Patrik Polyák

Hey all, Check out my initial experiences with OpenClaw! I write about set up experiences with Virtual Machines of OpenNebula Systems on Canonical's Ubuntu, and the skill I developed for my agent so it can have a software engineering playground. See the full post here: https://lnkd.in/ejbhFC87

9

How do you make sure that your site survives a sudden increase in traffic; for example in a festival ticket sale? We worked this on together with zdrei.com for a shared customer, and wrote up our learnings in the mittwald blog. The case in question is a TYPO3 site of a major german music festival, who wanted to make sure that their site remained available during the sudden traffic spike when 50k people want to buy their festival tickets at exactly the same time. The article covers how to build realistic load test scenarios with Grafana k6, why throwing ungodly amounts of hardware at problems is often not a solution, and how (as so often) smart caching ended up being the hero of the day. Some specific learnings: 1️⃣ Browser-based tests can measure more appropriately how the performance of a site feels for the end user, but end up eating a lot of RAM on your testing setup. Hybrid load tests, in which you combine browser-based with headless tests, are the way to go. 2️⃣ Core Web Vitals (like the Largest Contentful Paint) make for much more meaningful Service-Level Objectives than raw response times. 3️⃣ A single misconfigured Apache process limit can tank your server faster than any hardware bottleneck. Read the full thing here (in german): https://lnkd.in/evVmZ6X8

20

Building custom agents for GitHub Copilot, or any AI assistant, sounds fun until you hit the reality of testing them In this post, I will outline some of the difficulties faced during the development and testing phases. Additionally, I will share a more effective approach to developing and testing your agents locally. For a deeper dive into this topic, check out the full blog-post here: https://lnkd.in/e7qGERrE

66

4 kommentarer

Most Terraform pipelines are fine… until you have 10+ environments. Then it becomes: hardcoded folder lists 40 separate workflows “why didn’t prod run?” surprises reviewers drowning in plan outputs teams stepping on each other’s state So here’s the pattern I’ve seen work really well: A single GitHub Actions pipeline that scales to any number of Terraform environments — automatically. No updating YAML every time you add a new folder. The idea (simple) Terraform repo grows → pipeline should adapt automatically. So instead of listing environments manually, the workflow: detects which environment folders changed in the PR ignores module-only changes runs fmt/init/validate/plan per environment in parallel (matrix) runs security scanning (ex: Checkov) posts clean PR comments with per-env results on merge → plans + stores artifacts requires manual approval only for prod applies exactly what was approved (plan artifact → apply) Why this is a big deal Because this solves the real Terraform pain: Branch protection + dynamic matrix jobs don’t play nicely. So the trick is adding a final “check” job that aggregates results → one clean status check for GitHub protection rules. That one detail makes the whole system scalable. The workflow split that makes it clean 1) CI (on PR): detect changed envs run checks + plan comment results back to PR final “check” job = pass/fail for branch protection 2) CD (on merge): detect changed envs plan + upload artifact if env is prod/org → pause for approval download same plan artifact apply That “plan artifact” part matters. It prevents the classic drift problem: “what we applied is not what we reviewed.” If you’re doing multi-env Terraform… Here’s my question: Do you still maintain an environment list inside your pipeline YAML? Or are you auto-detecting changes? If you’re scaling Terraform across teams, this is one of those upgrades that instantly reduces incidents + pipeline babysitting. Drop your repo structure below (even anonymized) — I’ll tell you whether this approach fits. #Terraform #GitHubActions #DevOps #PlatformEngineering #IaC #CloudEngineering #Security #DevSecOps #SRE #Automation #SiteReliabilityEngineer #DevOpsEngineer # #Observability Beacon Hill TEKsystems

5

GitHub Copilot has officially transitioned to Usage-Based Billing as of today. Token management is now a priority for many teams, but this is just the beginning. The true challenge lies in agent quality; if your agent fails, your token strategy is compromised. Join us for a session focused on both Agent Quality and Token Optimization on 4 June from 16:00 to 17:00 CEST. We will discuss: - The connection between tokens and agent quality, highlighting their interdependence - How LLMs, harnesses, and the context window influence agent behavior - The controls available to enhance agent quality, optimize token usage, and minimize costs Register here: https://lnkd.in/eiTcDHJb

32

1 kommentar

GitHub Copilot CLI is officially GA — and this one’s a big deal for agentic development. 🤖 You can now plan, build, test, and review code directly from the terminal, with Copilot acting as a real coding agent across multi‑step workflows. No context switching. No broken momentum. This is AI showing up where developers already work — and actually helping them move faster. More here👇 https://lnkd.in/enjNHXXx

50

1 kommentar

Your Kubernetes YAML probably survived three migrations… and two interns. And it shows. Today’s Kube Today: 🧹 KubeTidy — Clean, merge, and optimize your Kubernetes configurations without breaking everything. Declarative doesn’t mean chaotic. 🗄️ MariaDB Operator — Manage MariaDB declaratively using Kubernetes CRDs instead of imperative command spaghetti. Define it once. Let Kubernetes handle the lifecycle. (Links in the comments 👇)

7

4 kommentarer

𝐈𝐭’𝐬 𝐅𝐫𝐢𝐝𝐚𝐲, 𝐚𝐧𝐝 𝐈’𝐦 𝐡𝐞𝐚𝐝𝐢𝐧𝐠 𝐛𝐚𝐜𝐤 𝐡𝐨𝐦𝐞 𝐭𝐨 𝐆𝐨𝐭𝐡𝐞𝐧𝐛𝐮𝐫𝐠 𝐚𝐟𝐭𝐞𝐫 𝐚 𝐛𝐮𝐬𝐲 𝐰𝐞𝐞𝐤 𝐢𝐧 𝐒𝐭𝐨𝐜𝐤𝐡𝐨𝐥𝐦. Customer meetings, old and new colleagues, and a lot of good discussions about the year ahead. One insight keeps coming back: 𝐌𝐨𝐬𝐭 𝐨𝐫𝐠𝐚𝐧𝐢𝐳𝐚𝐭𝐢𝐨𝐧𝐬 𝐝𝐨𝐧’𝐭 𝐡𝐚𝐯𝐞 𝐚 𝐭𝐨𝐨𝐥𝐢𝐧𝐠 𝐩𝐫𝐨𝐛𝐥𝐞𝐦. 𝐓𝐡𝐞𝐲 𝐡𝐚𝐯𝐞 𝐚 𝐠𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞-𝐚𝐭-𝐬𝐜𝐚𝐥𝐞 𝐩𝐫𝐨𝐛𝐥𝐞𝐦. Many teams already use world-class tools like GitLab, GitHub, Atlassian, and Microsoft Azure. Yet governance, risk, and compliance still feel fragmented — especially when regulations like 𝐍𝐈𝐒𝟐 and 𝐃𝐎𝐑𝐀 enter the picture. The result we see too often: • Siloed data • Manual audit evidence collection • Compliance work slowing down delivery • Increased security risk At Eficode, our consultants in Scandinavia help customers tackle this by connecting 𝐬𝐩𝐞𝐞𝐝 𝐚𝐧𝐝 𝐠𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞, not trading one for the other. A few examples of how: 1️⃣ 𝐀𝐮𝐭𝐨𝐦𝐚𝐭𝐞𝐝 𝐆𝐑𝐂 𝐰𝐢𝐭𝐡 𝐀𝐭𝐥𝐚𝐬𝐬𝐢𝐚𝐧 𝐂𝐥𝐨𝐮𝐝 Replacing spreadsheets with Jira Service Management and Confluence to build an audit-ready by default environment. • Real-time risk oversight instead of periodic audits • Automated evidence collection directly from DevOps workflows 2️⃣ 𝐌𝐨𝐝𝐞𝐫𝐧 𝐏𝐥𝐚𝐭𝐟𝐨𝐫𝐦 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 Removing friction so developers can focus on building. • Golden Paths with security and compliance built into GitHub or GitLab pipelines • Atlassian Cloud Enterprise for scalable, global setups without infrastructure overhead 3️⃣ 𝐌𝐚𝐧𝐚𝐠𝐞𝐝 𝐄𝐱𝐜𝐞𝐥𝐥𝐞𝐧𝐜𝐞 24/7 ISO-certified support for the full stack. • Seamless transition from on-prem to Cloud • Reduced operational cost (often €20,000+ annually) while strengthening security We’re not here to sell just another license. We help organizations 𝐚𝐥𝐢𝐠𝐧 𝐭𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐯𝐞𝐥𝐨𝐜𝐢𝐭𝐲 𝐰𝐢𝐭𝐡 𝐫𝐢𝐬𝐤 𝐚𝐧𝐝 𝐫𝐞𝐠𝐮𝐥𝐚𝐭𝐨𝐫𝐲 𝐫𝐞𝐪𝐮𝐢𝐫𝐞𝐦𝐞𝐧𝐭𝐬 — in a way that actually scales. 𝐈𝐬 𝐭𝐡𝐢𝐬 𝐬𝐨𝐦𝐞𝐭𝐡𝐢𝐧𝐠 𝐲𝐨𝐮’𝐫𝐞 𝐬𝐭𝐫𝐮𝐠𝐠𝐥𝐢𝐧𝐠 𝐰𝐢𝐭𝐡 𝐚𝐬 𝐰𝐞𝐥𝐥? How are you handling governance and compliance without slowing your teams down? Happy to compare notes — so send me a DM! #Scandinavia #DevOps #AtlassianCloud #GRC #NIS2 #DORA #PlatformEngineering #Eficode

48

CKA candidates often lose time, not because they don't know Kubernetes, but because their CLI skills are not refined enough. As an illustrative example, you can leverage tmux to significantly speed up your workflow. In the terminal, split the screen in half: tmux -> ctrl-b -> % On the left pane, you will troubleshoot with commands like: kubectl describe pod <name> On the right pane, you will continuously monitor status: kubectl get pod -w -----------------------sample output---------------------------- NAME    READY  STATUS        RESTARTS  AGE web-pod  0/1   ContainerCreating  0      3s web-pod  1/1   Running       0      8s ----------------------------------------------------------------- Just with this one simple setup, you can investigate while watching the cluster state being continuously updated on the right, saving you precious time. The workflow becomes: Observe → Investigate → Act → Continuously Monitor → Verify CLI adeptness becomes a real CKA advantage. Practice your commands inside tmux until your terminal layout, kubectl workflow, and monitoring become automatic.

44

12 kommentarer