[RFE] aggregate support in Ironic

[George Shuklin]

Right now Ironic have a single way to distinct one node from another during server creation: by using flavors and properties in the nodes, matching those in flavors.

Nova supports conception of aggregates/availability zones, when tenant has freedom to specify availability zone during instance creation (--availability-zone option for "openstack server create"/"nova boot"). It would be nice if ironic have a way to assign node to certain aggregate and take that information in account during instance scheduling.

[Hao Li (lihaosz)]

In my company, we do a lot of work for make ironic supported aggregate. I will commit the specs and then commit the codes soon.

[Ruby Loo (rloo)]

Are you still working on this Hao Li? There isn't any patch link here, so I'm setting it back to triaged.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Julia Kreger (<email address hidden>) on branch: master
Review: https://review.opendev.org/441924
Reason: Abandoning spec as it has not been updated in nearly three years. Functionally, the basic idea has been implemented with the conductor_groups functionality and the ability to select groups from the nova-compute.

[Julia Kreger (juliaashleykreger)]

Ironic implemented the idea of responsibility to a group of conductors which is assigned through the conductor_group field in the API, which then allows the nova configuration to move independently. In the case of scheduling to drive responsibility, it was just not possible with our model of management and the underlying needs for operators require more static architecture the closer we get to the hardware. As such, I'm going to mark this item as invalid.