Open tkinz27 opened 4 years ago
@akshaysngupta Any thoughts/update on this?
This would be a great feature and would make the tweaks described in this document less important: https://azure.github.io/application-gateway-kubernetes-ingress/how-tos/minimize-downtime-during-deployments/
Is your feature request related to a problem? Please describe. A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
Readiness Gates allow the ingress controller to provide feedback to the kubernetes scheduler about when the load balancer is ready to remove the pod or add the pod. By providing this feedback, you can get nearly zero downtime rolling deployments.
See https://www.youtube.com/watch?v=Vw9GmSeomFg for a better explanation of the issue.
I have a situation where we have to set
maxSurge
on our deployments rolling update to0
which makes this even more important (service uses ahostPort
). So what happens is we need our pods to go completely away, then get added and fully become ready (meaning seen as healthy by the load balancer) before moving on with the next set of pods to upgrade. I am coming from AWS, and no amount of tuning the connection draining, preStop commands was able to completely remove the 502s until the alb-ingress-controller added support for readinessGates.Describe the solution you'd like A clear and concise description of what you want to happen.
Something similar to alb-ingress-controller's readinessGate support.