feat(vm-consumption-report): add VM consumption reporting role

[stevefulme1]

Summary

• Adds vm_consumption_report role and playbook for generating consumption reports of OpenShift Virtualization VMs
• Queries VirtualMachine and VirtualMachineInstance resources via kubernetes.core.k8s_info, supporting namespace and label selector filtering
• Reconciles directly managed hosts (SSH/WinRM) with API-managed VMs via UUID, guest hostname, and IP address matching
• Outputs reports in JSON or CSV format with summary counts (indirect, direct, reconciled)

Moved from

Originally submitted as redhat-cop/openshift_virtualization_migration#29 — relocated here as this is day 2 ops functionality (reporting/inventory of running VMs), not migration logic.

FQCN updated from infra.openshift_virtualization_migration to infra.openshift_virtualization_ops.

Test plan

• Verify yamllint passes
• Verify ansible-lint --offline passes production profile
• Run playbook against an OpenShift cluster with KubeVirt VMs
• Verify JSON output includes correct VM entries and summary counts
• Verify CSV output renders correctly
• Test reconciliation with vm_consumption_report_direct_hosts entries
• Test with vm_consumption_report_include_status: false to skip VMI queries

🤖 Generated with Claude Code

[sabre1041]

@stevefulme1 Reviewed the report generation. There are some questions that I have based on this integration.

1. See the additional filter for removing null values in the lastTransitionTime property
2. The reconciled field will never contain a positive value as a match is not possible
3. Is the goal to have a set of hostnames/uuid's and check if a match is found in the cluster?

[sabre1041]

Suggested change

	| sort(attribute='lastTransitionTime')
	| selectattr('lastTransitionTime', 'defined') | rejectattr('lastTransitionTime', 'none')
	| sort(attribute='lastTransitionTime')

Need to filter out items that contain null values

[stevefulme1]

Good catch — pushed this fix in 89241f3. Added selectattr('lastTransitionTime', 'defined') | rejectattr('lastTransitionTime', 'none') before the sort.

[sabre1041]

hostname is not a valid field for guestOSInfo so we will never have a reconcile entry

[stevefulme1]

You're right — guestOSInfo doesn't expose a hostname field. The guest agent hostname isn't available at that path in the VMI status.

The reconciliation still works via UUID and IP address matching, but the hostname branch of _direct_match is effectively dead code. I'll remove guestOSInfo.hostname as a matching source. If there's a different VMI status path you'd recommend for guest hostname, happy to wire that in instead.

[sabre1041]

Targeting a VMI for reporting purposes is risky as no instance is created when a VM is not running.

There are a few options for looking into capturing VM information. It is not exposed in the status field, but you can check the .spec.template.spec.hostname field to see if it has been explicitly defined and maybe just fallback to metadata.name as that is used by default

[stevefulme1]

Addressing the review:

1. Null lastTransitionTime filter — Fixed in 89241f3. Added selectattr('lastTransitionTime', 'defined') | rejectattr('lastTransitionTime', 'none') before the sort.

2. guestOSInfo.hostname doesn't exist — Fixed in 6b84fa5. Removed the invalid field reference. _guest_hostname is now set to empty string, which disables the hostname branch of _direct_match. Reconciliation still works via UUID and IP address matching. If there's a valid VMI status path for guest hostname I'm not aware of, happy to wire it back in.

3. Goal of the reconciliation — Yes, the intent is to take a list of known hosts (from AAP inventory or a static list via vm_consumption_report_direct_hosts) and check which ones have a matching VM in the cluster. Matches are tagged reconciled, cluster VMs with no match are indirect, and inventory hosts with no cluster VM are direct_only. This gives a consumption report showing which VMs are managed by AAP vs. discovered on the cluster without a corresponding inventory entry.