[Mailman-Users] Mailman on Solaris-based web server

[Richard Barrett]

Just my last comment.

On 22 Oct 2004, at 01:57, Brad Knowles wrote:

> At 12:17 AM +0100 2004-10-22, Richard Barrett wrote:
>
>>  1. handling of files containing messages queued for processing as 
>> they
>>  are moved along the chain of queues from initial delivery by the MTA
>>  to handoff to the outbound MTA and to the archiver. The code in
>>  $prefix/Mailman/Queue/Switchboard.py provides the enqueueing and
>>  dequeueing functions.
>
> 	That's fine once the message gets to Mailman.  This means that your 
> MTA queues (and probably your user mailboxes, unless you use an 
> NFS-friendly storage method for them as well) should be on local 
> filesystems, however.  If the server goes down, you risk losing those 
> messages which are in the MTA queue but haven't been delivered yet.
>

It turns out that the period of risk for incoming messages to my 
Mailman server config which uses NFS is quite small. The local MTA does 
not send its SMTP response accepting delivery to the external MTA until 
it has successfully run the Mailman delivery script. The Mailman 
delivery script writes the incoming messages to the Mailman qfiles 
directory on NFS before returning its response to the local MTA. Thus 
the window of risk for unrecoverable loss of incoming messages is quite 
small, being just what is in unflushed I/O buffers at the time of a 
server failure. If the outgoing MTA used by Mailman is not running on 
the Mailman server then the risk of loss for outbound traffic due to a 
Mailman server failure is effectively nil.

As for user mailboxes, our Mailman servers (primary and backup) do not 
support anything but the list aliases plus a minimum set of 
RFC-recommended other aliases. The normal user mailboxes for the domain 
are held by a different system which provides the IMAP and POP 
services.

<snip>