Ironic nodes report CUSTOM_CUSTOM_FOO resource class

[John Garbutt]

Currently if you set CUSTOM_FOO in ironic, the virt driver now sends CUSTOM_CUSTOM_FOO to placement.

Really we shouldn't force users to drop the CUSTOM_ inside ironic.

Expected:
User sets CUSTOM_FOO in ironic.
Placement shows CUSTOM_FOO resources.

Actual:
Placement shows CUSTOM_CUSTOM_FOO resources

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/513001

[Matt Riedemann (mriedem)]

Being discussed, this might not be a bug.

[John Garbutt (johngarbutt)]

This is considered as designed, so will not be fixed.

It complicates the general case of mapping from ironic resource classes to placement resource classes.

I still think this is going to trip up many Ironic administrators. The docs are not clear that CUSTOM_FOO maps to CUSTOM_CUSTOM_FOO.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/513001
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=f73d6eb1dc928c811231e059f293901405bed1ee
Submitter: Zuul
Branch: master

commit f73d6eb1dc928c811231e059f293901405bed1ee
Author: John Garbutt <email address hidden>
Date: Wed Oct 18 11:47:49 2017 +0100

    Add test so we remember why CUSTOM_ prefix added

    While it is tempting as an ironic operator to have CUSTOM_FOO in
    placement, nova flavors and ironic resource class attributes,
    allow that makes for a more complicated mapping between ironic
    resource classes and the resource classes in the general case,
    and could cause confusion when there is a typo of CUSTOM vs CUSTM
    that could get mapped very differently into placement.

    Adding this test so it is clear this is a deliberate choice, and
    git blame will find this git commit message.

    Related-Bug: #1724524

    Change-Id: Ie38baea53e41ca7ba8875fd113cd777810f17db7