Port/volume status not refreshed after server attachment due to SSA field owner race

[mandre]

Problem description

After a port is attached to a server via Nova's interface-attach API, the port controller may never observe the Neutron status transition from DOWN to ACTIVE. The same race affects volumes (available → in-use). This causes test assertions on status.resource.status to time out.

Observed

CI flake in server-update test step 1: assertion port2.status.resource.status == 'ACTIVE' timed out after 4 minutes.

• https://github.com/k-orc/openstack-resource-controller/actions/runs/28871247632/job/85634311590
• Environment: OpenStack epoxy only (flamingo and gazpacho passed)

Neutron API logs confirm the port went ACTIVE at 14:08:56, but ORC's last fetch was at 14:08:54 (saw DOWN). No subsequent fetch occurred until namespace deletion at 14:12:45.

Root cause

serverToPortMapFunc detects that a port attached to a server still has status DOWN and attempts to re-trigger reconciliation by:

1. Patching the port's Progressing condition to True via SSA
2. Returning a reconcile request

However, the SSA patch uses field owner GetSSAFieldOwner(controllerName) (openstack.k-orc.cloud/portcontroller), while the generic reconciler's UpdateStatus uses GetSSAFieldOwnerWithTxn(controllerName, SSATransactionStatus) (openstack.k-orc.cloud/portcontroller/status). These are different SSA field owners, both with ForceOwnership.

This creates a race:

1. Port controller reconciles port — fetches from Neutron (DOWN), starts writing status
2. Server controller writes Server status with port in interfaces → serverToPortMapFunc fires
3. MapFunc patches Progressing=True (field owner: portcontroller)
4. Port controller's UpdateStatus writes Progressing=False (field owner: portcontroller/status), overwriting step 3
5. MapFunc's enqueued reconcile request is processed — shouldReconcile() sees Progressing=False → skips

The port is never re-fetched from Neutron. serverToVolumeMapFunc has the identical pattern.

Affected code

• internal/controllers/port/controller.go — serverToPortMapFunc SSA patch
• internal/controllers/volume/controller.go — serverToVolumeMapFunc SSA patch

ORC version

main

Additional information

No response

Relevant log output

[mandre]

We can't use the same SSA owner otherwise we're clearing all the fields. I've simply added another defense mechanism to reconcile when we detect a mismatch (in addition to what was already done in #765).