Re: .d.o machines which are down (Re: Questions for the DPL candidates)

[Thomas Bushnell BSG <tb@becket.net>]

Bernd Eckenfels <ecki@lina.inka.de> writes:

> In article <[🔎] 87ll8nx8o9.fsf@becket.becket.net> you wrote:
> > getting things back.  The point of the N+1 rule, as I understand it,
> > is to give a different kind of redundancy, so that we don't have to
> > wait a day or two.
> 
> How many current debian services are hosted that way?

You mean serious services, that if they are broken make everything
else wait?  Only the ftp archive, which is mirrored.

Why should this one get a rule that, say, the BTS or the mailing lists
do not?  Because in my history of being involved with Debian, which is
something like eight years now, I have never seen the BTS be gone for
a day; I have never seen the mailing lists stop working for a day.
There may have been one or two that didn't hit my radar screen.

By contrast, I have seen an arch's autobuilders all be gone for a
day--or much longer.

So experience tells us that redundancy in the autobuilders is more
important, because they have historically been less reliable as a
class.

Thomas