The event is also notified even it is not a versioned notification and not whitelisted

[Wenping Song]

IMO, we shouldnot send the event that is not a versioned notification and not whitelisted.
https://github.com/openstack/nova/blob/master/nova/rpc.py#L413, need to put L420 to else condiftion.

[Balazs Gibizer (balazs-gibizer)]

We either raise an Error or emit a warning depending on the fact that the code runs in a test env or in production. The self.fatal is set to True by ForbidNewLegacyNotificationFixture that is used in our base TestCase class.

The logic here is that we should not introduce new legacy notifications.

If that happens then the test case covering the new legacy notification will fail with the exception raised at L416.

Originally we did not block new legacy notifications in production (when self.fatal is False) environment as nova allows out of tree code to be imported like scheduler filters, or virt drivers, and we don't want to break those code in production if they emitting legacy notifications.

What do you suggest here? Should we raise exception for non whitelisted legacy notifications in production env? If so then please bring this up on the ML or weekly meeting as this needs a wider discussion.

I set this to incomplete. Please set it back to new if the wider discussion results in an agreement to raise such exception in production.

[Wenping Song (wenping1)]

We neednot raise exeption for non whitelisted legacy notifications in production env. I suggest not to send the event that not in legacy notifications whitelist. Because in our production env, we may not need so much legacy notifications, when we change the whitelist, it doesnot take effect.

[Launchpad Janitor (janitor)]

[Expired for OpenStack Compute (nova) because there has been no activity for 60 days.]