Ditching HAProxy (In My Homelab)
a year ago
- #homelab
- #kubernetes
- #networking
- Single HAProxy VM was a single point of failure in the homelab, causing access issues to Kubernetes services.
- Explored solutions like keepalived, VIPs, and MetalLB to address redundancy and high availability.
- MetalLB in layer 2 mode worked seamlessly for IPv4, simplifying service exposure via LoadBalancer.
- Challenges remain with IPv6 implementation despite successful IPv4 routing.
- Kubernetes distribution supported VIPs for kube-apiserver access, enhancing control plane reliability.
- Tailscale Kubernetes operator improved, enabling direct routing and LoadBalancer functionality with failover capabilities.
- Successfully eliminated HAProxy dependency, ensuring no service disruption during tests.
- Gained confidence in networking, motivating deeper exploration into BGP and IPv6.
- DNS played a crucial role in minimizing downtime during the transition.
- Homelab now operates with improved resilience against server failures.