[mod.computers.apollo] ApolloHelp: syslog error message

dennis%bugs@NOSC.ARPA (Dennis Cottel) (12/09/85)

Summary: need help with Domain/IX, Mail, and syslog

I've done the "sysadmin" installation of BSD4.2 on a disked DSP80 file
server, and then a "user" installation on my own DN320.  When I
executed Mail, I got two error messages, presumably from delivermail.
One of these pointed out that "/dev/../syslog" couldn't be written, so
I created one and got rid of that error message.  The other message is
"syslog: socket: IO error".  Although the mail is delivered correctly, I
have been unable to get rid of this message.  I have tried running
delivermail SUID to root, with all permission bits turned on on the
abovementioned syslog file, with no improvement.

I thought it might be because /etc/syslog wasn't running.  When I
execute /etc/syslog on my node, it returns immediately with an error
status of 72.  This means "an important system file is missing."
According to the syslog source on our VAX 4.2 BSD, this might be
because /etc/services doesn't contain a syslog entry, but /etc/services
has that entry, and exists on the file server (my node has a link to
/etc).  When I execute /etc/syslog after CRPing onto the server, it no
longer returns immediately--I take it to be running.  Nevertheless, I
still get the same message when executing Mail.  I haven't tried
rebooting the server with /etc/syslog in etc.rc.

Thanks for listening; can anyone explain how to fix the problem?

Dennis Cottel     Naval Ocean Systems Center     San Diego, CA 92152
(619) 225-2406        dennis@NOSC.ARPA        sdcsvax!noscvax!dennis

apollo@ucbvax.UUCP (12/10/85)

In-Reply-To: dennis%bugs@nosc.ARPA (Dennis Cottel), sun, 8 dec 85 23:00:14

    Summary: need help with Domain/IX, Mail, and syslog
    
Let me try to explain this situation:

First, the message "syslog: socket: IO error" means that when syslog(3)
tried to do a "socket" call, it failed.  (This is standard 4.2 behavior.)
This can happen if TCP service is not available (i.e. you don't have
a TCP server running on your node or you don't have a link to a node
that IS running one).

If syslog(3) fails in "socket", the standard 4.2 "syslog" then tries
to write its output to "/dev/console".  Since this doesn't seem
particularly desirable (even if we had a "/dev/console" -- which we don't
have right now -- it seems wrong to blat out "syslog" messages to it),
we made it try to create/open "/dev/../syslog" (which is
"`node_data/syslog" expressed in a slightly peculiar way).  I don't
know why it wasn't able to create the "syslog" file, except perhaps
because you have things protected strangely.

Note that if you have TCP installed, but aren't running syslog(8), logging
messages will simply be dropped on the floor.  (Again, this is standard
4.2 behavior.)

So, in short, if you don't have TCP installed, there's no way to suppress
the "IO error" message.

As for running "syslog" and its returning immediately, this is what it
appears to do.  However, it forks off a copy of itself and the parent
then exits so if you do a "ps", you should see a "syslog" process off
in the background.  (Once more -- standard 4.2 behavior.)

Anyway, I was the one who tested the whole sendmail/syslog mess for sr9
and it worked for me.  I know also that some customers are using it
successfully.  If you are still having problems, let me know.

                -- Nat Mishkin
                   {yale, wanginst, uw-beaver}!apollo!mishkin
-------