[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: mimedefang on alpha

On Fri, Jul 30, 2004 at 04:46:13PM +0200, Thimo Neubauer wrote:
> On Fri, Jul 30, 2004 at 12:54:24PM +0100, Tim Cutts wrote:
> > No-one replied before, but we are willing to host a build machine for 
> > Alpha here.  We have several machines we could use - we have several 
> > sizeable machines approaching the end of their maintenance contracts 
> > which we will not want to use for production work once that time 
> > arrives.  I don't foresee much difficulty putting one or two of them in 
> > our DMZ for Debian's use.

> First of all thanks for your generous offer but I think that the
> problems will be resolved soon. I kept quiet because I knew that the
> machine HP offered some time ago already was on it's way. Now that HP
> put time and money into sending me a machine I'd definitely want this
> box to become a Debian server. If I should get big problems with the
> thing I'll tell you privately in the next couple of days. But the
> first steps with the new box (sitting right next to me) seem good: the
> SRM already showed. :)

> As I know them the Debian Admins are not very keen on changing the
> server situation rapidly and I guess they mostly have reasons for
> that. Technically speaking, two Alphas are enough. For the longest
> time lully could build the packages for unstable without a big queue
> and escher was mostly idling if not for libc tests once in a
> while. With the new server building will be no problem at all and
> escher will still be a stable development machine. That's my view
> anyway, feel free to put me wrong :)

> I'll work on getting the new box up and running as fast as RL permits

Having only one buildd per arch, or having our only backup buildd be
unable to handle the load on its own in the event of a failure of the
primary, is a big problem, IMHO.  Escher has been having a terrible time
trying to keep up with uploads.  The reality is that failures like
lully's can take weeks to recover from, which can put a serious dent in
release planning if there's no real redundancy in place.

If Brainfood isn't planning to resurrect lully, I think it would indeed
be appropriate for the project to solicit another alpha buildd beyond
escher and the new box.

Steve Langasek
postmodern programmer

Attachment: signature.asc
Description: Digital signature

Reply to: