Talos Linux is an operating system built to run nothing but Kubernetes: it has
no SSH, no shell and no package manager, and everything is driven through an
API with the talosctl command. Choose it when you are building a new
fleet, you want identical nodes described by a YAML file, and your team
accepts no longer logging in to the machines. Rule it out if your teams
troubleshoot their servers over SSH or automate with Ansible. This page helps
you decide, shows the creation of a local cluster, and explains what Talos
changes day to day.
What you will learn
- Recognize what sets Talos apart from a classic Linux distribution
- Compare Talos with kubeadm, Kubespray, RKE2 and k0s
- Choose Talos, or rule it out, based on your team, your fleet and the target Kubernetes version
- Spot what Talos changes for troubleshooting and upgrades
What does Talos Linux solve, and how does it work?
Talos solves configuration drift: on Ubuntu or Rocky servers, every SSH login, every installed package and every setting made by hand moves a machine a little further from its neighbors. Talos removes those possibilities. The root file system is read-only, there is no shell access, and a machine only changes state through a call to its API.
A Talos node is described by a YAML file, the machine config. It contains
the network, the disks, the node's role (control plane or worker) and the
kubelet options. You send it to the node with talosctl apply-config. Two
nodes that receive the same file are identical, and the file is versioned in
Git like any other code.
The talosctl client talks to the Talos API over gRPC, with mutual TLS
authentication. It distinguishes two notions: endpoints, which are the
addresses where the API can be reached (often a load balancer in production),
and nodes, which are the machines targeted by the command. To form a cluster,
a bootstrap command on the first control plane node initializes etcd, and the
other nodes join it.
Talos replaces the operating system, not Kubernetes: your manifests, your Helm charts and your tools are used as on a kubeadm cluster. What changes is how you administer the nodes.
When should you choose Talos, and when should you avoid it?
Choose Talos when three conditions are met: the fleet is new or being rebuilt
(bare metal, virtual machines or cloud), you want a reduced attack surface and
strictly identical nodes, and the team is ready to learn talosctl in place of
its Linux reflexes. The gain shows in security and in configuration drift. The
cost is human: the usual gestures disappear.
Rule Talos out in three situations. If your teams troubleshoot over SSH or deploy with Ansible, Kubespray keeps those methods. If you must install Kubernetes on existing Linux machines without reinstalling them, Talos does not fit: it replaces the system. If you want to learn Kubernetes or pass the CKA, kubeadm remains the right starting point.
The table compares the same administration task with kubeadm on a classic distribution and with Talos. The difference lies in day-to-day work, not in Kubernetes: that is where the choice is made.
| Task | kubeadm on Ubuntu or Rocky | Talos |
|---|---|---|
| Log in to a node | SSH | Impossible: no SSH or shell |
| Read the state of services | systemctl status kubelet | talosctl services |
| Read a log | journalctl -u kubelet | talosctl logs kubelet |
| Change a setting | Edit a file on the machine | Change the machine config, then talosctl apply-config |
| Upgrade the system | apt or dnf, then a reboot | talosctl upgrade with a system image |
| Upgrade Kubernetes | kubeadm upgrade node by node | talosctl upgrade-k8s |
The comparison with the other distributions is on the Kubespray page, and the RKE2 and k0s pages detail their own cases. The guide Deploy a Kubernetes cluster in production also covers managed services.
Which Kubernetes version does Talos install?
The latest release, Talos v1.14.6, dates from 11 November 2026. In the code of
that tag, the Kubernetes version installed by default is 1.37.1, and the
Talos 1.14 support matrix lists Kubernetes from 1.33 to 1.37. Checked on
2026-11-13 against the
v1.14.6 release, the
constants.go file of the tag and the
support matrix.
Talos is therefore not behind the version taught in this course, 1.37. The matrix still marks the end date of community support for 1.14 as to be confirmed (release of 1.15 planned for late 2026): check it when you choose.
What does a local Talos cluster look like?
The demonstration below follows the Quickstart of the official documentation. It is not replayed for this page and is not graded: it shows the shape of the procedure. The procedure for installing a cluster is practiced in the kubeadm lesson.
You need Docker, talosctl (the documentation gives its installation) and
kubectl. A single command creates the cluster: the docker subcommand tells
it to start the nodes as containers.
talosctl cluster create dockerkubectl get nodes -o widetalosctl cluster destroyThe documentation shows a two-node cluster, one control plane and one worker,
both Ready, whose OS-IMAGE column shows Talos followed by the version.
That is what you should look at: the system image is Talos, not a classic
distribution. The last command destroys the cluster and frees the resources.
How do you go from this demonstration to production?
A production cluster is prepared differently: you download an ISO image from the Talos image builder (Image Factory), boot three control plane machines and the workers from it, then apply their machine config to them. The Kubernetes API entry point must be a load balancer or a virtual address that reaches the three control plane nodes: without it, high availability does not exist. The High availability lesson explains this principle.
One network detail is surprising: the Talos API uses gRPC with mutual TLS on
port 50000. Behind a firewall, this is the port you must let through to the
nodes in order to drive them with talosctl. To manage a large number of Talos
clusters, the project also offers Omni, a separate service.
What does Talos change for operations?
Troubleshooting moves first. Without SSH, you read state and logs through
the API: talosctl services gives the state of each system service, and
talosctl logs returns their logs. Talos's services (etcd, kubelet) are not
Pods visible in kube-system: they belong to the system. Linux diagnostic
habits no longer help, and teams take a while to find their bearings again.
Upgrades then happen in two separate steps. talosctl upgrade updates a
node's system by telling it which image to use: Talos keeps the previous image
and rolls back if the new system does not boot, or on request with
talosctl rollback. This upgrade does not change the Kubernetes version. That
one is upgraded with
talosctl --nodes <control-plane-ip> upgrade-k8s --to <version>, and the
--dry-run option first shows what will be changed.
The system upgrade path follows adjacent minor versions: the migration is only tested between adjacent minors, so from 1.12 to 1.14 you go through the latest patch release of 1.13. The lesson Upgrade a Kubernetes cluster covers the kubeadm cluster case, for comparison.
Key takeaways
- Talos Linux is a system with no SSH, no shell and no package manager, driven by the
talosctlAPI. - Each node is described by a YAML machine config, versioned in Git and applied with
talosctl apply-config. - Talos replaces the operating system, not Kubernetes: manifests and charts are used unchanged.
- It suits a new fleet and a team ready to give up Linux reflexes; it does not suit a team that troubleshoots over SSH or automates with Ansible.
- v1.14.6 (11 November 2026) installs Kubernetes 1.37.1 by default and also supports earlier versions.
- The system and Kubernetes are upgraded separately, with
talosctl upgradeandtalosctl upgrade-k8s. - In production, the Kubernetes API goes through a load balancer, and port 50000 of the Talos API must stay reachable through the firewall.
- To practice installing a cluster, the kubeadm lesson remains the reference.
Test your knowledge
Seven questions on what makes Talos specific: the no-SSH model, the comparison with kubeadm, the Kubernetes version installed and the upgrade operations.
Knowledge check
Check what you have learned with this interactive quiz
How it works
- The timer starts when you click Start
- Multiple choice, true/false and short answer questions
- You can move between questions
- Detailed results are shown at the end
Starts the quiz and the timer
Review
(0/0)Score by domain
A domain groups several guides and gets several questions, so its score is more reliable than a per-topic score, which often rests on a single question.
Summary
Strengths
No topic passed at 80% on at least two questions yet.
Weak areas
No topic below the pass mark.
Recommended next
Hints before you try again?
Continue the course: RKE2: when to choose this hardened Kubernetes distribution
A new full quiz with random questions
Work again only on the questions you missed