[Systems] First weekly meeting and other matters...

[Bernie Innocenti]

On Thu, 2010-02-18 at 15:27 +0100, Sascha Silbe wrote:
> On Thu, Feb 18, 2010 at 08:24:33AM -0300, Bernie Innocenti wrote:
> 
> > Making all primary services fully redundant is probably not an option,
> > because not all services are as "easy" to cluster as ASLO.
> Out of curiosity, what are the problematic ones (if we assume a properly 
> clustered PostgreSQL installation which e.g. Trac can use)?

Clustering the DB is the #1 problem, but not the hardest.

Most applications, including Trac and Mediawiki, also use filesystem
storage for attachments and media. This would have to be clustered too.

And once you have to manage a cluster instead of a single machine,
sysadmin complexity grows 10x.

This is why I'd rather have one solid machine that can guarantee between
99.9% and 99.99% of uptime, more than enough for our current scale. We
only have 3 part-time volunteer sysadmins here, let's leave the
five-nines reliability problems to Google and Amazon, who hire dozens
just for running the backups :-)

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