HA k3s deployment failed to proxy to backend

[falkia34]

Describe the bug
I deployed authentik with K8s, using ArgoCD, cloudnativepg 17 with read replica, and minio. I got "failed to proxy to backend" on the log, and the pod status on amd64 node stuck on progressing, with message "Readiness probe failed: Get "http://10.15.0.123:9000/-/health/ready/": context deadline exceeded (Client.Timeout exceeded while awaiting headers)".

To Reproduce
Steps to reproduce the behavior:

1. Deploy argocd, cloudnativepg with read replica, and minio
2. Deploy authentik with kustomization and helm chart, with 3 server replicas

Expected behavior
The deployment run smoothly

Logs
https://hastebin.com/share/igupoxoful.json
https://hastebin.com/share/ilecucoquz.python

Version and Deployment (please complete the following information):

• authentik version: 2024.12.2
• Deployment: k3s cluster with 3 nodes mixed arch, 1 amd64 and 2 arm64

[falkia34]

I think this problem is related to #5059, but in my case, it was caused by slow storage on redis instance because it was using longhorn storage class. I switched it to local-path, and everything seems to be fine now.

• Even after changing redis storage class, when I increased the server replicas to more than 1 ( in my case was 3), it would still encounter the same error.
• Looks like this issue related with Add redis cluster support #3979, Redis Cluster or other mechanisms available as ref: #3979 #5531, and Support redis sentinel #1285. Also, I'm not sure whether this related or not, but there is redis cluster support PR said ready to be merged support redis cluster transport  celery/kombu#2204