PMDF System Manager's Guide

Mail messages in the PMDF message queue directories generally have a two-digit file extension. If PMDF detects a message that is looping, it will rename the file so that it has an extension of .HELD . Message files with such a file extension will not be processed by PMDF channel programs and therefore will not be delivered. This is a safety mechanism to prevent messages from looping indefinitely. Looping messages are detected by having a large number of Received: headers lines.

34.4.7.1 Diagnosing .HELD Files

Another common cause of message loops is PMDF receiving a message that was addressed to your host with a network name that PMDF does not recognize as one of the host's own names. For example, imagine a host which is known to the TCP/IP domain name system and to other hosts and users as example.com, but whose PMDF configuration does not know that. A message is sent to joe@example.com and is accepted by the network and delivered to this host. Since PMDF does not know itself as example.com, it will likely assume that example.com is elsewhere and direct the message back out to the network and unwittingly loop the message back to itself. This loop will continue until PMDF detects the loop and puts the message on hold.

If you detect such a situation you should try to determine by examination of the message file whether there is a name you should add to your PMDF configuration as a synonym for your official local host name. The Received: lines should show the path the message travels through the loop.

Example 34-2 pmdf.cnf For milan.example.com

! pmdf.cnf - PMDF configuration file for milan.example.com ! Written by SYSTEM, 19-AUG-2012 21:23 ! This file was created by the PMDF configuration generator. ! ! Rewrite rules for the local host/cluster ! milan $u@milan.example.com (1) milan.example.com $u@milan.example.com naples $u@milan.example.com (2) naples.example.com $u@milan.example.com example.com $u@milan.example.com (3) ! ! Rewrite rules for the Internet . . . l nox_env_to milan.example.com (4) . . .

1. Rewrite rules for first cluster member to official local host.
2. Rewrite rules for second cluster member to official local host.
3. We have added this rule so example.com is recognized.
4. The host name on the L channel (local channel) is always the official local host.

In Example 34-2 we have added example.com as another name for the cluster consisting of milan.example.com and naples.example.com, where the official local host name has been milan.example.com. Mail addressed to joe@example.com will now be properly recognized and locally delivered.

If you do not believe that the name in question should be directed to your host, then you can have to address the problem with a network configuration change or by changing the behavior of a remote mailer.

34.4.7.2 Cleaning Up .HELD Files

# cd /pmdf/queue/tcp_local # foreach N ({.,*}/*.HELD) ? mv $N `dirname $N`/`basename $N \.HELD`.00 ? end

Then synchronize the queue cache with the command

# pmdf cache -synchronize

(Alternatively, the pmdf qm -maint utility's release command can be used to cause message files to cease to be .HELD and the PMDF queue cache database to be synchronized.) Then use the pmdf startup post to run the PMDF periodic delivery retry job immediately. Now, after the resulting jobs have run, look around to see if there are new .HELD files. There can very well be some. If there are still .HELD files in the original queue directory, then you can not have solved the looping problem. However, you can also find .HELD files in other queue directories such as the local channel (L channel) queue directory. This is because PMDF marks a message .HELD when it has too many Received: lines in which the local host appears. As the message moved from the original directory to another directory (i.e., moved from one channel to another), PMDF again saw too many Received: lines in the message's header and again marked it .HELD . This is to be expected. Simply repeat the process of renaming the .HELD files to .00 , synchronizing the queue cache, and again submitting PMDF processing jobs. Repeat this process until there are no more .HELD files in any of the channel queue directories.