🐛 Update to UBI9 to fix arm64 builds

[jmontleon]

Signed-off-by: Jason Montleon jmontleo@redhat.com

I know we weren't in a hurry to do this, however multi arch builds are experiencing odd failures on arm64 using UBI8.
https://github.com/konveyor/tackle2-ui/actions/runs/4168992831/jobs/7216425953#step:2:1067
https://github.com/konveyor/tackle2-ui/actions/runs/4174604421/jobs/7228394767#step:2:1077

I can reproduce much of this locally, and was able to workaround it using UBI9 node16 and node18 images without a problem both locally and within github actions:
https://github.com/jmontleon/tackle2-ui/actions/runs/4175222062/jobs/7229841318
https://quay.io/repository/jmontleon/tackle2-ui?tab=tags

I've opened an issue with qemu-user-static to see if there are any known issues we might be hitting, or other debugging steps we can try to resolve the issue, but this is probably our best bet to move on at the moment.
multiarch/qemu-user-static#185

[gildub]

Upgrade to UBI9 put on hold because the entire tackle2 stack needs to be aligned.

See #638 for follow up.

[gildub]

@jmontleon, that's interesting issue which gives an opportunity to bump to latest UBI.
Meanwhile we have to wait for the whole project to bump UBI along.

N/A

[gildub]

Just to make sure we don't merge too early.

[jmontleon]

@jmontleon, that's interesting issue which gives an opportunity to bump to latest UBI. Meanwhile we have to wait for the whole project to bump UBI along.

• I don't agree with this. tackle2-addon-test-generator has been UBI 9 all along.
• tackle2-addon-windup had to be bumped in order to get a publicly available s390x svn package (CentOS 8 did not have an s390x release). The windisco.com RPMS were originally for EL6 and i686 and x86_64 only, and were not used downstream.
• in general for node and golang builds, all other things being equal, I don't view changing the distribution version as high risk. There's some, but not much.
• Downstream does not have to change with us. I get that the desire is to keep drift minimal, but I know for a fact it won't be the first time ever we've been on different releases upstream and down (ansible-runner upstream moved WAY faster than downstream, and then I/we missed a product rename that had us sitting longer on EL7 down while containers upstream were well and away on 8. Also, see SVN above.)