The mailing list for listmasters using Sympa

[Sylvain Amrani <address@concealed>]

Johann Spies wrote:
> I have thought of doing something like this, but I would like some
> advice from this group on a better way of doing it:
>
> 1. Inform the mailing list owners of a period of possible unstability
>    for sympa.
> 2. Create an alias (in the Apache configuration) for wws/ on mail2
>    referring to sympa/.
> 3. Copy the ssl-certificate of mail3 to mail2 and restart apache on mail2.
> 4. Dump the postgresql-database on mail3.
> 5. Stop Sympa on mail3.
> 6. Import the database on mail2. (In future I would like to get some sort of
>    database replication going.)
> 7. Copy the content of /var/lib/sympa/expl from mail3 to mail2
> 8. Update /etc/mail/sympa.aliases on mail2 to reflect the situation of 
> sympa on mail3.
> 9. Stop the network on mail3
> 10. Create a virtual IP on mail2 with mail3's IP address.
> 11. Restart sympa on mail2 and see what is malfunctioning...
>   
Hi,

I did the same few months ago without any trouble. We have now 2 servers
in a heartbeat configuration, with a virtual IP address shared by both.
Now I run sympa upgrades on the spare and then swap the active server to
upgrade the second one.

The spools, configuration and expl/ are sync'ed from master to spare and
I will put these files on a DRBD volume in a near future. The spare act
as slave in a MySQL replication.

For the ease of administration, the spare will never become a master,
just an active server. During this time, users can't administrate their
lists (they get an apologies page as they try to authenticate via SSO),
and there is no task_manager daemon on this server. This way, I don't
have to re-sync expl/ on failover takeback.

These difficulties will disappear as soon as DRBD is configured. Of
course you can use a SAN or an NFS share. Anyone using that on this list ?

Regards,
Sylvain.

-- 
Sylvain Amrani
Direction Générale de la Gendarmerie Nationale
DGGN/SPM/SDTI/BSA
Section Architecture des Systèmes
01.56.28.86.47