On 7/11/19 4:13 AM, Stephen John Smoogen wrote:
On Thu, 11 Jul 2019 at 06:47, Philip Kovacs via devel <
devel(a)lists.fedoraproject.org> wrote:
> It's likely the big endian emulation running on little endian machines
> which is killing performance.
> I also have some time sensitive package tests failing on s390x.
>
I am a bit confused by what you are saying. Could you restate? Compilation
on s390x happens on native hardware. But are you meaning that the erlang VM
is little endian and the big-endian s390x is having to emulate it?
We have been having problems with the s390x systems lately due to multiple
reasons outside of infrastructure's control. Kevin Fenzi has been working
on trying to get a replacement set of builders in place over the last week
to try and deal with this. I do not know if your build got stuck on a
builder which was transitioning or something similar.
So, yeah, part of this is my fault. :(
If you look at
https://koji.fedoraproject.org/koji/taskinfo?taskID=36131065
you will note that there are 4 buildroots listed. That means the build
was restarted 4 times. Likely this happened when I was
installing/updating/shuffling builders around there. Sorry about that.
That said, it still should have finished long ago.
The current state of s390x builders:
buildvm-s390x-01 to 14 are all the existing Z/VM instances we have had
for a long while. Likely we are going to keep them for a while until we
are sure the kvm instances are reliable and happy.
buildvm-s390x-15 to 24 are new kvm instances. They have a higher
'weight' than the others and from my testing are much faster.
Your build seems perhaps stuck in tests?
I've freed it (again) and it's now running on buildvm-s390x-22 (a kvm
instance). Lets see how it does there. I will in the meantime
update/reboot the Z/VM instances.
kevin