09e678ee14
When rescheduling from a failed cold migrate / resize, the compute service does not pass the request spec back to conductor so we create one based on the in-scope variables. This introduces a problem for some scheduler filters like the AggregateMultiTenancyIsolation filter since it will create the RequestSpec using the project and user information from the current context, which for a cold migrate is the admin and might not be the owner of the instance (which could be in some other project). So the AggregateMultiTenancyIsolation filter might reject the request or select a host that fits an aggregate for the admin but not the end user. This fixes the problem by using the instance project/user information when constructing the RequestSpec which will take priority over the context in RequestSpec.from_components(). Long-term we need the compute service to pass the request spec back to the conductor during a reschedule, but we do this first since we can backport it. NOTE(mriedem): RequestSpec.user_id was added in Rocky in commit |
||
|---|---|---|
| .. | ||
| tasks | ||
| __init__.py | ||
| api.py | ||
| manager.py | ||
| rpcapi.py | ||