home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!decwrl!decwrl!infopiz!mccall!ipmdf-newsgate!list
- From: dan@hmcvax.claremont.edu (Daniel C. Newman)
- Newsgroups: vmsnet.mail.pmdf
- Subject: RE: Postmaster notification for held msgs
- Message-ID: <01GMX0EOW7AOAM2ZDZ@HMCVAX.CLAREMONT.EDU>
- Date: 28 Jul 92 23:19:00 GMT
- Organization: The Internet
- Lines: 78
- Return-Path: <epmdf@YMIR.CLAREMONT.EDU>
- Resent-Date: 28 Jul 1992 15:19 -0800 (PST)
- Resent-From: epmdf@YMIR.CLAREMONT.EDU
- Errors-To: epmdf@YMIR.CLAREMONT.EDU
- Resent-Message-ID: <01GMX0FK23ZM9GVWIR@YMIR.CLAREMONT.EDU>
- X-Vms-To: IN%"bob@camb.com"
- X-Vms-Cc: IPMDF
- Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
- Content-Transfer-Encoding: 7BIT
-
- > I had situation where pmdf held some msgs after detecting a probable mail
- > loop. I don't think I'd received any notification from pmdf (e.g., via
- > a mail msg to postmaster) when this occurred. Did I miss it? Should the
- > daily RETURN job report on held files if they stick around for too many
- > days?
-
- No, you did not miss seeing an error report. When a mail loop is detected,
- sending more mail is very, very dangerous. Consider the case where the local
- host is not recognizing one of its names -- the typical cause of a mail loop.
- Now, suppose we send mail to the postmaster and the postmaster's address
- involves the unrecognized name. The message to the postmaster will loop and
- loop until it is marked .HELD (20 loops) and yet another message will be sent
- to the postmaster. Mail then piles up and the postmaster still has no clue as
- to the problem.
-
- Now, it's not hard to conceive of a loop whereby the looping mail multiplies.
- In such a case generating new mail is really, really dangerous.
-
- The bottom line here is that when a mail loop occurs, something, somewhere is
- misconfigured (the local host, a remote host, who knows) and generating new
- mail messages (warnings to the postmaster) is clearly not a safe thing to do.
- PMDF therefore does nothing other than to rename the message file to *.HELD
- (this prevents PMDF from attempting further delivery attempts). At present,
- it's up to the postmaster to notice such things. About the only safe thing
- PMDF could do would be to issue an OPCOM broadcast of some sort.
-
- Now for some general advice on what to do when you find *.HELD messages. [I
- know that Bob knows what to do, but I figure other people on this list may
- not.]
-
- First, the occurrence of these loops are very rare and always indicate a
- misconfiguration somewhere. There are two typical causes:
-
- (1) A user has set up a forwarding loop (e.g., their mail on system A goes
- to system B; their mail on system B goes to system A), or
-
- (2) System A thinks that mail for user@host goes to system B; B receives
- the mail but doesn't doesn't recognize that it is @host and sends the
- mail back to A (or elsewhere) which then sends it back to B.
-
- The way to determine what is going on is to examine the Received: lines in a
- *.HELD message. PMDF, when it sees the local host occur 10 or more times
- (configurable with the PMDF option file options MAX_LOCAL_RECEIVED_LINES and
- MAX_RECEIVED_LINES) in the Received: header lines, will mark the message *.HELD
- and stop trying to deliver it. Look at these received lines, see who the
- message is for and make sure that (1) is not the problem. If (1) is the
- problem then undo the user's forwarding and discuss e-mail concepts with them.
- If (1) is not the problem, then it is undoubtedly (2).
-
- If (2) seems to be the case, then the problem is either (2a) that the local
- host does not recognize one of its names, or (2b) the remote host which keeps
- on sending mail to the local host is confused. In the former case, you
- probably need to add a rewrite rule to your PMDF.CNF of the form
-
- unknown-host $U@local-host
-
- where "unknown-host" is the host name that is not being recognized and
- "local-host" is the official local host name associated with the local (l)
- channel. It may be that the local host is a gateway and should be passing this
- mail on to another system, "other-host", in which case a rule of the form
-
- unknown-host $U@other-host
-
- is probably appropriate.
-
- In the case of (2b), you have to call the postmaster of the remote host and
- get them to fix things....
-
- Now, after things are straightened out, rename the *.HELD file to *.00, [run
- pmdf_root:[exe]synch_cache.exe if you're using PMDF V4.1,] and then release the
- PMDF Delivery job in the MAIL$BATCH queue. There's a good chance that the
- message will pass on to another PMDF channel queue directory (e.g., the local
- channel directory) and again be marked *.HELD. This is normal; it happens
- because PMDF again sees the local host name appearing 10+ times in the received
- line. Merely rename the file to *.00, [rerun synch_cache,] and then release
- the PMDF Delivery job yet again.
-
- Dan
-