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

Re: mips and mipsel qualification for Wheezy


regarding both mips and mipsel, a few remarks from the porters. Let's start
with our current buildd hardware:

1. swarm: can work as mips and mipsel. We have five such boxes, where one
is used as mips buildd (ball), two as mipsel (rem and mayer), one is
currently with Aurel, and one has a dead disk and hosting issues (mayr) and
should be brought back online again. Machines have dual-core CPUs. 

2. Loongson 2e/f: Only mipsel. We have five 2e machines, and one 2f. One 2e
is used as porter box, and one as buildd. A second one is just about being
setup as buildd. Machines have single-core CPUs, and therefor sometimes
"different" fast then the swarms.

3. Cavium Octeon: Only mips. We have three such machines, which are
installed equal. Two of them run buildds, one a porterbox (gabrielli). Two
of the machines have stability issues (not totally unstable and in general
works well enough, but not as it should be; means machines sometime just
reboot by themself), whereas one doesn't. Machines have 16-core-CPUs but no
FPU (that's why ressource-hungry packages and/or FPU-intensive ones build
*very* slow here). We currently run two buildds per machine, each using up
to six cores. We have been offered a replacement for the porterbox, which
is currently being stability-tested, and after succeeding could be

Regarding the lines on http://release.debian.org/wheezy/arch_qualify.html :

mips porterbox: We currently have gabrielli, but as noted above it
sometimes isn't as stable as it should. Also, usually issues are same
between mips and mipsel, so the mipsel porterbox could also often be
used. In sum I think red is not the appropriate colour; I'm not sure
green is appropriate for the current status but then yellow might be.

mipsel buildds: In the last month, we had two buildds eating their hard
disk, so all the time only three buildds are active. The three can just
keep up but are obviously not how it should be. The currently broken buildd
is the non-DSAed, and I don't intend to bring that one back again (at least
not for unstable), but instead setup another DSAed 2e buildd. If wanted, I
could also setup yet another one - hardware is available for that.

So, the topic "buildd-dsa" is already "yes" now as phrixos is down (and
please either decrement the number of buildds by one, or set the
"rendundancy" to yes - probably decrementing is more appropriate until the
new DSAed buildd is installed).

Of course, just as arm* both architectures might be not too fast in certain
cases like linking large C++-files - same issues and reasons apply for
mips* as well, so I don't want to repeat what's discussed already for arm*


Reply to: