[Mimedefang] redundancy in mimedefang

Dave Ellenberger dave at nofuture.ch
Mon Oct 6 05:21:01 EDT 2003


On Mon, 6 Oct 2003 00:12:35 -0600 (MDT), Lucas Albers wrote
> Considering this situation:
> Two installations of mimedefang configured as incoming mx servers 
> for a mail domain, configured to relay mail to some internal mail server.
> M1 and M2 are their names.
> 
> Fault Tolerance:
> M1 is down, then all mail is routed through M2.
> Based on clients point to the next mail server on the mx record.
> If M1 dies anything on the ramdisk spool is lost.
> How to prevent this?
> How to prevent any mail from every being lost, even if 1 of the cluster
> dies?
> Set up seperate mail spools? A disk spool and ramdisk spool?
> Synchronize the ramdisk to disk?
> 
> Ramdisk is being used for performance requirements.

Hi Lucas,

Mails queued by sendmail's mta are usually located in /var/spool/mqueue on
disk, not in /var/spool/MIMEDefang where the ramdisk is located. The remote
MTA will resend the email if the mimedefang scanners go down while recieving.
The mail will be recieved delayed, but it will not be lost.

I suggest that you create a directory outside the ramdisk to store the data
you don't want to lose (if you have not yet). Example: /var/state/defang, and
changed the home of the defang user in /etc/passwd to this.
If you are using quarantine, set the var $QuarantineDir to where the
quarantine mails should be stored (outside the ramdisk...), Example:
$QuarantineDir = "/var/state/defang/quarantine/virus";

> Bayes Database:
> How to share the bayes database between the systems?

On my defang machine sa learns about 5'000 spams and more than 10'000 hams a
day. I see no need to share the database on busy defang servers. The suggested
bayes database size of max 10MB is reached with about 5'000-10'000 ham/spams.

Maybe if both servers really have low amounts of mails to learn rsync or
intermezo could do it. The real easy way would be NFS, but the redundancy will
be lost again if the NFS server is not aviable. I suggest you just keep it
simple so it stays stable in redundancy and such.

-Dave




More information about the MIMEDefang mailing list