George Fleury

🚀 New in Kubernetes v1.33: Smarter, Region-Aware Traffic with PreferClose 🌍 Kubernetes v1.33 just made it easier than ever to optimize service traffic with topology-aware routing, now supporting trafficDistribution: PreferClose as a stable feature. 🎉 🔎 What’s the deal? With PreferClose, Kubernetes will prefer sending traffic to the closest (same zone) healthy endpoints, but gracefully fallback to others if needed. 📌 Why it matters ✅ Reduces cross-zone data transfer (and $$$) ✅ Improves latency ✅ Requires no complex service mesh setup 🛠️ Just add trafficDistribution: PreferClose to your Service and make sure your cluster is on v1.33+ with ServiceTopology enabled. Learn More: https://lnkd.in/e7YYUtTD #Kubernetes #K8s #CloudNative #PlatformEngineering #DevOps #GKE #EKS #AKS #Networking

914

23 Comments

Hiring in DevOps is broken Ever seen a job post asking for 10 years of Kubernetes experience when Kubernetes itself is just about that old? That is DevOps hiring today. Everyone wants "experts" but forgets that DevOps is a mindset, not just a toolset. They list 20 tools and expect one person to be a master of all. They want deep cloud knowledge, infra as code, security, CI/CD, monitoring, and incident response. Basically, an entire platform team in one person. Interviews are another story. Coding challenges that have nothing to do with real world work. Whiteboarding algorithms when the real job is about understanding systems, troubleshooting, and automation. Then comes the irony. Companies want engineers who think in automation and self service but run hiring manually like it is the early 2000s. My view? Hire for problem solving. Look for practical experience. Assess real world skills. And stop treating DevOps like a shopping list of tools. What do you think? Have you seen this too?

737

60 Comments

Kubernetes is free. Terraform is free. Docker is free. Prometheus, Grafana, Helm, Vault, GitHub is also free. Cloud free tiers exist. Docs are everywhere. Tutorials, examples, open source repos are unlimited. What’s not free? Execution. Stop over-planning. Stop saving posts. Stop waiting for the “perfect roadmap”. Open your terminal. Break something. Fix it. Build again. No permission needed. Just ship.

351

7 Comments

5 Levels of Kubernetes. - level 1 Knows basic concepts. Limited troubleshooting skill. - level 2 Managing deployments, namespaces & scaling apps. Troubleshoots issues with pods and deployments. - level 3 Deep understanding of Kubernetes. Automate deployments. Debug advanced issues. - level 4 Optimizes cluster performance. Follow Kubernetes security best practices. Designs & implements secure and scalable clusters. - level 5 Ability to say - "We don't need Kubernetes for this problem." So, what do you think? Am I missing anything? 🔁 Do a repost if this is helpful.

256

24 Comments

🚀 New blog post out now: GitHub Actions + AWS CodeBuild – The Ultimate Guide for Container Nerds 🐳🔥 Ever tried switching from GitHub-managed runners to AWS CodeBuild and thought: "This should be easy..." It is — kind of. But there are gotchas. In this post, I cover: ✅ What really happens when you run GitHub Actions on CodeBuild ✅ How runs-on can override your carefully crafted project setup ✅ Why your container image might break everything (or eat your HOME) ✅ And how a single repo can spawn a 72xlarge instance if you're not careful 😅 If you're deep into CI/CD, containers, or just like weird edge cases — this post is for you. 👉 Read it here: https://lnkd.in/d9UaM3jg 👇 Have you hit any wild GitHub Actions behaviors with CodeBuild or containers? Drop your story in the comments. #GitHubActions #AWS #CodeBuild #Containers #DevOps #CICD #CloudComputing

50

3 Comments

𝗞𝘂𝗯𝗲𝗿𝗻𝗲𝘁𝗲𝘀 1.34: 𝗣𝗲𝗿 𝗰𝗼𝗻𝘁𝗮𝗶𝗻𝗲𝗿 𝗿𝗲𝘀𝘁𝗮𝗿𝘁 𝗽𝗼𝗹𝗶𝗰𝗶𝗲𝘀 𝗮𝗿𝗲 𝗵𝗲𝗿𝗲 A long-standing limitation in Kubernetes has been removed. Restart behavior is no longer tied to the Pod level. With Kubernetes 1.34, each container can now have its own restart policy, and rules can be applied to specific exit codes. The new model is far more flexible: • 𝗚𝗿𝗮𝗻𝘂𝗹𝗮𝗿𝗶𝘁𝘆: Policies like Always, Never, or OnFailure can now be set per container. • 𝗖𝗼𝗻𝘁𝗿𝗼𝗹: Exit-code rules allow containers to retry recoverable failures while failing fast on critical errors. • 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆: Restarts happen in place, cutting recovery from up to a minute down to seconds, while preserving volumes and networking. This matters in practice. A migration container can now run once and never restart. A machine learning job can retry GPU memory errors without being rescheduled. A microservice with logging and monitoring sidecars can give each one restart logic suited to its role. The feature is still in alpha and gated, but it is worth testing. It reduces complexity, improves reliability, and reflects the way modern workloads actually fail. I break down examples, exit-code design patterns, and migration guidance in the full write-up: https://lnkd.in/enXiUvjU #Kubernetes #CloudNative #DevOps #MLOps #PlatformEngineering

7

🚀 Crossplane: The Game-Changer for Infrastructure Management Ever wished you could manage your cloud infrastructure the same way you manage your Kubernetes workloads? Enter Crossplane! This open-source Kubernetes add-on transforms how we think about Infrastructure as Code by bringing multi-cloud management directly into the Kubernetes ecosystem. What makes Crossplane special? ✅ Kubernetes-Native - Use familiar kubectl commands and YAML manifests ✅ Multi-Cloud Ready - AWS, Azure, GCP... all from one place ✅ Declarative - Define your desired state, let Crossplane handle the rest ✅ Extensible - Add new providers and resources as needed The magic happens when: Your infrastructure becomes just another Kubernetes resource Teams use the same tools for apps AND infrastructure Multi-cloud deployments become as simple as applying a YAML Unlike traditional tools like Terraform, Crossplane lives inside your K8s cluster, eliminating the need for separate state management and bringing true GitOps to infrastructure. 🎥 Want to see it in action? Check out this excellent technical deep-dive by Marcel Dempers (Marcel Dempers) - his explanations make complex concepts crystal clear! [https://lnkd.in/dapX2fuR] Thanks Marcel for creating such valuable content for the community! 🙏 #Crossplane #Kubernetes #InfrastructureAsCode #MultiCloud #DevOps #CloudNative

14

1 Comment

📊 Prometheus Monitoring Guide Everything you need to master Prometheus: ✅ Install with Docker, Helm & Operators ✅ Use exporters for MongoDB, PostgreSQL, Nginx ✅ Instrument apps in Go, Java, Python, NodeJS ✅ Monitor Golden Signals (Latency, Traffic, Errors, Saturation) ✅ Scale with federation, remote-write & Thanos/Cortex 💡 Perfect for DevOps engineers, SREs & Kubernetes teams. 📌 Save it. Share it. Use it for production-ready monitoring. Follow me for more DevOps & Cloud-native guides. #Prometheus #Monitoring #DevOps #Kubernetes #AWS #Azure #GCP #Observability #CloudComputing #Grafana #Sysdig #SRE #CloudNative #FollowMeForMore

147

4 Comments

Thinking about Kubernetes probes and their usage? 🤔 Try to simplify each probe to a basic question: Liveness Probe: Is this container still functioning and should it continue running? Readiness Probe: Should this pod be taking ingress/traffic? Startup Probe: Has this container finished its initialization and is it ready to be monitored by other probes? If the probe answers its key question correctly with a reasonable cost, then you have a good probe. 👍 The acceptable cost and complexity depends on the probe type: Liveness probes should be kept simple and fast since false positives cause disruptive restarts. Readiness probes can be more thorough since failure only affects traffic routing. Startup probes can be the most complex since they only run once during initialization. The key is matching the probe type to what you're actually trying to detect and what action you want Kubernetes to take when it fails. As developers we know the state our application should be in to provide the operations with the information it needs to do a good job. #IWorkForGM #developers #operations

7

𝐓𝐡𝐞 𝐬𝐦𝐚𝐫𝐭𝐞𝐬𝐭 𝐭𝐞𝐚𝐦𝐬 𝐝𝐨𝐧’𝐭 𝐣𝐮𝐬𝐭 𝐮𝐬𝐞 𝐊𝐮𝐛𝐞𝐫𝐧𝐞𝐭𝐞𝐬 𝐭𝐨 𝐫𝐮𝐧 𝐜𝐨𝐧𝐭𝐚𝐢𝐧𝐞𝐫𝐬. They use it to control everything—from clusters to databases. That’s exactly what Thomas Güttler explored at this year’s Chemnitzer Linux-Tage. Everyone knows Kubernetes as the go-to for container orchestration. But what really makes Kubernetes brilliant is how it works—and how its core ideas have reshaped infrastructure. At its heart, Kubernetes is built on two powerful principles: – A declarative API (you describe the desired state) – A reactive system (controllers constantly work to keep reality in sync) That combination makes Kubernetes so much more than a tool for running apps. In his talk, Thomas breaks down: 🔍 What’s actually needed to run Kubernetes (beyond just spinning up a cluster) 🔍 Why the controller pattern is so powerful and flexible 🔍 The shift from Infrastructure-as-Code to Infrastructure-as-Software 🔍 How Kubernetes' own architecture is now being used to manage Kubernetes itself A few years ago, we at Syself started the cluster-api-provider-hetzner (CAPH) open-source project to bring Cluster API support to Hetzner. Not long after, Thomas began contributing to the project. Fast-forward to today, and he’s part of our team, still pushing CAPH forward. It’s a great reminder of how open source doesn’t just build software—it builds teams. The best part of this talk? It shows the bigger picture. Kubernetes started as a tool to manage containers. Now it’s a platform for managing everything—including itself. The same controller pattern that orchestrates your apps is now managing Kubernetes clusters, and even running databases and other critical systems in a self-healing, declarative way. This is the side of Kubernetes I wish more people talked about. Because it’s where the real transformation is happening. The talk is in German 🇩🇪 (sorry, international friends — we have other great talks in English too) ▶️ Watch here: https://lnkd.in/eaa_2gPU

31

5 Comments

Control Plane and Data Plane are often confusing and used in various contexts. Let’s break it down in Kubernetes terms: → Control Plane: The Brain of Kubernetes The control plane is where all the big decisions happen. It manages cluster state, schedules workloads, and enforces policies. Core components like the API server, scheduler, controllers, and etcd work together to keep your cluster running smoothly. Every time you use kubectl, you’re talking to the control plane to set rules or check on your cluster. → Data Plane: The Muscle of Kubernetes The data plane is all about execution. Worker nodes, along with their pods and containers, carry out the instructions from the control plane. Each node runs a kubelet (to ensure the desired state) and kube-proxy (to handle networking), making sure your applications actually run and traffic flows as intended. In short, Control plane = policy, scheduling, cluster management Data plane = running workloads, execution as well as load management By separating decision-making (control plane) from execution (data plane), Kubernetes can scale massively and recover quickly from failures. Visit this blog from Kong to learn more: https://lnkd.in/eRQzxCyf

25

Lifecycle arguments help control the flow of your Terraform operations by creating custom rules for resource creation and destruction. Instead of Terraform managing operations in the built-in dependency graph, lifecycle arguments help minimize potential downtime based on your resource needs as well as protect specific resources from changing or impacting infrastructure. https://lnkd.in/gwUSYC2u #terraform #iac #devops #devsecops #awscloud #azurecloud #googlecloud #opensource #githubactions #github #jenkins

10

 𝐓𝐞𝐫𝐫𝐚𝐟𝐨𝐫𝐦 𝐨𝐫 𝐎𝐩𝐞𝐧𝐓𝐨𝐟𝐮 𝐢𝐧 2025—𝐖𝐡𝐚𝐭’𝐬 𝐑𝐢𝐠𝐡𝐭 𝐟𝐨𝐫 𝐘𝐨𝐮? Since Terraform moved to a BSL license, the open-source world answered with OpenTofu, backed by the Linux Foundation. Both are great for Infrastructure as Code, but they follow different paths. 🔹 Terraform offers enterprise stability and official support. 🔹 OpenTofu stands for open-source values, community input, and vendor-neutral governance. Whether you're in DevOps, SRE, or cloud, choose what best fits your team’s future. 👇 See the comparison chart below to decide.

75

1 Comment