Is there a working solution for K8s autoscale out for Ignition?

Ignition
1

Been working on a new deployment of Ignition lately. I have it tied to into a K8s cluster with ArgoCD.
Current Layout:
Frontend container w/loadbalancer,pvc,autoscale
Backend container w/loadbancer,pvc,autoscale
Mariadb containers (3 containers statefulset with replication enabled)

The only issue I am having and I am honestly not sure if it is supported is the autoscaleout. I have tried having the containers share the same PVC, seed the data, different pvcs, etc. My connection to the gateway will consistently want me to login, (fixed by sessionaffinity) but if it were to hit the second replica that was created, the login screen will just continue to spin and not function correctly.

If this is not supported, is my only option created individual gateways and utilizing EAM on each one?

Any help is appreciated.

2

I am working on a similar setup to have Ignition running behind K8 with the ability to scale up and down accordingly with a PVC. I understand that Ignition will need sticky sessions and there are methods to do this within K8.

I have setup sticky session before for other application and one needs to ensure session timeouts are configured as such that users don't get stuck on one pod. Ideally ignition should be updated to not require sticky sessions as this adds complexity to updates because you have to wait for the few users to end there sessions. It also makes rolling restarts take longer.

K8 autoscale is the way to go. I am just starting my project so if you find the magic please share.