ISSUE 479 | Week in Review: Solving Kubernetes’ Biggest Problem

“It’s funny: In migrations, you think you’re about 70% of the way done, and you start doing significant testing, you realize you’re only 10% of the way through, because you’re finding so many issues.”

Week in Review: Solving Kubernetes’ Biggest Problem

Thank you to my colleague Richard MacManus for writing last week’s Week in Review — I took a few days off to go “downy ocean,” as we say here in the U.S. Mid-Atlantic, and am the better for it. Time spent with friends in a pleasant (albeit sweltering) locale is always good for what ails you.

And speaking of what ails us, Kubernetes has long struggled with one big design flaw: it’s not built for running or managing stateful workloads. Databases can be Kryptonite to Kubernetes, and the situation gets even more confusingly complex as you scale.

No surprise, then, that last week’s most popular post with The New Stack’s readers was one contributed by Rotem Tamir, co-founder and CTO of Ariga, and Gabriele Bartolini, vice president and chief architect of Kubernetes at EDB. 

The two introduce the notion of “reconciliation loops” for stateless infrastructure. Such loops operate much like air conditioners keep a room’s temperature consistent, even when the outdoors temperature changes: “They aren’t just automation. They’re autonomous,” Tamir and Bartolini wrote. “They provide us with infrastructure that responds to failure, handles change safely, and continues to move forward without requiring constant oversight at every step.”

But such a model doesn’t work for stateful infrastructure and workloads, like databases. Instead, Kubernetes operators — ”purpose-built controllers that extend Kubernetes to manage complex, domain-specific, stateful resources” — can get the job done. In their TNS post, Tamir and Bartolini walked readers through an example of using two open source projects — CloudNativePG and Atlas Operator — to manage PostgreSQL schema changes on Kubernetes. 

The result represents what the authors call “a truly GitOps-native approach to schema change management.” 

They sum up why solving this particular problem is vital for organizations that use K8s: “As Kubernetes becomes the control plane for everything, state can’t be the exception.” 

Databases, they add, cannot be treated like “untouchable snowflakes — segregated from the rest of our infrastructure automation … to achieve truly modern, scalable, and repeatable systems, we must integrate them into the same workflows we use for everything else.”

— Heather Joslyn, editor-in-chief, The New Stack

Trending Stories

MCP + SQL: The Secret Weapon To Connect AI to Enterprise Systems

By speaking the language your AI already knows and providing secure access through MCP, you can transform the way your organization works with data.

Bash 5.3 Has Some Big Improvements — Here’s How You Can Test It

Version 5.3 adds some significant new features, the most important of which is a new form of command substitution that executes a command in the current shell context.

Kiro Is AWS’s Specs-Centric Answer to Windsurf and Cursor

AWS today launched Kiro, its answer to Windsurf and Cursor, with a focus on writing specs over prompts.

MCP: A Practical Security Blueprint for Developers

Model Context Protocol unlocks powerful AI-first integrations for developers, but without care, it can introduce serious risks.

Open Source Is Worth Defending

The definition of "open source" is quietly eroding. When these lines blur, trust breaks — and open source doesn’t work without trust.

Building Real-Time SMS, WhatsApp and Slack Alerts

Learn how to create a comprehensive alerting system across three critical channels using InfluxDB 3's new Python processing engine.

How To Build Smarter Factories With Edge Computing

Learn the essential components of the industrial edge, a three-phase approach to put your plan into action and success factors to measure.