The epel10 tag is a special locked tag we use for the baseurl of the
local repo in the epel-10 template of mock-core-configs. It is similar
to the locked rawhide tag. During mass branching we need to change its
inheritance to the next epel10.x minor version tag.
Signed-off-by: Carl George <carlwgeorge@gmail.com>
The original make-koji-release-tags script this playbook is based on had
a significant flaw. The original design was to have external repos for
each epel10.x-base tag that changed URLs over time. For example,
epel10.0-base would have an external repo of el10.0-baseos that would
migrate through the following repos:
* https://infrastructure.fedoraproject.org/repo/centos/centos-10/BaseOS/$arch/os/
* https://infrastructure.fedoraproject.org/repo/centos/centos-10-snapshot/BaseOS/$arch/os/
* https://infrastructure.fedoraproject.org/repo/rhel/rhel10/10.0/repos/$arch/baseos/
During the epel10.0 mass branching, we discovered that wouldn't work.
CentOS and RHEL have many RPMs with the same filename but different
checksums as they have different metadata and are signed with a
different key. Koji build error out when the checksum of an RPM that
has already been indexed changes. The solution to this is to use
external repos whose URLs don't change, but then move the external repos
around to different tags over time. That is what we have already
implemented in production, we just never updated make-koji-release-tags
to reflect it.
This changes the playbook create the new epel10.x-base tag with the
correct external repo names and switch the existing epel10.x-base tag
from the centos-10 repos to the centos-10-snapshot repos.
Signed-off-by: Carl George <carlwgeorge@gmail.com>
bvmhost-a64-05 has less memory than 01-04, so fewer buildvm's on it.
However, even with that it sometimes is OOM killing one.
So, lets drop 48 for now to give it more free memory.
Signed-off-by: Kevin Fenzi <kevin@scrye.com>
These vm's run tasks like newrepos and tagging, so they can handle a lot
more smaller jobs like that. Increase things to 128 for now.
Signed-off-by: Kevin Fenzi <kevin@scrye.com>
OpenShift was failing to build mdapi because it was looking for Dockerfile
but the repository contains Containerfile. Added dockerStrategy.dockerfilePath
configuration to specify the correct container build file.
Signed-off-by: Akashdeep Dhar <akashdeep.dhar@gmail.com>
Remove volumemount references to myconfigpy-volume from cronjob and
deploymentconfig as the corresponding volume definition was missing,
causing OpenShift deployment failures.
Signed-off-by: Akashdeep Dhar <akashdeep.dhar@gmail.com>
Non rdu3 hardware doesn't have access to tang.
There's some slight exceptions on some hosts that use tpm,
but we can override that in hosts vars.
Also drop the rdu3 datacenter check, we can just use the
nbde boolean to determine if we run it or not.
Signed-off-by: Kevin Fenzi <kevin@scrye.com>
