Kubernetes Ingress HA

I wonder how to solve a highly available proxy/ingress to my web services running on Kubernetes cluster on AWS.

Right now, I have a nginx-ingress-controller deployed on a node, this node has assigned EIP and this acts as an edge reverse proxy to access web services deployed on the cluster.

This is pretty much SPOF, when the nginx controller is down, or the EC2 node is broken or something, there is no way to access the services.

Normally, I'd deploy keepalived and setup it to reassign the EIP to the standby proxy node when the primary nginx proxy fail.

What the pattern to implement this with Kubernetes? I'd go for AWS ELB, but it doesn't support HTTP2 for example :(

Solution 1:

ELB doesn't support http2 per se, but it does support TCP. Your best bet is to front your k8s cluster nodes with the ELB. If you want fancier load balancing features, you could front an HAProxy or nginx cluster with an ELB, and rely on HAProxy or nginx to do whatever additional rewrites/inspect/routing/etc prior to hitting your k8s cluster.