Hard to Inject Provider-specifc Logic into Quota Check on Provision Requests

[miyamotoh]

It became obvious, while working on adding proper quota checks for CPU upon VM provisioning on IBM Power, that it's not easy to inject provider-specific logic in the workflow. This may be stemming from the fact that IBM Power cloud manager doesn't use the Flavor field to indicate how much CPU for the provisioning, that's assumed by the core class MiqProvisionRequest and ServiceTemplateProvisionRequest. Also, IBM Power allows fraction, instead of integer, to count its CPU allocation.

So, this may well go much wider, but to get my point across, I think we want some or all of ./content/automate/ManageIQ/System/CommonMethods/QuotaMethods.class (in manageiq-content) to lessen assumptions on how certain ManageIQ constructs (ie. Flavor) are used and become friendlier to provider overrides (ie. used, requested).

[jaywcarman]

See ManageIQ/manageiq-providers-ibm_cloud#454 for details specific to the IBM Cloud PowerVS provider.

[jaywcarman]

Here's an example of logic that should be pluggable in the provider repos: https://github.com/ManageIQ/manageiq/blob/master/app/models/mixins/miq_provision_quota_mixin.rb#L394

[Fryguy]

cc @agrare

[agrare]

The key is going to be finding something which is provider specific to key off of, then move the quota logic somewhere that a provider can override. In the miq_provision_quota_mixin.rb there is a def vendor() method which checks request.source.vendor. This request.source is very likely the source vm/template object which would be of a specific provider subclass. Whatever logic is needed to be overridden can then be moved into that class and the "default" behavior moved into the base VmOrTemplate class.

[jaywcarman]

This request.source is very likely the source vm/template object which would be of a specific provider subclass.

I dropped a byebug in the vm_quota_values method here and can confirm that the "source" is the template subclass.

(byebug) l =

[288, 297] in /home/jwcarman/src/github.com/ManageIQ/manageiq/app/models/mixins/miq_provision_quota_mixin.rb
   288:      })
   289:   end
   290: 
   291:   def vm_quota_values(pr, result)
   292:     byebug
=> 293:     num_vms_for_request = number_of_vms(pr)
   294:     return if num_vms_for_request.zero?
   295:     flavor_obj = flavor(pr)
   296:     result[:count] += num_vms_for_request
   297:     result[:memory] += memory(pr, cloud?(pr), vendor(pr), flavor_obj) * num_vms_for_request
(byebug) pr.source
  VmOrTemplate Load (1.0ms)  SELECT "vms".* FROM "vms" WHERE "vms"."id" = $1 LIMIT $2  [["id", 8], ["LIMIT", 1]]
  VmOrTemplate Inst Including Associations (0.4ms - 1rows)
#<ManageIQ::Providers::IbmCloud::PowerVirtualServers::CloudManager::Template id: 8, vendor: "ibm_power_vs", format: "", version: nil, name: "CentOS-Stream-8", description: "stock", location: "unknown", config_xml: nil, autostart: nil, host_id: nil, last_sync_on: nil, created_on: "2023-08-18 16:02:12.022095000 +0000", updated_on: "2023-08-18 16:02:12.022095000 +0000", storage_id: nil, guid: "e8844402-543c-43da-a916-bf7e5af54d9b", ems_id: 3, last_scan_on: nil, last_scan_attempt_on: nil, uid_ems: "31e22a48-6dee-4e2b-89f1-e669b936acd1", retires_on: nil, retired: nil, boot_time: nil, tools_status: nil, standby_action: nil, power_state: "never", state_changed_on: "2023-08-18 16:02:12.012183000 +0000", previous_state: nil, connection_state: nil, last_perf_capture_on: nil, registered: nil, busy: nil, smart: nil, memory_reserve: nil, memory_reserve_expand: nil, memory_limit: nil, memory_shares: nil, memory_shares_level: nil, cpu_reserve: nil, cpu_reserve_expand: nil, cpu_limit: nil, cpu_shares: nil, cpu_shares_level: nil, cpu_affinity: nil, ems_created_on: nil, template: true, evm_owner_id: nil, miq_group_id: 1, linked_clone: nil, fault_tolerance: nil, type: "ManageIQ::Providers::IbmCloud::PowerVirtualServers...", ems_ref: "31e22a48-6dee-4e2b-89f1-e669b936acd1", ems_cluster_id: nil, retirement_warn: nil, retirement_last_warn: nil, vnc_port: nil, flavor_id: nil, availability_zone_id: nil, cloud: true, retirement_state: nil, cloud_network_id: nil, cloud_subnet_id: nil, cloud_tenant_id: nil, raw_power_state: "never", publicly_available: nil, orchestration_stack_id: nil, retirement_requester: nil, tenant_id: 1, resource_group_id: nil, deprecated: nil, storage_profile_id: nil, cpu_hot_add_enabled: nil, cpu_hot_remove_enabled: nil, memory_hot_add_enabled: nil, memory_hot_add_limit: nil, memory_hot_add_increment: nil, hostname: nil, ems_ref_type: nil, restart_needed: nil, ancestry: nil, placement_group_id: nil>

Whatever logic is needed to be overridden can then be moved into that class and the "default" behavior moved into the base VmOrTemplate class.

I'll start working on a set of PRs to move number_of_cpus from the MiqProvisionQuotaMixin module into the ManageIQ::Providers::CloudManager::Template class and the IbmCloud::PowerVirtualServers plugin's ManageIQ::Providers::IbmCloud::PowerVirtualServers::CloudManager::Template subclass.

[miq-bot]

This issue has been automatically marked as stale because it has not been updated for at least 3 months.

If you can still reproduce this issue on the current release or on master, please reply with all of the information you have about it in order to keep the issue open.