[Mailman-Users] Keeping subscription info in sync across servers
R.Barrett at ftel.co.uk
Mon Feb 10 18:58:45 CET 2003
At 16:41 10/02/2003, Ian Chilton wrote:
>Please cc replies to ian at ichilton.co.uk.
>I was wondering if there was an easy way to keep subscription info in
>sync across multiple list servers.
>What I want to do is setup a backup list server which will still
>deliver list mail if the main box is down. So, I only need a single way
>replication, i.e all changes are made on the main server and the 2nd
>server just has to stay in sync.
>I was thinking an easy way to do this would be to have a function in
>mailman which made it automatically sent an e-mail or made a socket
>connection to the other server when any options or subscriptions were
>Is such a thing possible?
Just an idea of how we do things but it may not suit your needs/equipment
Faced with the problem of mailing list server resilience and redundancy we
turned to NFS and keep almost all of MM's $prefix subtree (all except the
stuff in $exec-prefix) on shared storage. I do have the advantage of a high
uptime file server, with all the data security features to act as NFS
server for this.
I run a live server with a backup server as hot standby, that is, the
secondary sits there until the primary server breaks and we initiate the
switchover. This involves a minimal manual intervention.
Basically we switch the MAC address of the machines known as
primary-mailman-server.our.domain and secondary-mailman-server.our.domain
in the DHCP config. A brief pause, shutdown the old primary and reboot the
erstwhile secondary which now comes up with the primary IP number. A little
conditionality in some of the start scripts means that coming up with the
primary server IP's number is all that is needed for the new primary server
to run as such.
A slight pause for the switchover period is OK for us; incoming mail just
backs up on the MTA's trying to deliver to the Mailman server until the
It is a bit crude but it works; had the internal hard drive on the primary
die a few weeks back and switched over with less than 15 minutes
interruption in service and no lost mail.
The secondary server does not lie fallow. It has its own MX record and I
can run test lists on it. I use it (not mounting NFS but using its local
hard drive for MM's $prefix tree) for checking out new MM software releases
and getting the wrinkles out before upgrading the primary server in the NFS
space; the stress when doing a major upgrade to the primary server is much
reduced having already done it for real as a rehearsal on the secondary
machine first. The secondary is also used for other non-critical tasks that
can be suspended if we have to perform a switchover.
If you have other ideas for providing low-cost resilience and redundancy
for a Mailman server then I would like to hear them.
More information about the Mailman-Users