New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Why are you using Kubernetes ?
lowprofile
Member
Hello folks!
I would like to hear why you are using Kubernetes (Real case scenarios)
I might be very classic, but i still cant justify the usage of kubernetes due to its "complexity" in term of management.
I have approx 100 servers (web, db, app), and honestly i still prefer to manual spinup a VM and configure it manually. I feel more control and insight. Even scaling can be done in many ways (classic ways) So this discusssion is just about why you are using kubernetes ? Enlight me please. When is the right time to implement kubernetes in ones business model ?
Thanked by 1xms
Use of Kubernetes
- Are you using Kubernetes in your infrastructure?72 votes
- YES!31.94%
- Yes, but i dont know why. Sounds fancy and trending maybe?2.78%
- No. I prefer the old classic way.45.83%
- No, but time has come to reconsider....19.44%
Comments
it may be not fit for all business models, but if you handle a million requests, I think it can be a pain if you install from vanilla when you need hundred a servers in seconds.
Are you handling that many servers manually or use Ansible playbook?
I still use VPS but I've moved a lot of my workload over to Kubernetes simply because it's so easy to scale (automatically & manually), it's highly available out of the box and it's excellent for handling an application with lots of requests/load.
i don't use it (yet) but being able to quickly scale up without breaking the bank (while avoiding services like aws/vercel) and having highly-available, fault-tolerant clusters are the biggest reasons i want to get into it.
it's not that you cant achieve that using more 'vanilla' methods but I'm lazy and k8s sounds like the most straightforward method🤷
I am looking into Kubernetes and Docker myself, the reasons are many, and keep in mind you can still take advantage of each benefits using a virtual machine instance first. But Kubernetes is a container orchestration platform for managing clusters across what could be many machines. Hardware resources are, in theory abstracted at the infrastructure layer, while virtualization software and hypervisors like Hyper-V VMware or KVM allow for virtualized instances of physical hardware to run on the same machine. Obviously one major benefit aside from scalability and resource management is deployment. There are also benefit in management of backups and versioning, especially if considering the deployment of applications. When considering vanilla configuration and setup of a VM really we're bundling many different activities into the concept of a machine which ultimately serves a purpose in the traditional sense of insert_purpose server. I am very interested in replacing and more efficiently managing several of these activities through platform allocation cluster management and container deployment. I think the future will leverage the benefit of both technologies when using managed or unmanaged web and application services, we're talking about saving time and resources in the end not necessarily comparing features of either.
You don't need that many severs to handle millions of requests (likely one tbh) a day, even billions of requests a month is fine. Terraform and ansible is all you need unless you want to scale eslastically, even then, you don't need k8s.
Very very few people need hundreds of servers in seconds, if any. The pain point is always then going to be the database layer anyway.
K8s is fine, it's good, I'd always go for a managed service where you're just paying for the compute anyway like AKS. But unless you have the expertise to manage and maintain it, you're likely better off going for a different approach imo. Like simple terrraform & ansible or terraform and nomad, etc
If you really prefer doing every manually well that's weird to me. Besides, you are missing the point. Kubernetes is huge. It's not just an awesome way to manage infrastructure, it's also a huge ecosystem with a shit ton of functionality built on top of it. I am pretty sure that if you gave yourself the chance to peek over the complex "look" you would realize that it's not all that difficult. It's just different from traditional ways of doing things, and frankly a ton better. But if you stop your consideration a the "it's too complex" though, without exploring it, then a) you are missing out, b) you are not able to make a qualified conclusion about vs what you do know.
I've used both, there's a time a place for both. Especially if you have an existing infrastructure in place and have to weigh up migration vs business needs. If you're already containerised and have a devops team who can learn the ins and outs for edge cases etc, I'd lean towards it, especially for greenfield projects, but having taken part in a large migration, I'd be a little more cautious migrating without true experts in place.
Personal projects, ansible for me, albeit a bit more manual.
Edit: I know you have a bias towards k8s given the time you've invested in it, I do intend to use your project on hetzner some day. When I have time, which seems to be never at the minute
If it's just one or two servers then it's definitely overkill. But if it's a bunch of servers that cooperate together for a common purpose, then Kubernetes IMO makes things easier than doing everything manually or with basic automation. I have been working with Kubernetes for a long time so perhaps I am biased, but today it's really not as complex as many people think. You can pick up the basic on how to use it very quickly. The administration side is of course more complex, but there are so many managed Kubernetes services out there really making it so easy. And if you want to save money there are tools like mine that make it easy to set up a cluster in minute even with a provider that doesn't offer a managed service directly, limiting the amount of maintenance and setup you need to do yourself.
If people today think that Kubernetes is complex I wonder what they would think of it in the beginning
ppl who use k8s can you tell me how you handle upgrades?
even using manged services they need to upgraded frequently because these only support handful of versions at a time.
Managed is the way!! In a corporate environment at least, most of the benefits with less downsides
If you use a managed service like GKE you can either click a couple of buttons every month or two, or just let GKE do automatic upgrades if you keep the workloads up to date so they are compatible with new APIs. It's so easy that I really don't see where the difficulty with this is. it's not like you are maintaining etcd or things like that yourself.
If money is not an issue, managed is definitely the easiest option. You literally just need to look after your workloads and that's it especially if you use GKE. With other services you may have to do a few more things but I am pretty positive that the benefits always overcome the cons of using k8s over some more traditional approach.
And I speak as someone who has in the past managed clusters with thousands of nodes so I have seen the real complexity with administration. But if you use managed, all of that is something you don't need to worry about at all.
Personally I think k8s is over-engineering (means intentionly make things more complex), bringing more trouble than benefits.
Instead, you could use scripts to do most of the management for the many nodes case.
Flexible and controllable, and you can code whatever you want.
k8s is for the corporate environment,
it's not for the pragmatic result (conveniently managing), but for a job barrier (to protect the administartor's job from being replaced easily). I saw many meaningless case like people use it only to manage 2 ~ 4 nodes.
Not hard to understand, image that in a traditional (I mean the office culture) company if you make things too easy and simple, your boss will think what you do valueless and you're replaceable. And on macro view, simple job pulling down the salary of administartor engineers.
But,
truthfully saying, this kind of thing is a poison.
And maybe the theory behind this is the reason why cool things are mostly coming from hobby projects than enterprise projects.
If you ask whether you need Kubernetes or not, the answer is mostly not. You'll know when it's time to use it.
Kubernetes is complex, kinda overkill for most of times, and challenging especially is when upgrading. Using managed like GKE is really helpful here.
Where Kubernetes is really shine is to manage large scale workload with dozen to hundred of thousands request per second, with hundreds of services. Or with workloads that might have bursty pattern. The autoscaling (especially when using managed like GKE) is so simple and fully automated. You can scaling from few nodes to hundreds node in 1-2 minutes.
Another use case is if you need doing complex operation like blue-green/canary deployment, traffic engineering etc.
To what extent have you actually used it to state that it brings "more trouble than benefits"? Or are you just echoing comments you have heard?
Are you... seriously comparing "scripts" with a platform like Kubernetes? Wow.
With a few nodes it may be overkill depending on the case. If you just use a managed service or a tool like mine that sets up a cluster for you in 2 minutes, then I'd still prefer using Kubernetes if I have more than a couple of nodes that work together e.g. for the same app.
I'm using k3s on my VPS'es. It's just a great way to manage infra/applications IMO. The VPS basically only has iptables and k3s installed, everything else is managed true k3s with yamls stored in version control. Makes moving to a different VPS a lot easier.
Exactly. We’ll reach a million requests over a 24 hour period. Not all the time, but sometimes.
Mostly for database actions, viewing often cached.
I switch up my approach every now and then, not gonna lie, but right now a Netcup root server does just fine.
Simple, yet enough.
People like adding layer of complexity where it’s simply not needed.
Although, for learning purposes, I have no issues with that.
I never echo others opinion, what's more is that I never easily effected by other's speaking.
. And I also added "Personally", because I thought most people like it instead of hate it, I'm the minority side in my opinion.
The hate to k8s comes from my own brain and nobody can change it
Yes, seriously. At least I can code to fulfill anything I want without using this so-called meaningless "platform".
Any "platform" is built by small scripts. Don't look down upon scripts. And those who write scripts maybe exactly the authors and creators of many platforms and frameworks you like.
I feel proud of being part of them as a creator.
I really don't like this kind of declaration platform.But I have got a need to scale easily and migration easily and reboot easily.My last project failed because after a reboot I can't recall how can I start it on a dedi-server:'(
what is kubernetes actually?
Look at the job market, you might be surprised how much the salary for a Kubernetes pro is going for. I'm just saying, you got to keep up with the times
I can tell you that I receive a shit ton of requests for calls etc from recruiters all the time because my profile is senior dev + kubernetes and devops + security. But even just good knowledge with Kubernetes can land very well paid jobs.
If you use Kubernetes as much as possible stay with managed services.
At $work we use multiple kubernetes cluster per project, its easier to maintan because we have control almost on everything compare to on prem, which managed by different team depend on the domain, like DB, Linux/Windows/Mac
Debugging on Kubenetes is harder than debuggin on VM, but you’ll used to it once you spend sometime thinkering with it.
For personal project use whatever technology you’re familiar with.