Installation·Certified Kubernetes Administrator
Three-Node kubeadm Cluster Setup
Build a control plane and two workers, then use the second worker for what it is actually for: watch six replicas spread across both, drain one node and see every pod reschedule onto the other while the Service keeps answering.
Installation and Cluster Builds Guide 9 of 103 Intermediate
- Kubernetes1.36.4
- Cluster3 nodes
- Runtimecontainerd 2.2.6
- CNICalico v3.32.1
- TimeAbout 50 min
Tokens, hashes, pod name suffixes and timings differ on every run. The token here expired within a day of being created.
- Kernel7.0.0-29-generic
- Calicov3.32.1
| Server Name | IP Address | OS | Roles | CPU | RAM | HDD |
|---|---|---|---|---|---|---|
| CKA2001 | 192.168.0.181 | Ubuntu 26.04 LTS | Control Plane Node | 2 Core | 4 GB | 50 GB |
| CKA2001-NODE01 | 192.168.0.182 | Ubuntu 26.04 LTS | Worker Node | 2 Core | 4 GB | 50 GB |
| CKA2001-NODE02 | 192.168.0.183 | Ubuntu 26.04 LTS | Worker Node | 2 Core | 4 GB | 50 GB |
This guide includes
Use this when two nodes are not enough to rehearse the thing you actually need - taking a machine out of service without dropping traffic. This matters because draining only means anything once there is somewhere else for the Pods to go.
- preparing three machines identically - swap off, modules, sysctls - regardless of their role
- initialising one control plane and joining two workers with the same token
- asking for six replicas and watching them spread across both workers
- draining a worker, watching the cluster absorb its Pods, and proving the Service never stopped answering
- reversing the cordon half of a drain with
kubectl uncordon
Before you start
- Three Ubuntu 26.04 machines, 2 CPUs and 2 GB of RAM minimum each, each with a unique hostname.
- All three on the same network. Both workers need to reach the control plane on TCP 6443.
sudoand a fixed address on each.- A pod CIDR that does not overlap your LAN. This lab is 192.168.0.0/24 and uses 10.244.0.0/16.
- The two-node guide covers the join mechanics in detail. This guide runs the same join twice and then spends its time on what two workers let you do.
-
Prepare all three machines
-
Install the packages on all three machines
-
Initialise the control plane
-
Join both workers with the same token
-
Install Calico
-
Watch replicas spread across both workers
-
Drain a worker and watch the cluster absorb it
-
Confirm the Service never stopped answering
-
Bring the node back and clean up