Shulou(Shulou.com)06/02 Report--

Source: Rancher Labs

Since the k3s came out, many friends in the community have asked the question "What is the difference between k3s and K8s besides the middle number?" How should I choose between the two? "。This article will briefly explain the difference between the two.

What is Kubernetes?

As you know, Kubernetes is a "container orchestration platform," meaning you can choose from a set of machines to run the containers you need.

It also handles things like upgrading your containers, so if you publish a new version of your site, it gradually uses the new version to launch the container and drops the old version, a process that takes only one to two minutes.

So, what exactly is K8s?

K8s is an abbreviation for Kubernetes, because there are eight letters between K and s, so it is called K8s. Often, however, whether people are talking about Kubernetes or K8s, they are talking about native upstream Kubernetes, a truly highly available and scalable platform designed by Google.

The problem is that while you can run Kubernetes on your local computer using tools such as Minikube, if you want to run it in a production environment, you will quickly get some "best practices" advice, such as:

Separate your nodes from the master, use your master to run the control plane, use your nodes to run the workload, and the two will never meet

Run etcd on a separate cluster to ensure it can handle the load

Ideally, Ingress() nodes are separated so that they can easily handle incoming traffic, even if some of the underlying nodes are already busy

Soon, you will have 3 times as many K8S masters, 3 times as many etcds, 2 times as many Ingress, and your nodes. So before you get to the stage where you need to ask,"How many nodes do I need for my site," you actually have at least eight medium-sized instances.

Don't get me wrong, I'm not accusing the advice of being bad. On the contrary, if you are running a production workload, these suggestions are very sensible. After all, there's nothing worse than debugging an overloaded shutdown production cluster on a Friday night!

But if you just want to learn Kubernetes, or host a development/staging cluster for some non-core apps, then adopting the above advice is a bit of a "kill the chicken," isn't it? At least for me. If I just want to launch the cluster to see if my Kubernetes manifest (deployment configuration, etc.) is correct, I don't want to pay hundreds of dollars a month for it.

What are the advantages of K3S?

Rancher Labs is the industry's leading container software provider and its flagship product Rancher is an open source, enterprise-class Kubernetes management platform that excels at managing and installing Kubernetes clusters. They have released a range of products that make up their ecosystem, such as Longhorn, a lightweight and reliable containerized distributed block storage solution that can be used in Kubernetes and has recently been included in the CNCF sandbox project. Rancher Labs is also the creator of k3s, a lightweight Kubernetes distribution.

k3s packs everything you need to install Kubernetes into a binary file that is only 60MB in size and fully implements the Kubernetes API. To reduce the memory required to run Kubernetes, Rancher removed many unnecessary drivers and replaced them with add-ons.

k3s is a fully CNCF-certified Kubernetes distribution, which means you can write YAML to work with the full Kubernetes version, and they will work with k3s clusters as well.

Since it requires very low resources to run, it can run clusters on any device with more than 512MB of RAM, in other words, we can have pods running on master and nodes.

Of course, since it's a small binary file, we can install it in a fraction of the time it takes to start a regular Kubernetes cluster. It usually takes less than 2 minutes to start a k3s cluster with several nodes, meaning you can deploy applications for learning or testing whenever you get the chance.

That sounds good, but what about it?

When people think of Kubernetes, they think of containers automatically starting on other nodes if a node dies, Load Balancer between containers, isolation and rolling deployment, all of which are the same between the full version of Kubernetes and k3s.

However, k3s are not always good, otherwise everyone would use k3s. So why aren't some people using k3s?

First of all, the current version of k3s (k3s v0.8.1) can only run a single master, which means that if your master goes down, then you can't manage your cluster, even if there is already a cluster that needs to continue running. But multi-master mode is already an experimental feature in k3s v0.10 and may be GA capable in the next release.

Second, in k3s for a single master, the default data store is SQLite, which is friendly for small databases, but SQLite becomes the main pain point if hit hard. However, the changes that occur in the Kubernetes control plane are more related to frequently updated deployments, scheduling pods, etc., so the database does not cause too much load for small development/test clusters.

conclusion

K8s and k3s have their own advantages and disadvantages, and their use scenarios are also different, so they cannot be generalized. If you're doing a large cluster deployment, I recommend K8s; if you're in a small deployment scenario like edge computing or just need to deploy some non-core clusters for development/testing, K3s is the more cost-effective option.

Give it a try!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.