neoaggelos
commented
7 months ago Hi @zerowebcorp, pasting my response from a similar Slack thread this week:
Using a LoadBalancer service for the control plane is a chicken-egg problem, therefore it is typically not recommended
Most often, you need to setup a load balancer that proxies requests to the control plane nodes. The way to do this varies by cloud. The common step is to make sure that the loadbalancer IP or DNS is added as a SAN in the kube-apiserver certificates (see https://microk8s.io/docs/services-and-ports) for details.
Solutions include:
- using a virtual IP across the nodes (by means of keepalived or kube-vip)
- using an external load balancer provided by the cloud (e.g. Elastic LoadBalancer on AWS, octavia on openstack, etc)
Hello, I've checked the documentation on how to setup ha with microk8s. However I am not finding anything that explains the additional steps to do in order to put a load balancer ( I am using an azure load balancer) in front of the 3 VMs where microk8s is installed on HA mode so that I when I run
microk8s add-nodeI get the IP address of the load balancer instead of the node's ip where I executed the command so that the worker nodes use it to connect to one of the 3 manager nodes.