Sascha Eglau

Every engineer that has extensively used Kubernetes can easily explain the Kubernetes Schedular. We all know it is the control plane component responsible for assigning newly created Pods to suitable Nodes within a cluster. But things will take a different turn when asked to explain how the scheduler assigns or schedules these pods to nodes. I was once in this state, so I decided to go down the rabbit hole to deeply understand the kubernetes schedular which I consider the beauty of the kubernetes control plane. If you want to understand the scheduling logic, phases and advanced Scheduling specifications of the Kubernetes schedular, then this article for you. https://lnkd.in/enqdbJJV

25

2 kommentarer

While building with different LLM workflows, I realized we are missing a universal packaging layer. Right now, we define agents using loose markdown files and scattered system prompts. That might work for a quick local test, but they don't version, they don't distribute natively, and they can't carry executable tools or persistent memory. We have Docker for apps, but what do we have for Agents? I decided to build the solution: Abbyfile. Abbyfile brings the Unix philosophy to AI: small, focused, composable. It takes your entire agent definition (system prompts, tools, and memory) and turns it into a versioned, distributable CLI binary. How it works: 🧠 The LLM (like Claude Code) is the brain. 🤖 The binary is the body (providing the instructions, tools, and memory via built-in MCP auto-discovery). The Workflow is just 3 commands: 1️⃣ agentfile build — Compiles a standalone binary from your YAML + Markdown definitions. No Go code required. 2️⃣ agentfile publish — Cross-compiles and creates a GitHub Release. 3️⃣ agentfile install — One command to download the right binary and wire it up via MCP. Done. By treating agents as real, versioned software artifacts, we move away from "vibes-based engineering" and fragile text updates. Your context overhead stays minimal, your tools are actually executable, and your agents can be shared and installed anywhere instantly. 🏗️ This is just the start. Check out the project, see the benchmarks, and let’s build the standard together. 👇 Link in comments #PromptAsCode #AI #SoftwareArchitecture #OpenSource #AIAgents #Abbyfile #MCP

16

3 kommentarer

Hey, fellow engineers. Does anyone here actually write Contract Tests? Here is the talk about Contract Testing of Websockets, which I made back in 2018 at the Nordic APIs in Stockholm. And back then, it was a hot topic; I was not even the only one speaking about it at the same event. But what I observed within the years since then, while working with different projects, is that this type of testing is still a luxury that not everyone can afford. Do you write Contract Tests with Pact or any other similar software in your company? Btw. the link to the full talk is in the first comment. #testing #QA #unittests #websockets #API

17

6 kommentarer

🌟 Day 6 of My Backend Journey: Understanding Middleware in Express.js! If you’re learning backend with me, today’s topic is one of the MOST important building blocks of Express — Middleware 👇 🧩 What is Middleware? Middleware functions are like checkpoints that your request has to pass through in an Express application. They can: ✅ Modify the request/response ✅ Run some logic ✅ Stop the request ✅ Or pass it forward to the next function Think of middleware as the security guard, data cleaner, logger, and traffic controller of your app! 🔧 Types of Middleware 1️⃣ Application-level middleware Runs on every request unless filtered. app.use((req, res, next) => { console.log("Request received"); next(); }); 2️⃣ Route-level middleware Works only for specific routes. app.get('/user', loggerMiddleware, (req, res) => { res.send("User data"); }); 3️⃣ Built-in middleware Like express.json(), express.urlencoded() 4️⃣ Third-party middleware Like cors, morgan, cookie-parser etc. 🚦 Why is Middleware Important? It helps you: ✔ Handle authentication ✔ Validate inputs ✔ Log requests ✔ Parse data ✔ Protect routes ✔ Handle errors smoothly Without middleware, an Express app would feel like traffic without signals 😂 🛠️ Common Use Case Example const auth = (req, res, next) => { if (req.headers.token) { next(); } else { res.status(401).send("Unauthorized"); } }; app.get('/dashboard', auth, (req, res) => { res.send("Welcome to Dashboard"); }); 😂 Today's Meme : "When you forget to call next() inside middleware…" 🏁 Outro Learning Express becomes super fun when you understand middleware — it's literally the brain of your app! See you tomorrow for Day 7 👋🔥 #ExpressJS#NodeJS#BackendDevelopment#WebDevelopment #JavaScript #LearnWithMe #CodingJourney#FullStackDeveloper #BackendEnginee #SoftwareEngineering#100DaysOfCode#TechLearning#APIDevelopment#DevelopersCommunity#CodeNewbie

4

Interesting article...I'd say software engineering involves understanding what people truly need, not just their initial requests. It's more dependent on context and often overlooked in discussions that focus too narrowly on tool selection rather than actual outcomes. Such as the way everyone is implementing LLMs to the MOON! without thoughtful integration and a gradual rollout. https://lnkd.in/gmMdmYi9

2

Is Terraform the Backbone of Modern Infrastructure as Code? If you work with cloud infrastructure, chances are you’ve come across Terraform. It’s one of the most widely used tools for managing environments. Everything starts with a few simple files: main.tf, variables.tf, provider.tf, outputs.tf, and more. The terraform state file usually stored securely in the cloud keeps track of what’s actually deployed. What really makes Terraform powerful is its cloud-agnostic nature. Whether you’re using AWS, Azure, or GCP, you can manage everything through a single workflow without being tied to one provider. Why It Matters? 1.Consistency Infrastructure is defined as code, no manual setup or “it works on my machine” issues. 2.Version Control All changes are tracked in Git for easy collaboration and rollback. 3.Scalability  Adjust resources or environments in minutes as business needs evolve. 4.Cloud Agnostic  One tool and syntax for multiple clouds. 5.Modularisation Terraform encourages reusable, composable modules meaning teams can define a standard building block (like a VPC, Database, or Kubernetes cluster) once and reuse it across multiple projects. This saves time, reduces duplication, and ensures every environment follows the same best practices something that’s harder to achieve with tools like CloudFormation or ARM templates, where reuse and portability are more limited. What to take? Whether you’re spinning up a new Kubernetes cluster, testing disaster recovery, onboarding a client environment, or rolling out a multi-cloud strategy. Terraform makes it simple, repeatable, and reliable. At the end of the day, whether it’s a small tweak or a full-scale deployment. Terraform just works. Do you agree? #DevOps #Learning #Terraform #Infrastructure #Consistency #VersionControl #Scalability #Resilience #CoderCo #SystemsThinking

12

If you're currently using Promtail for your Grafana Loki setup, you’ll want to read this just incase👇 🔔 Promtail is now deprecated and is on Long-Term Support (LTS) since February 13, 2025. No new feature updates, only critical bug and security fixes. Commercial support will end after the LTS phase, expected to run until February 28, 2026. End-of-Life (EOL) begins March 2, 2026, after which no further support or updates will be provided. All future development will shift to Grafana Alloy, the next-generation agent for collecting and shipping data. If you’re using Promtail today, it’s time to start planning your migration to Alloy. Source: https://lnkd.in/dAs-ssha #Grafana #Loki #Promtail #Observability #DevOps #CloudNative

11

2 kommentarer

Stockholm Engineers Quick heads-up: Equal IT has an upcoming Kubernetes role, and it’s not another “deploy and call it platform engineering” situation. It's working with a product-driven start-up that you may recognise. This one is for someone who’s actually run Kubernetes: building clusters, owning upgrades, handling networking and storage, and treating Kubernetes as an internal platform that product teams rely on. It’s on-prem/private cloud, infra-heavy, and very hands-on. If most of your experience is managed Kubernetes, this probably isn’t for you. We need those who have: • Run Kubernetes on-prem or in a private cloud • Owned cluster upgrades yourself • Worked deeply with networking & storage • Built platforms that other engineers depend on • Preferred substance over shiny tools More details soon, or feel free to DM if this sounds like you. Photo from the best spot in town.

12

5 kommentarer

AutoMQ: Diskless Kafka® on S3 https://lnkd.in/evJAjNAx Cost effective: The first true cloud-native streaming storage system, designed for optimal cost and efficiency on the cloud. Refer to this report to see how we cut Apache Kafka billing by 90% on the cloud. High Reliability: Leverage object storage service to achieve zero RPO, RTO in seconds and 99.999999999% durability. Zero Cross-AZ Traffic: By using cloud object storage as the priority storage solution, AutoMQ eliminates cross-AZ traffic costs on AWS and GCP. In traditional Kafka setups, over 80% of costs arise from cross-AZ traffic, including producer, consumer, and replication sides. Serverless: Auto Scaling: Monitor cluster metrics and automatically scale in/out to align with your workload, enabling a pay-as-you-go model. Scaling in seconds: The computing layer (broker) is stateless and can scale in/out within seconds, making AutoMQ a truly serverless solution. Infinite scalable: Utilize cloud object storage as the primary storage solution, eliminating concerns about storage capacity. Manage-less: The built-in auto-balancer component automatically schedules partitions and network traffic between brokers, eliminating manual partition reassignment. High performance: High throughput: Leverage pre-fetching, batch processing, and parallel technologies to maximize the capabilities of cloud object storage. Refer to the AutoMQ Performance White Paper to see how we achieve this. Low Latency: AutoMQ defaults to running on S3 directly, resulting in hundreds of milliseconds of latency. The enterprise version offers single-digit millisecond latency. Contact us for more details. Built-in Metrics Export: Natively export Prometheus and OpenTelemetry metrics, supporting both push and pull. Ditch inefficient JMX and monitor your cluster with modern tools. Refer to full metrics list provided by AutoMQ. 100% Kafka Compatible: Fully compatible with Apache Kafka, offering all features with greater cost-effectiveness and operational efficiency.

3

I recently watched an interview with Lars Hvam Petersen where he mentioned that if you have a tool that helps you in your daily work, you should just share it. It doesn't have to be perfect, and it doesn't need to cover every possible edge case—as long as it’s useful. So, I decided to make the ABAP Smartform Translator public. It allows you to manage labels dynamically via a database table (SM30,Business Configuration), eliminating hardcoding and the need for transport requests for simple text changes. You can check it out at the following link: https://lnkd.in/dchnJy3K #ABAP #OpenSource #SAP #Smartforms #GitHub #CleanCode

104

4 kommentarer

🚀 🔥 Helsinki (Hetzner) DuckDB Server with Postgres interface, try it out. It's BoilStream. I live in Helsinki, so I spun up a Hetzner instance in Helsinki and downloaded boilstream there and started it. ‣ It's Hetzner 16 vCPU 32GB RAM instance with 320GB local fast SSD. ‣ 💶 It costs 24€ per month, and the performance with DuckDB 🦆 is just awesome! ‣ 🚄 From may laptop, using DBeaver this time, end to end latency over WLAN is about 16ms. Not bad at all. ➡️ You can try it out yourself -- but if you fill up the disk, I blame You 😄 jdbc:postgresql://boilstream:boilstream@65.21.3.221:5432/memory 👉🏻 This is the free tier version you can download your self for free and create similar instance if you like. It has configured rate limiting per connection (data ingestion speed) and limited concurrent sessions. So, what can you do with it, like easily with only SQL? ✅ Stream data with e.g. DuckDB Airport through the server into cloud object storages with optimal Parquet files, including materialised views, i.e. derived topics. Like this: D load airport; D ATTACH 'boilstream' (TYPE AIRPORT, location 'grpc://65.21.3.221:50051/'); D insert into boilstream.s3.nyc select * from parquet_scan('*.parquet'); ‣ You can create more topics with plain CREATE TABLE onto the boilstream.s3 schema and they become visible on the FlightRPC interface the Airport extension uses. ‣ You can create realtime derived topics with CTAS (CREATE TABLE derived_topic AS SELECT ..) and they will land on object storage or physical DuckDB on-disk table in realtime. ✅ While you stream, run analytics e.g. with DBeaver, Power BI, or Tableou. In other words, you can investigate in realtime what data flows through. ✅ Or just use it as a DuckDB Cloud DWH, connect with your favourite BI Tool and don't worry about backups as the data lands on S3 already (the topics). ✅ You can add the Airport extension into BoilStream itself and write from e.g. DBeaver directly to the topics (or through FlightSQL interface). Pull data from various sources, APIs etc. with DuckDB httpfs extension and many other extensions and write to topics. Just familiar DuckDB SQL, 16ms latency end-to-end from BI Tool.

79

10 kommentarer

You can scale code in seconds with Kubernetes. Scaling trust takes years. As I reflect on the journey of building Testriq QA Lab global service provider, the biggest lesson I’ve learned has nothing to do with technology. It’s about consistency. When we started, it was easy to guarantee quality because I could personally review every major project. But when you grow to serving clients across E-commerce, Healthcare, and FinTech globally, "the CEO checking it" is no longer a strategy. The hardest part of scaling a service business isn't acquiring customers. It’s ensuring that a tester in our Mumbai lab approaches a problem with the exact same rigor, curiosity, and ownership as a tester in London or New York. If we are going to offer Trust Certifications to matrimonial apps (https://lnkd.in/d4PUPHsD) or trading platforms (https://lnkd.in/dHbWyS7v), we have to embody that trust internally first. We learned that you cannot write an SOP (Standard Operating Procedure) for "caring about the outcome." Quality isn't a department. It's a culture of constructive paranoia—a shared belief that if we miss something, it matters to a real human being on the other end of that screen. To my fellow leaders scaling service-based organizations: The process is important, but the people are everything. What is the biggest non-technical challenge you’ve faced while scaling your business globally? #Leadership #ScalingBusiness #CEOInsights #QualityCulture #TestriqQALab #Trust

4

What a great way to start presentstion - with working demo! And the one at the end blew my mind. Here’s sample from GraalWasn docs: Using GraalWasm, you can load WebAssembly modules in your application, access them from Java, and make them interoperate with other Graal languages (in this case Rust) https://lnkd.in/ejkCSJxx

7

The next Yocto Project LTS will use systemd as the default init manager. For a long time, the default in Yocto Project was sysvinit, the traditional approach based on shell scripts. In practice, many, if not all, recent configurations and distributions had already moved to systemd, even if it was not the default. At the OpenEmbedded FOSDEM BoF, a quick survey confirmed it. The large majority of users reported using systemd. Now the default has caught up with usage. Systemd is the new default. Commit link: https://lnkd.in/eZYsFcNn

182

23 kommentarer

💡 Why pull the same container image 20 times when you can pull it once? Meet Spegel - the Swedish word for "mirror" and a clever solution to K8s image distribution inefficiency. Instead of every StatefulSet replica triggering separate image pulls Spegel: Stateless cluster local OCI registry mirror. | Kubernetes Architect posted on the topic | LinkedIn , Spegel creates a peer-to-peer network where nodes share images locally. One pull = cluster-wide availability. ✨ The magic: Zero configuration caching, faster scaling, reduced bandwidth costs, and resilience against registry downtime. Now embedded by default in K3s/RKE2 Embedded Registry Mirror | K3s - because sometimes the best solutions are the simplest ones. Worth exploring: https://lnkd.in/gVaRzteG #Kubernetes #Efficiency #CloudCosts #DevOps

8