The mailing list for listmasters using Sympa

[Miles Fidelman <address@concealed>]

Warren Anderson wrote:
> Does anyone out there require high-availability for their sympa lists? If so, 
> how do you provide redundancy? Having a back up service that we can fail-over 
> to would be the obvious thing to do, but then we have the problem of 
> synchronizing the production service and the backup. We are using LDAP to 
> provision subscription lists and moderator lists, and are using kerberos for 
> wwwsympa authentication, so we don't really need to replicate user 
> information for the lists. However, we would want to somehow sync the 
> archives if possible, and I think there is information in the DB back-end (in 
> our case mySQL) that would need to be synched too.
>
> We can probably figure this out ourselves over time, but I would prefer to 
> learn from the experience of others if  anyone has some.
>    
We take a pretty simple, brute force approach:

- we run Sympa (and its associated files, databases, etc.) in a Xen 
virtual machine
- we run the VM on a 2-machine cluster set up for failover (pacemaker, 
heartbeat, DRBD based)
- for good measure, the disks on each machine are RAIDed (and we do 
nightly backups as well)

It all just works.  If one machine crashes, or otherwise fails, the data 
is already replicated on the other machine (courtesy of DRBD), and the 
VM just starts up on the 2nd machine and picks up where it left off 
(courtesy of pacemaker and heartbeat).

Now, if you want off-site failover, that gets a little trickier - though 
the latest versions of DRBD will replicate across disks that are located 
at different physical sites.

Miles Fidelman

--
In theory, there is no difference between theory and practice.
In<fnord>  practice, there is.   .... Yogi Berra