(RE-15156) Add redhat-9-arm64 support

[yachub]

I'm not sure about all the places that the abs and vmpooler platform data keys are used, but was a bit confused because the redhat8-AARCH64 specifies the x86 template name redhat-8-x86_64, is that a mistake?

beaker-hostgenerator/lib/beaker-hostgenerator/data.rb

Line 348 in da8ac9a

'template' => 'redhat-8-x86_64',

So for the new platform's vmpooler key I used the expected name that we would use in vmpooler for that platform, is that correct? Or should that key just be omitted?

EDIT: Ah it looks like those keys related to the hypervisor gem. The above definitely seems like a bug per:

> beaker-hostgenerator redhat8-AARCH64 --hypervisor vmpooler
---
HOSTS:
  redhat8-AARCH64-1:
    platform: el-8-aarch64
    template: redhat-8-x86_64
    hypervisor: vmpooler
    roles:
    - agent
CONFIG:
  pooling_api: https://vmpooler-prod.k8s.infracore.puppet.net/

[codecov]

Codecov Report

Patch coverage has no change and project coverage change: -0.01% ⚠️

Comparison is base (da8ac9a) 0.73% compared to head (0bca3ce) 0.73%.

❗ Current head 0bca3ce differs from pull request most recent head 86da134. Consider uploading reports for the commit 86da134 to get more accurate results

Additional details and impacted files

@@            Coverage Diff            @@
##           master    #327      +/-   ##
=========================================
- Coverage    0.73%   0.73%   -0.01%     
=========================================
  Files          15      15              
  Lines        1765    1776      +11     
=========================================
  Hits           13      13              
- Misses       1752    1763      +11     

Files Changed	Coverage Δ
lib/beaker-hostgenerator/data.rb	0.00% <0.00%> (ø)

☔ View full report in Codecov by Sentry.

📢 Have feedback on the report? Share it here.

[ekohl]

For RHEL I haven't looked at correct generation yet. As you noticed, the data was inconsistent. Like redhat7-AARCH64 uses centos-7-arm64 in abs.

Ideally we would generate the ABS template in https://github.com/voxpupuli/beaker-hostgenerator/blob/master/lib/beaker-hostgenerator/hypervisor/abs.rb and vmpooler in https://github.com/voxpupuli/beaker-hostgenerator/blob/master/lib/beaker-hostgenerator/hypervisor/vmpooler.rb. Then it would be reduced to just listing the versions in generate_os_info.

The biggest problem is that Vox Pupuli doesn't use ABS or vmpooler so we really don't know what's available and what makes sense.

[joshcooper]

Seems like a bug to me too. IMO BHG should always return a template/platform name that is consistent with the key (BHG string):

We do cross compile on some platforms, but the decision about whether to cross-compile and which build host to use lives in the platform definition for the vanagon project, for example: https://github.com/puppetlabs/puppet-runtime/blob/6c8bbc664aea36cbadc12ab0d62773250faf70f1/configs/platforms/sles-12-ppc64le.rb#L31-L32

In other words, I don't see a use case for BHG ever returning an Intel-based template for an ARM-based key.

[joshcooper]

Not sure why, but I'm having trouble getting this working. This line is causes an error if platform is not defined:

https://github.com/voxpupuli/beaker-hostgenerator/blob/013dab5a0ab39c5d1b7383ec1265c382eb325dac/lib/beaker-hostgenerator/hypervisor/abs.rb#L27C39-L27C50

$ bx beaker-hostgenerator redhat9-ARM64a
bundler: failed to load command: beaker-hostgenerator (/home/josh/.rbenv/versions/2.7.6/lib/ruby/gems/2.7.0/bin/beaker-hostgenerator)
Traceback (most recent call last):
...
	 6: from /home/josh/work/beaker-hostgenerator/bin/beaker-hostgenerator:10:in `<top (required)>'
	 5: from /home/josh/work/beaker-hostgenerator/lib/beaker-hostgenerator/cli.rb:174:in `execute!'
	 4: from /home/josh/work/beaker-hostgenerator/lib/beaker-hostgenerator/cli.rb:168:in `execute'
	 3: from /home/josh/work/beaker-hostgenerator/lib/beaker-hostgenerator/generator.rb:30:in `generate'
	 2: from /home/josh/work/beaker-hostgenerator/lib/beaker-hostgenerator/generator.rb:30:in `each'
	 1: from /home/josh/work/beaker-hostgenerator/lib/beaker-hostgenerator/generator.rb:52:in `block in generate'
/home/josh/work/beaker-hostgenerator/lib/beaker-hostgenerator/hypervisor/vmpooler.rb:22:in `generate_node': undefined method `gsub' for nil:NilClass (NoMethodError)

I get the same when passing abs or vmpooler as the hypervisor

[yachub]

@joshcooper Following precedent of most recently added platforms I used "AARCH64" instead of "ARM64". Does that work for you?

> beaker --version
      wWWWw
      |o o|
      | O |  5.3.1!
      |(")|
     / \X/ \
    |   V   |
    |   |   |
    
> beaker-hostgenerator -v
2.4.0

> beaker-hostgenerator redhat9-AARCH64a --hypervisor abs
---
HOSTS:
  redhat9-AARCH64-1:
    platform: el-9-aarch64
    template: redhat-9-arm64
    hypervisor: abs
    roles:
    - agent
CONFIG: {}

[ekohl]

Still, would be good to either silently ignore unknown architectures or fail with a clear error instead of a stack trace.

[joshcooper]

Still, would be good to either silently ignore unknown architectures or fail with a clear error instead of a stack trace.

yeah that could be handled better. Took me a bit to understand what was going on. I also run into errors like that when trying to run beaker tests and I've forgotted to do bundle update to pull in a version of BHG that recognizes the new platform.

I used "AARCH64" instead of "ARM64". Does that work for you?

Yeah I think it's good for the architecture to be consistent for all versions of the same OS. As for why we use different names for one OS vs another, I can only refer to https://twitter.com/lorenc_dan/status/1699794514586763408:

[ekohl]

$ bx beaker-hostgenerator redhat9-ARM64a

Funny, I use be as the bundle exec alias.

Still, would be good to either silently ignore unknown architectures or fail with a clear error instead of a stack trace.

#337