,
and start it twice (in background). Here is my test program:
If YP is the problem it should dwarf about unknown service sometimes.


#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <netdb.h>
#include <stdio.h>

#define SERVICE	"time"

main(argc,argv)
int argc;
char *argv[];
{
	struct servent *sp;
	long	i;

	i = 0;
	for (;;i++)
	{
		sp = getservbyname(SERVICE, "tcp");
		if (sp==NULL) {
			fprintf(stderr, "%s: unknown service, loop %d\n",SERVICE,i);
		}
	}
}


Sincerely,
Klaus Steinberger
--
Klaus Steinberger               Beschleunigerlabor der TU und LMU Muenchen
Phone: (+49 89)3209 4287        Hochschulgelaende
FAX:   (+49 89)3209 4280        D-8046 Garching, Germany
BITNET: K2@DGABLG5P             Internet: k2@bl.physik.tu-muenchen.de