Optimize Live Migration Scheduling by Preemptively Failing Non-Migratable PCI Requests

[Uggla]

Description:

Currently, if a boot request contains non-live-migratable PCI devices, the system proceeds with scheduling live migration before eventually failing. This leads to unnecessary resource consumption.

Expected Behavior:

The system should detect non-live-migratable PCI devices before scheduling the migration and fail immediately, preventing unnecessary scheduling attempts.

Actual Behavior:

The system schedules live migration before checking for non-migratable PCI devices, leading to wasted resources and delayed failure.

Steps to Reproduce:

    Submit a boot request containing non-live-migratable PCI devices.
    Observe that the system proceeds with scheduling live migration.
    Eventually, the migration fails due to the non-migratable PCI devices.

Proposed Fix:

Move the existing checks from the conductor earlier in the process, ideally before the scheduling part, so the request fails immediately.

Note:
This bug could be addressed as part of the refactoring of the migration code path, which currently consists of two separate code paths and should ideally be unified into one.

[Uggla (rene-ribaud)]

The code is working today, this is more an optimization of the migration process.