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

Re: saving sparc for etch requalification

[Note: all the information I list in that mail have been grabbed from
the web / irc / ssh access to the machine. It would be nice if that can
be confirmed by somebody]

On Thu, Oct 27, 2005 at 10:13:55AM +0200, Robert Lemmen wrote:
> as far as i see it (which isn't very far) auric is having problems, vore
> is considered to be too weak to count as a full buildd, and some new
> buildd is in the works. the new buildd is probably able to keep up by
> itself, so the only problem is the redundancy. judging from what
> db.d.o says vore isn't very fast, but it *should* be able to keep up.
> perhaps vore could be beefed up with a bit more memory or more/faster
> spindles? what hardware exactly is vore?

AFAIK, with the information I grabbed from #debian-release, the main
problem with the sparc port is that the kernel is not stable, thus the
buildd has stability problem. This is the case of vore, auric and

vore is an UltraSPARC II 300Mhz, with 512 MB of RAM, so I doubt that we
can upgrade a part to make it faster, except the processor. It is not
able to follow unstable, I would say it could build 80/90% of the

auric is able to follow unstable without problem.

I don't know a lot about mrpurply, but looking at the logs, it seems it
is the same kind of hardware (maybe a bit slower) than vore.

So if the 3 machines are able to run 24x7, we have buildd redundancy:
 auric itself is able to follow unstable
 vore + mrpurply are able to follow unstable

The other point of the release team, is that vore and auric are also
developer machines. If somebody is using the machine, that would slow
down the buildd. I think we could remove developer access to vore, and keep
only developer access to auric. It has 2 CPUs, one only being used by
the buildd.

So that would say the only problem is to make the buildd running 24x7. I
have seen that vore and auric are running 2.6.12 kernels. Does the
instability comes from there? AFAIK there is no stability problem with
the kernels from sarge. Maybe the machine could use them.


  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Reply to: