Hi Kenji Kagei,
Proposal 2:
You asked if setting the deployment slot 100% traffic (with a production slot at 0%) can serve as a workaround to avoid downtime during a scale-up.
- While it seems that it can separate production, deployment slots cannot spread to various app service schemes.
So if you are scaling from S2 to P0V3, you cannot assign your deployment slot in a separate app service plan as compared to your production slot.
In other words, both the production slots and all the purposeless slots share the same app service plan. So even if you direct 100% traffic to a slot, the underlying infrastructure is still affected when you scale - and it means:
All slots will restart, in which a traffic is obtained.
Users can experience a brief disruption, even if it is just an eye.
- Unfortunately, slot swapping works within only one app service scheme. You cannot swap one slot from one app service plan to another, so the proposal 1 is not viable
Creating a new app service in a separate plan (P0V3),
Posting your app there with the same settings,
And then users root it (perhaps through DNS or similar),
Then it is not really a slot swap, but a blue-green deployment, and it is a valid and recommended zero-downtime strategy.
Yes, when you score an app service plan, the azure will restart all the applications running on it, as the underlying VM SKU and calculation environment change.
This is true, even if you remove the traffic using a deployment slot, as the slots still live under the same plan.
To avoid the restart application, create new azure service plan
if you have any further concerns or queries, please feel free to reach out to us.