Packt SysOps

Figma’s S-1 dropped a quiet bombshell: they’re spending nearly $300K a day on AWS. That’s over $545 million spent over five years. All for compute, storage, bandwidth, and other AWS services, with no multi-cloud or diversification. It’s not just the cost that stands out. The filing also notes that Figma is *entirely dependent* on AWS. If terms change or an outage hits, their business is directly at risk. And replatforming isn’t a casual option. Their infrastructure is tightly bound to AWS. This isn’t a knock on Figma. It’s a reminder of the invisible gravity that comes with cloud scale. What starts as speed and flexibility can turn into deep lock-in, both technically and contractually. Compare that to 37signals, who took the opposite route: leavinf cloud and shifting workloads to their own hardware and saving millions annually. There’s no one-size-fits-all answer here. But if you're an infra engineer, it’s worth asking: Are your infra choices helping you move fast, or locking you in deeper than you realize? Follow Packt SysOps for more informative posts everyday.

📢 Only 1 day to go! 🚀 Mastering Enterprise Platform Engineering officially launches tomorrow! If you’re a platform engineer, SRE, DevOps leader, looking to set up your first internal developer platform, or scaling one to support AI-driven workloads, this book is packed with frameworks, strategies, and actionable insight. securely and sustainably. 🔍 Why this book matters: 🎯 A hands-on guide to platform engineering as a discipline, not just tooling 🏗️ Build golden paths and composable services that accelerate dev productivity 🤖 Integrate GenAI capabilities within internal platforms responsibly 🧭 Navigate platform team org models, from developer UX to product thinking 📏 Establish KPIs that align engineering impact with business goals 💡 Learn from real-world patterns used by leading enterprise teams 📚 Pre-order or grab your copy here → https://lnkd.in/dHeWcR-3 #PlatformEngineering #DevOps #InternalDeveloperPlatform #GenerativeAI #SRE #EnterpriseTech #BookLaunch

If you’re a #PlatformEngineer, #SRE, or #DevOps leader, looking to set up your first internal developer platform, or scaling one to support AI-driven workloads, this book is packed with frameworks, strategies, and actionable insight. securely and sustainably. 📚 Pre-order or grab your copy here: https://lnkd.in/eZsx64rz 🤖 Why this book matters: - A hands-on guide to platform engineering as a discipline, not just tooling - Learn from real-world patterns used by leading enterprise teams - Build golden paths and composable services that accelerate dev productivity - Integrate GenAI capabilities within internal platforms responsibly

By 2021, Uber knew Mesos was on its way out. It had served them well, but without active support, it wasn’t the future. So the team started migrating all stateless workloads to Kubernetes. The scale was wild: 3 million cores, 7,500-node clusters, hundreds of thousands of pods. But what really shaped their approach wasn’t just the infra. It was a principle: developers shouldn’t even notice. That meant zero changes to workflows. No manual migrations. No downtime. They built on “Up,” Uber’s internal federation layer, so services could silently move from Mesos to Kubernetes in the background. Most developers never knew it was happening. But under the hood, it was a different story. The team had to rebuild key integrations from scratch, optimize the API server to survive high churn, and even rework how Kubernetes handles artifacts and rollbacks. One lesson stuck: scale breaks defaults. Kubernetes worked great out of the box, until it didn’t. They hit weird UI crashes, informer delays, and rollout blind spots. At every turn, they had to dig in and tune the system to behave. By mid-2024, the migration was done. Seamlessly, by design. But it only worked because the team treated this like infrastructure *for developers*, not just infra for infra’s sake. Follow Packt SysOps to learn how the biggest teams actually run infra in production. Source: Uber Tech Blog

All cloud & sysadmin books: now just $9.99 Yes, really. Every title. Pre-orders included.   From Linux scripting and Kubernetes ops to running GenAI workloads across clouds, this is your chance to level up your infra skills without draining your budget.   You’ll find: - Automation & scripting in real-world Linux setups - Kubernetes guides that don’t stop at “hello world” - AI pipeline deployment at scale - Infra cost tracking with Prometheus, Grafana & Kubecost   Whether you’re deep in DevOps or just stepping into SysOps, these are the tools top teams are betting on.   Grab them now. Every book is $9.99: https://packt.link/Gh8Ra

Unlock the full #GenAI lifecycle on #Kubernetes, from first deployment to scaling #LLMs in production. This hands-on guide helps you optimize GPU usage, automate #infrastructure, and secure your workloads, so you can ship smarter, faster, and more cost-effectively in the AI era. 👉 Get your copy of Kubernetes for Generative AI Solutions today: https://lnkd.in/en6bddcB

Most engineers reach for Python by default. It’s powerful, readable, and battle-tested. But if you’re automating infrastructure, especially on Linux systems, that default can work against you. Here’s the uncomfortable truth: In many day-to-day scenarios, Bash is the better tool. It’s not about nostalgia. It’s about friction. Consider this: 📌Need to parse logs on a remote VPS or IoT device? Python might not even be installed. Bash is already there. 📌Writing a quick file-sorting script? Python needs imports and structure. Bash does it in one line. 📌Managing EC2 from the CLI? A few interactive prompts in Bash can be faster (and safer) than rewriting the same aws commands over and over. And here’s the kicker: Shell scripting doesn’t require “learning to program” in the traditional sense. It just requires knowing your environment, and using the tools it already gives you. Of course, Bash isn’t a silver bullet. It breaks down for complex logic, large datasets, or anything that needs serious error handling. But if you’re a cloud engineer who’s only reaching for Python, you might be overengineering your automation. Donald Tevault dives into this topic with real-world examples in our latest CloudPro special issue. Link to the full article in comments below👇 Follow Packt SysOps for more infrastructure insights.