So after hitting my head a while against the HA setup with dqlite that k3s support in “Experimental” phase, had to let go after hitting https://github.com/rancher/k3s/issues/1391
Good thing is that with only one master, things are way faster.
I have also deployed a DaemonSet with DigitalOcean Agent on it so that I can have some metrics that don’t rely on k3s.
apiVersion: apps/v1
kind: DaemonSet
metadata:
name: do-agent
spec:
selector:
matchLabels:
app: do-agent
template:
metadata:
name: do-agent
labels:
app: do-agent
spec:
containers:
- name: do-agent
image: digitalocean/do-agent:stable
imagePullPolicy: Always
volumeMounts:
- mountPath: /host/proc
name: host-proc
- mountPath: /host/sys
name: host-sys
volumes:
- name: host-sys
hostPath:
path: /sys
- name: host-proc
hostPath:
path: /proc
BTW, the kubernetes plugin for IntelliJ based IDEs works like a charm.