We needed to reupload the image with @nikromen because of a changed
hostname for the openstack instance. And I then forgot to note the
correct image name.
Nourhane @ IBM confirmed this should be OK, considering we spend about
the same cloud budget as before (more machines means we still process
the same amount of tasks, but faster).
These are about 1.5x more expensive, so we'll have only 2/3 of them for
now. The reason for this experiment is that the old instances cz2-2x4
are often useless for memory-intensive tasks that simply do too much
swapping - so we too much time on them (which is even often wasted for
unnecessary failure).
Relates: https://github.com/fedora-copr/copr/issues/3060
It was notoriously colliding with other system users like copr-signer
and others.
Revert "copr: test without nrpe_client_uid specified"
This reverts commit 435b71a695.
Everything should now be using linux-system-roles/network, so we drop
our hacky nmcli calls and everything that referred to them, including
exclude variables. Also, lets just let NM handle resolv.conf so it's not
wrong all the time on reboots.
Signed-off-by: Kevin Fenzi <kevin@scrye.com>
- add copr_builder_fedora_version variable, because production sometimes
differs from devel for a short period of time
- drop the epel-8 -> centos-stream+epel-8 links, this is now done all
against RHEL
- katello hack package is not in s390x images anymore, since we generate
the images by 'copr-image', not by IBM Cloud snapshots
- kernel-core is now fixed for aws builders, update it as well as other
packages when updating the image
- don't install subscription-manager when spawning builder, it is
already available from image preparation time
- copr-rh-subscribe.sh script is also prepare_base_image only
- we don't modify /etc/copr-rpmbuild/mock.cfg.j2 for quite some time,
drop the hack
- the /usr/bin/copr-update-builder is fixed upstream, drop the
lineinfile hack
- epel-8 configs are installed via files/mock/epel now
- no need to touch /etc/rpkg.conf nowadays on f35
- prepare mockbuilder user only when preparing image, not spawning
- disable coredumps only when preparing the image
- disable services only when preparing the image
- don't update rubygem-gem2rpm all the time, on F35 we have new enough
version
- the /etc/copr-rpmbuild/rpkg.conf.j2 is to be kept as is on F35
