[packman] Tumbleweed uses Leap kernel

[Stefan Seyfried]

Hi Stefan,

Am 21.03.2017 um 22:17 schrieb Stefan Botter:
> On Mon, 20 Mar 2017 21:53:46 +0100
> Stefan Seyfried <stefan.seyfried at googlemail.com> wrote:
>> Stefan, do you have OBS_VM_KERNEL set in /etc/sysconfig/obs-server on
>> the workers?
>> That would override the kernel-obs-build stuff...
> 
> Yes, of course, otherwise the KVM stuff should not work according to
> the comments in /etc/sysconfig/obs-server...
> It seems to work without such definitons, though.

These comments may be old and outdated ;-)

> Since a few hours there is an additional worker there (swkjt02), which
> has 16 cores and runs 16 worker instances, but unfortunately only on 5
> spindles, which may harm the I/O performance a little bit.
> This is a fresh installed 42.2 with the obs-worker package from the 2.8
> tree.

It is more or less totally irrelevant which obs-worker package is
installed, because all it does is

 curl http://api:$port/give_me_the_workercode | tar xf

and then executes the downloaded code. So the only thing important is,
which build version is on the OBS server :-)

> It seems to work quite good so far. There are sometimes error
> messages regarding too few hugetable pages, and I do not know how to
> deal with it. Perhaps next week, if I can keep the machine that long.

Never had that at work, but my workers are relatively big machines
(144GB RAM, running the workers on tmpfs), and in my "research
installation" at home, I just do small testbuilds to check if it works.

> If I find time in the next few weeks, I will try to update all
> currently available workers to a more up-to-date OS

That's not really necessary, if kernel-obs-build is used then the host
kernel is not really that interesting to the build VMs anymore.

> and also update PMBS to 2.8.

I have not yet updated to 2.8, but the update from 2.6 to 2.7 was pretty
painless in my environments.

> But I cannot promise anything ATM - it is even possible,
> that I have to retire from PMBS hosting, but I will give enough heads
> up before that, if happening.
-- 
Stefan Seyfried

"For a successful technology, reality must take precedence over
 public relations, for nature cannot be fooled." -- Richard Feynman