Whenever i execute helm upgrade my applications scale to the max because on startup they need quite some resources to initiate and the autoscaler thinks we are under high load.
Is there a way to make this process a bit smarter?
I thought about these things, but would like to verify if there is maybe a consensus about the best way:
- Delete the autoscaler before deployment, so when you do helm-upgrade it gets created again
- Increasing stabilizationWindowSeconds on scaleUp to 60 seconds, but this makes the application less likely to really catch a high load when it happens.