Re: DSA concerns for jessie architectures (mips*)

On Sat, 22 Jun 2013, Andreas Barth wrote:

> > * mips: existing machines are either not reliable or too slow to keep
> >   up; we suspect that they may not be easily replaceable.

> Also, if we buy more mipsel machines we could convert the mipsel
> swarms to mips ones (and so replace broken machines, see below) -
> mostly depends on how urgent you think this is.

If our existing eight-year old hardware is the only mips machines we can
reasonably get then that doesn't bode well for mips.  We don't think
relying on the SWARMs (alone) is an option.

> > * mipsel: the porter machine and some of the buildd machines have an
> >   implementation error for one opcode; missing kernel in the archive
>
> Different answers - select the one you like most:
> 1. We could buy a some loongson 2f machines (or newer), see e.g.
> http://www.tekmote.nl/epages/61504599.sf/nl_NL/?ObjectPath=/Shops/61504599/Products/CFL-006
> plus some memory. These machines have kernels in the archive, and not
> the hardware bug with choking on too many nop-instructions in a row.

AIUI these machines have a maximum memory of only 1GB.  That's probably
OK for now but might be problematic in the long term.


> 3. We have currently two new machines with loongson 3a processors to
> test. It will take a bit of time to finally get a working kernel on
> these, but that would also decrease build-times quite much.

When do you expect them to be usable?

If not any time soon then maybe we should try to get a couple of
loongson 2f machines.  Would four machines of this type be sufficient to
replace all our exist swarm and 2e machines as buildds?  If so, should
we just get 5 (4buildd+1porterbox)?

Cheers,
-- 
                           |  .''`.       ** Debian **
      Peter Palfrader      | : :' :      The  universal
 http://www.palfrader.org/ | `. `'      Operating System
                           |   `-    http://www.debian.org/