[Systems] munin.sl.org updated to version 2.0-beta4-1

[Bernie Innocenti]

On Wed, 2011-09-28 at 15:27 +0000, Aleksey Lim wrote:
> > It does not seem to be used yet:
> > 
> >  http://munin.sugarlabs.org/sugarlabs.org/sunjammer.sugarlabs.org/memcached_rates.html
> 
> I pointed sunjammer's node to aslo-web's memcache (as it was before my
> investigation).

Ah, then we have a dependency inversion between sunjammer and aslo-web.

Sunjammer is acting 3 roles: load balancing frontend, aslo node and
database backend. Thus, if sunjammer goes down, the entire service goes
down. If instead aslo-web goes down, we halve our capacity (*).

Given the current architecture, it makes sense to run memcached on
sunjammer as well. If ASLO supports failover with multiple memcache
instances, we could keep both running. For the current amount of
traffic, I don't think we need to complicate our architecture beyond
this point in order to eliminate all single points of failure or to
improve the maximum throughput.

(*) actually, because there's no HA monitoring in place, we just keep
forwarding half of the incoming requests to a dead node.

-- 
Bernie Innocenti
Sugar Labs Infrastructure Team
http://wiki.sugarlabs.org/go/Infrastructure_Team