Scale down pods on the previous stack after switching #443
Description
Feature Request
Is your feature request related to a problem or unsupported use case? Please describe.
After a blue/green deployment, keeping the previous stack for a few days is important in case we need to quickly rollback the code. However, on projects that are scaled up with many replica pods, keeping a stack with all its pods mobilizes a lot of resources for nothing.
Describe the solution you'd like
- Automatically scale down a stack after switching it to
previouse.g all pods set to x1 replica - Automatically scale up the
previousstack before initiating a rollback ie all pods set to the number of replica specified in their intial specification.
Activity
jmaupetit commentedon Jan 21, 2020
madmatah commentedon Jan 22, 2020
sampaccoud commentedon Jan 22, 2020
To go a bit further in our brainstorming:
nextstack should also be deployed with a limited number of pods and only scaled up just before switching,previousstack, we should not scale down celery workers because they still have tasks to handle after switching (can take up to a few hours to finish them so maybe we should also find a way to forbid deleting the previous stack before its workers have finished handling their tasks...)