Do I Need Kubernetes?(doineedkubernetes.com)
doineedkubernetes.com
Do I Need Kubernetes?
https://doineedkubernetes.com/
16 comments
If you’re running containers you’ll want some kind of container management - but I’ve been fine with ECS for running on AWS. It’s well integrated, standard enough, configurable via OpenTofu/Terraform/CloudFormation etc and free for container hosts running in AWS
Do I need it? No. Do I want to use it because of the superior tooling? Yes.
Honestly. Give yourself a month to learn it and migrate your home lab and it's actually not too difficult. Yes it takes a lot of debugging initially, but like most things, once you're familiar, it becomes easy.
Honestly. Give yourself a month to learn it and migrate your home lab and it's actually not too difficult. Yes it takes a lot of debugging initially, but like most things, once you're familiar, it becomes easy.
The fact that another engineer stands a chance of being able to see my cluster & have some basic idea how it works—that a k8s system self describes itself fairly well— is a superpower. Even if it remains yours and only yours, having a common reference frame to discuss your computing adventures in, being able to break out the great pieces others have made: it's so much more compelling to me than cobbling together your one off special pet of a system.
What you’re arguing is in favour of standards.
And that’s an undeniable feature of Kubernetes: people know about it, and smart people use the same (CNCF) building blocks, to keep maximising the probability that your next Kubernetes engineer has also knows about it.
But do you want to turn your software system distributed at the cost of immense complexity?
Only if you’re already cultivating something immensely complex, and you’d be better off with a common language.
And that’s an undeniable feature of Kubernetes: people know about it, and smart people use the same (CNCF) building blocks, to keep maximising the probability that your next Kubernetes engineer has also knows about it.
But do you want to turn your software system distributed at the cost of immense complexity?
Only if you’re already cultivating something immensely complex, and you’d be better off with a common language.
I initiated and led the Kubernetes adoption at the startup I'm with currently. Handed it off to the new ops team recently. The company is dramatically better off now in terms of scaling, deployment automation, configuration management, service discovery, internal service interactions, stability, reliability... The list goes on.
If your needs are met by a VM or three then sure, you may not need Kubernetes, although as other comments have pointed out, distributions like k3s can be useful even in those environments. But as you climb the scale and complexity ladder, there soon comes a point where it's very hard to beat Kubernetes, which is why it has become so widespread.
If your needs are met by a VM or three then sure, you may not need Kubernetes, although as other comments have pointed out, distributions like k3s can be useful even in those environments. But as you climb the scale and complexity ladder, there soon comes a point where it's very hard to beat Kubernetes, which is why it has become so widespread.
I did the same with managed services and they're in a better place than they would have been with k8s - which is where they started and had to be rescued from... It had become an inescapable gravity well of pointless busy work. It isn't about the tooling, it's knowing the business.
Yes, that works if the system is simple and standard enough to be handled by a managed service provider.
> It isn't about the tooling, it's knowing the business.
The sort of statement I expect to see on LinkedIn. Once you know the business, you decide on appropriate tooling. As I said, as you climb the scale and complexity ladder, Kubernetes starts to make more sense. At that point, the tooling can become very important.
> It isn't about the tooling, it's knowing the business.
The sort of statement I expect to see on LinkedIn. Once you know the business, you decide on appropriate tooling. As I said, as you climb the scale and complexity ladder, Kubernetes starts to make more sense. At that point, the tooling can become very important.
I use to believe that too until someone rightfully reminded me of all of the ugly solutions I’ve used over the years to deploy to VMs. Then I became a big believer in always containerizing.
This is the insight. You'll a worse job even for something as certificate management with letsencrypt. Let alone all the other stuff. If your workload doesn't need that, and you are fine with some downtime, then and only then don't do it.
When I see questions like this, I wonder what the asker does instead to cobble together all that K8s does for you that you’d have to figure out. I spent years refining those ops processes. And when k8s came along and I got over the initial learning curve, it felt like I’d entered the promised land. All operational tasks that were such a pain all ended up being easier in k8s. I’ve put a small start up on k8s because of this.
Congrats. Unfortunatly, not every usecase in the world is your usecase.
k3s is dead simple to install. Install rancher on top if you like a nice web GUI, or don't if you are fine with kubectl. The software you want to run on k8s probably has helm charts which make it easy to run.
I use Rancher for a hosted Kubernetes cluster on top of dozens of dedicated servers, and so far, it has been super nice. What are the alternatives for CI/CD for a small team (30)?
You don't need kubernetes in the same way you don't need a fast internet service. Sure you can work with dial up, but working with 1gbps internet line is just so much more convenient
You don't need kubernetes in the same way a small one man foodtruck doesn't need a McDonalds scale logistical backoffice.
https://news.ycombinator.com/item?id=735568