This how to guide describes how to use the MicroK8s charm to deploy and manage MicroK8s clusters
Requirements
- An existing bootstrapped Juju controller on a cloud like AWS, Azure, OpenStack, LXD, etc
Deploy a single-node cluster
Create a new model called k8s-1, then deploy the microk8s charm. Ensure the specified constraints are enough to account for your workloads:
juju add-model k8s-1
juju deploy microk8s --constraints 'mem=4G root-disk=20G' --channel 1.28/stable
# make sure we can reach the microk8s cluster with a kubeconfig
juju expose microk8s
You can check the progress of the deployment with juju status. After some time, it should look like this:
Model Controller Cloud/Region Version SLA Timestamp
k8s-1 zs zerostack/KHY 3.1.5 unsupported 19:25:47+03:00
App Version Status Scale Charm Channel Rev Exposed Message
microk8s 1.28.1 active 1 microk8s 1.28/stable 213 no node is ready
Unit Workload Agent Machine Public address Ports Message
microk8s/0* active idle 0 172.16.100.206 16443/tcp node is ready
Machine State Address Inst id Base AZ Message
0 started 172.16.100.206 ccef3b37-8696-4e8f-8d7f-fff3392748f6 ubuntu@22.04 nova ACTIVE
You can also use juju exec to run kubectl commands:
juju exec --unit microk8s/leader -- microk8s kubectl get node
The kubectl output should look like this:
NAME STATUS ROLES AGE VERSION
juju-131248-k8s-1-0 Ready <none> 26m v1.28.1
Scale to a 3-node cluster
In order to scale the cluster to 3 nodes, you can use juju add-unit
# add two more units to microk8s
juju add-unit microk8s -n 2
This will spin two new machines, install MicroK8s and form a 3-node cluster. You can check the progress with juju debug-log and juju status. After the cluster is formed, the status should look like this:
Model Controller Cloud/Region Version SLA Timestamp
k8s-1 zs zerostack/KHY 3.1.5 unsupported 19:48:47+03:00
App Version Status Scale Charm Channel Rev Exposed Message
microk8s 1.28.1 active 3 microk8s 1.28/stable 213 no node is ready
Unit Workload Agent Machine Public address Ports Message
microk8s/0* active idle 0 172.16.100.206 16443/tcp node is ready
microk8s/1 active idle 1 172.16.100.180 16443/tcp node is ready
microk8s/2 active idle 2 172.16.100.85 16443/tcp node is ready
Machine State Address Inst id Base AZ Message
0 started 172.16.100.206 ccef3b37-8696-4e8f-8d7f-fff3392748f6 ubuntu@22.04 nova ACTIVE
1 started 172.16.100.180 db1384ed-6b95-408a-8d63-e74175314168 ubuntu@22.04 nova ACTIVE
2 started 172.16.100.85 83aca2ff-bb52-4b05-9a94-375a3501865d ubuntu@22.04 nova ACTIVE
We can also verify that the MicroK8s cluster is formed and has 3 nodes:
juju exec --unit microk8s/leader -- microk8s kubectl get node
now returns:
NAME STATUS ROLES AGE VERSION
juju-131248-k8s-1-0 Ready <none> 30m v1.28.1
juju-131248-k8s-1-1 Ready <none> 53s v1.28.1
juju-131248-k8s-1-2 Ready <none> 54s v1.28.1
Add worker nodes
The 3 nodes act as control-plane nodes by default, but can also execute any workloads. For larger environments, it is desirable to limit the number of control plane nodes to an odd-number (e.g. 3 or 5) and then add more nodes as workers. The worker nodes will not run any of the control plane services:
To add worker nodes, deploy a new microk8s-worker application and set role=worker.
# deploy 3 microk8s worker nodes
juju deploy microk8s microk8s-worker --channel 1.28/stable --config role=worker -n 3
When deploying with role=worker, the charm will stay in a waiting state until it is related to a control plane:
Model Controller Cloud/Region Version SLA Timestamp
k8s-1 zs zerostack/KHY 3.1.5 unsupported 19:59:47+03:00
App Version Status Scale Charm Channel Rev Exposed Message
microk8s 1.28.1 active 3 microk8s 1.28/stable 213 no node is ready
microk8s-worker waiting 3 microk8s 1.28/stable 213 no waiting for control plane
Unit Workload Agent Machine Public address Ports Message
microk8s-worker/0 waiting idle 3 172.16.100.74 waiting for control plane
microk8s-worker/1* waiting idle 4 172.16.100.110 waiting for control plane
microk8s-worker/2 waiting idle 5 172.16.100.161 waiting for control plane
microk8s/0* active idle 0 172.16.100.206 16443/tcp node is ready
microk8s/1 active idle 1 172.16.100.180 16443/tcp node is ready
microk8s/2 active idle 2 172.16.100.85 16443/tcp node is ready
Machine State Address Inst id Base AZ Message
0 started 172.16.100.206 ccef3b37-8696-4e8f-8d7f-fff3392748f6 ubuntu@22.04 nova ACTIVE
1 started 172.16.100.180 db1384ed-6b95-408a-8d63-e74175314168 ubuntu@22.04 nova ACTIVE
2 started 172.16.100.85 83aca2ff-bb52-4b05-9a94-375a3501865d ubuntu@22.04 nova ACTIVE
3 started 172.16.100.74 d5c83472-1e25-4980-9f2f-b5599f5035aa ubuntu@22.04 nova ACTIVE
4 started 172.16.100.110 e3b7ee86-804b-4d31-b4b9-83b2b2e1f770 ubuntu@22.04 nova ACTIVE
5 started 172.16.100.161 ff2a84f9-2dba-413d-b803-bda3d41de9c7 ubuntu@22.04 nova ACTIVE
To proceed, integrate microk8s-worker with microk8s, so that the nodes can join the cluster:
# 'microk8s' will be the control plane to 'microk8s-worker'
juju integrate microk8s:workers microk8s-worker:control-plane
After some time, the juju status output should look like this:
Model Controller Cloud/Region Version SLA Timestamp
k8s-1 zs zerostack/KHY 3.1.5 unsupported 20:01:00+03:00
App Version Status Scale Charm Channel Rev Exposed Message
microk8s 1.28.1 active 3 microk8s 1.28/stable 213 no node is ready
microk8s-worker 1.28.1 active 3 microk8s 1.28/stable 213 no node is ready
Unit Workload Agent Machine Public address Ports Message
microk8s-worker/0 active idle 3 172.16.100.74 node is ready
microk8s-worker/1* active idle 4 172.16.100.110 node is ready
microk8s-worker/2 active idle 5 172.16.100.161 node is ready
microk8s/0* active idle 0 172.16.100.206 16443/tcp node is ready
microk8s/1 active idle 1 172.16.100.180 16443/tcp node is ready
microk8s/2 active idle 2 172.16.100.85 16443/tcp node is ready
Machine State Address Inst id Base AZ Message
0 started 172.16.100.206 ccef3b37-8696-4e8f-8d7f-fff3392748f6 ubuntu@22.04 nova ACTIVE
1 started 172.16.100.180 db1384ed-6b95-408a-8d63-e74175314168 ubuntu@22.04 nova ACTIVE
2 started 172.16.100.85 83aca2ff-bb52-4b05-9a94-375a3501865d ubuntu@22.04 nova ACTIVE
3 started 172.16.100.74 d5c83472-1e25-4980-9f2f-b5599f5035aa ubuntu@22.04 nova ACTIVE
4 started 172.16.100.110 e3b7ee86-804b-4d31-b4b9-83b2b2e1f770 ubuntu@22.04 nova ACTIVE
5 started 172.16.100.161 ff2a84f9-2dba-413d-b803-bda3d41de9c7 ubuntu@22.04 nova ACTIVE
Verify with:
juju exec --unit microk8s/leader -- microk8s kubectl get node
Which should now print something like this:
NAME STATUS ROLES AGE VERSION
juju-131248-k8s-1-1 Ready <none> 13m v1.28.1
juju-131248-k8s-1-2 Ready <none> 13m v1.28.1
juju-131248-k8s-1-3 Ready <none> 58s v1.28.1
juju-131248-k8s-1-0 Ready <none> 42m v1.28.1
juju-131248-k8s-1-5 Ready <none> 56s v1.28.1
juju-131248-k8s-1-4 Ready <none> 56s v1.28.1