[Systems] Migrating a.sl.o to beamrider.

[Bernie Innocenti]

[cc+=danny]

El Sat, 31-10-2009 a las 04:55 -0500, David Farning escribió:
> It looks like beamrider is ready to go into production.
> 
> I just cleared the memcached cache on sunjammer and turned on a 512M
> instances on beamrider.

Earlier this week, Danny installed a patch cable running between cloud9
(sunjammer's host) and treehouse (beamrider's host). The virtual
interface is not yet configured on sunjammer.

Once we have a fast and private communication channel between the two
VMs, we can do everything needed to cluster remora:

1) make our mysql available to both machines over TCP

2) mount /srv/www-sugarlabs/activities over NFS

3) (maybe) mount php sessions over nfs, or find a way
   to store them in memcached

4) (maybe) share a single memcached instance


Real fail-over would be ambitious:

 - We could setup slave MySQL instance on beamrider for improved
   performance and fault-tolerance for read-only operations

 - Instead of NFS, we could experiment with some of the fancy
   parallel networked filesystems.b Pohmelfs comes to mind.
   I'm currently evaluating DRBD and GFS2 for the FSF, but I'm
   not yet so confident about it.

 - We should not use memcached

For improved security, we may want to move the master MySQL instance to
a separate machine. Well, perhaps not very useful: both sunjammer and
beamrider store cleartext passwords allowing blanket access to the
various databases, so if they ever get rooted the database is also going
to be compromised.

We may still want to move MySQL away from the VMs for performance
reasons, but running non-trivial services right on treehouse sounds like
a bad strategy to me.

Comments?

-- 
   // Bernie Innocenti - http://codewiz.org/
 \X/  Sugar Labs       - http://sugarlabs.org/