I'm happy to help you understand how LiveKit supports autoscaling when self-hosting in AWS using Kubernetes Horizontal Pod Autoscalers (HPA), mainly configured through its Helm charts. Here’s a friendly overview of how it works: • LiveKit Server: Autoscaling is handled through the Helm chart's

autoscaling

configuration block. It leverages Kubernetes HPA based on CPU or memory usage. For instance, you can set a target CPU utilization of 60%, along with minimum and maximum replica counts to suit your needs. • Ingress and Egress Services: These services also support autoscaling! For Ingress, you can use a custom Prometheus metric

livekit_ingress_available

to scale based on how ready the service is, rather than just CPU or memory. This requires exposing metrics via the

prometheus_port

. • Graceful Scaling: When scaling down, pods enter a "draining mode," which means active rooms and sessions keep running smoothly while new connections are paused, helping to minimize any disruption. • Traffic Distribution: For AWS EKS setups, the Helm chart works with AWS ALB (Application Load Balancer) for TLS termination and routing traffic efficiently, supporting autoscaling by distributing traffic effectively. If you’d like, I’m here to walk you through any part of the setup or answer any questions you have about autoscaling! Sources: ingress-sample.yaml | Self-hosting the Ingress Service | LiveKit Docs | Deploy to Kubernetes | LiveKit Docs | Distributed Setup | LiveKit Docs | Self-hosting the Egress Service | LiveKit Docs