Well, this has been the argument for quite sometime that you do not need Kubernetes for everything. Kubernetes is a beast and in its earlier days when things were not managed by the cloud providers, it was supposed to be glue fixed and installed + maintained on computes by the engineers which was not at al an easy task.
Although managed Kubernetes has made enough progress overt he past years to simplify the process but the question of do you really need Kubernetes? still remain.
I have seen various posts from Dhh on how they have only used docker and glued it together for their applications, Whole Hey runs on Kamal. In general the convention of Kubernetes is Hard is true, it is a beast system originated at Google for Google complexities to it really makes sense to use Kubernetes for large scale microservices applications. But When you have very small apps you can simply run on VM and it should just work.
NO!! This is not the end game, So Kubernetes is hard, cloud providers make it easy by providing managed Kubernetes and you can use and you can have a single node cluster or a two node cluster for smaller apps. So the convention that Kubernetes by nature is expensive depends on how cloud providers are providing you with their offering, for example AWS, GCP charges for control plane and worker nodes, yes in that case you are paying a lot for your small app. But providers like Civo where control plane charges are free and you only pay for worker nodes where your wok load is running is quite an awesome deal. This is because you will run the workload and pay and be able to use all the powerful features of Kubernetes.
So case by case basis depending on how you are running you app on Kubernetes the cost can vary from architecture to architecture and provider to provider.
All in all in my opinions with k3s, Talos and smaller footprint Kubernetes distributions it takes less resources for the cluster and if there are managed kubernetes offering around these then it makes sense to use Kubernetes powers for any scale apps. But if you are paying for control plane and the kubernetes components and other cloud components running on the worker nodes also consume lot of resources, your simple web app can be run elsewhere without Kubernetes.
Do comment on what is your view on when to not use Kubernetes?
Coming to the stuff I have been working on:
I have a few conferences coming up that I need to prepare for:
- Civo Navigate - I love Civo Navigate not because I am part of making this event come to life but because the awesome people I get to meet, sessions I get to attend and innovations I get to see. This is a dope event happening on 20th and 21st Feb in Austin. If you are around then lets meet. Dm me if you want discounted code for the ticket.
- WasmIO - This is one of the best conference I attended last year and I am even more excited to be at the second edition. This time I have multiple sessions, so it will be fun. Also what a timing to release course before the event. Let’s hang out and chat Wasm if you are in/around Barcelona on 14th and 15th March.
- KubeCon EU - I was almost about to skip KubeCon EU but since WASM I/O and KubeCon dates are so close that i can cover both the conferences with less distance I have to travel, so I will be there at Kubecon having couple of sessions and also speaking at the co located event - Kubernetes on Edge Day. Again, if you are coming to Paris and Kubecon, do not forget to say HI!
Sponsored Content
Without the sponsors I won’t be able to give you an authentic newsletter with all the cool stuff, so please do check them out
Awesome Reads
- The State of eBPF 2024 - This report covers the evolution of eBPF, the revolution it created, what’s being built with it today, challenges, and where it is heading.
- Creating an Automated Documentation Pipeline in PHP with Autodocs and GitHub Actions - It shows how to create a pipeline to automate the build of markdown documents using a demo project, Autodocs, with Minicli and GitHub Actions.
- Introduction to Zig, a Potential Heir to C - David Eastman explores Zig, a promising successor to C, emphasizing its low-level capabilities that bring developers closer to the hardware in a modern context. Zig offers a general-purpose programming language and toolchain designed for creating robust, optimal, and reusable software, with features like no garbage collection and direct memory management, aiming to simplify and speed up the development process for programmers working on hardware-near projects.
- Build a Slack Emoji Bot in 20 Lines of Code - A simple slack bot build with python and spin.
- Cloud-Computing in the Post-Serverless Era: Current Trends and beyond- It explores the evolution of serverless computing beyond FaaS towards a future where cloud services integrate seamlessly with developer needs, emphasizing the shift from infrastructure-focused to developer-centric cloud constructs. It highlights the transition towards hyperspecialized vertical multi-cloud services, the move from Infrastructure as Code to Composition as Code, and redefines microservices in terms of organizational boundaries, signaling a move towards more intuitive, efficient, and user-friendly cloud-service configurations for developers.
- Dapr Deployment Models - it goes through the evolution from a Kubernetes microservices sidecar to a serverless API. It discusses different deployment scenarios ranging from VMs and containers to Kubernetes, showcasing Dapr's capacity to meet diverse architectural and operational needs while simplifying the development process across different platforms.
Kubesimplify Updates
The Wasm course is going strong with close to 4k views and I request everyone reading this newsletter to watch, learn Wasm and share in their network as your support.
We have also written three new shiny blogs in past 2 weeks
- Highly available Kubernetes - how to create a HA kubernetes cluster with 3 controlplane., 3 etcd nodes and 2 worker nodes with kubeadm and haproxy. Also the steps to automate that.
- Pure Cilium : A Guide for Local Load Balancing and BGP - In this guide, we'll walk through the steps to build a multi-node Kubernetes cluster on your local workstation or MacBook (M1, M2, or M3) using K3s and Cilium. We will also demonstrate using Cilium's powerful Load Balancer (L.B.) Use the IPAM feature to expose your service as a built-in load balancer in your K8s cluster.
- Ksctl: Making Kubernetes Easy Across Clouds - Ksctl, short for Kubernetes Simplify Control, is a tool designed to simplify how we handle Kubernetes clusters, especially across different cloud services. This blog will walk you through what Ksctl is, the problems it tackles, its standout features, and a step-by-step guide on creating clusters effortlessly.
Resources and Repos
Do subscribe to my free newsletter to stay updated in cloud native and WebAssembly. Thats the least you can do ;)
DevOps Engineer @NIC | Cloud Native | DevRel | Founder @GrowInCommunity
1 年Thanks for this insightful newsletter! Looking forward to learning more about K8s from the Kubesimplify resources.
?? Platform Evangelist & Enabler, Software Delivery Optimization Expert
1 年Well. I pretty much laid it all out here: https://www.dhirubhai.net/posts/antonweiss_kubernetes-activity-7157409500650483712-S-oB?utm_source=share&utm_medium=member_desktop The comments are also very enlightening!
Technical Project Manager | IIT Roorkee AIML | Product Owner | Domain Expertise Fintech & Banking | PMP | Certified Agile Scrum Master
1 年?It was supposed to be glue fixed and installed + maintained on computes by the engineers which was not at all an easy task....!