home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.mail.sendmail
- Path: sparky!uunet!pipex!warwick!cam-cl!doc.ic.ac.uk!rhbnc!csqx.cs.rhbnc.ac.uk!markd
- From: markd@csqx.cs.rhbnc.ac.uk (Mark Damerell)
- Subject: Configure sendmail, help please
- Message-ID: <1992Sep10.135436.9592@csqx.cs.rhbnc.ac.uk>
- Sender: news@csqx.cs.rhbnc.ac.uk (USENET News System)
- Nntp-Posting-Host: csqx.cs.rhbnc.ac.uk
- Reply-To: markd@csqx.cs.rhbnc.ac.uk (Mark Damerell)
- Organization: Dept of Comp Sci, Royal Holloway & Bedford New College Uni London
- Date: Thu, 10 Sep 1992 13:54:36 GMT
- Lines: 47
-
- Trying to configure sendmail, made a mess of it, please can anybody advise?
-
- 1. First problem is that any user might log in to any machine. Therefore it
- seems that the system mailboxes for incoming mail must all be in the same
- place. So I named one machine as the mailhost (call it HUB) & arranged for
- all incoming mail to appear on HUB in files called
-
- /usr/spool/mail/username
-
- Then all the other machines use NFS to mount the /usr/spool/mail directory.
-
- 2. The next problem is that on the satellite machines (say S1, S2...) NFS
- does not allow root to have privileged access to a mounted directory. So it
- seems that when sendmail on any satellite receives any piece of mail, it
- must explicitly use the tcp mailer to send that mail to the hub. So for
- each of the satellites, I wrote a .cf file whose Ruleset 0 says:
-
- R$+ $#tcp$@mailhost$:$1
-
- Sometimes this seems to work, at other times it goes badly wrong. The
- symptom is that large numbers of sendmail processes appear on satellite
- machines. After a while you start getting console messages about "too many
- hops". These processes cannot be killed by any ordinary method, a kill
- command gets a "no such process" error. What seems to be happening is that
- each sendmail process spawns a child, then the parent dies. If you do
- ps ax to find the PID of one of these runaways, then by the time you
- have read the PID and typed a kill, that process has killed itself.
-
- I dont know what is starting this off, but it seems likely that the
- sequence of events is something like this:
-
- User on S1 sends a message to somebody
- Sendmail reads the Ruleset 0, tries to pass the message to mailhost.
- Mailhost rejects it (or maybe is down).
- Sendmail tries to post an error.
- Sendmail reads Ruleset 0, tries to send error to mailhost.
-
-
- but after a while, the messages seem to be MAILER_DAEMON at S1
- sending to MAILER_DAEMON at S1 , so I suspect that the
- sequence of events suggested above is not the whole story.
-
- So please: what is the right way to configure sendmail in the situation of
- (1) above? and how to arrange a reliable route for error messages if the
- mailhost is down? THanks, MArk
-
-
-