The Potter's Wheel: On Centering Your Systems

There is a moment in pottery, just after the lump of clay is first thrown onto the wheel, that is pure, unadulterated chaos. The potter’s hands are wet, the wheel spins fast, and the clay wobbles violently, threatening to fly off entirely. The only way to prevent this disaster is to perform a process called ‘centering’. With steady, immense pressure from both hands, the potter forces the clay to find its true center, to spin in perfect, silent alignment with the axis of the wheel. Until it is centered, nothing else is possible. No vase, no bowl, no cup can be formed from a lump that is off-kilter.

Our small services are that lump of clay. We throw a new project onto the wheel—a fresh idea, a new microservice, a database cluster—and it is immediately a chaotic, spinning mess of dependencies, configurations, and potential points of failure. Our instinct is often to start shaping it immediately, to build features and add complexity. We try to make a beautiful vase from a wobbly foundation. This is how we end up with systems that are fragile, noisy, and exhausting to maintain.

The lesson from the potter’s shed is simple, yet profound: before you build anything, you must center your system. Centering, in our world, is the boring, reliable work that happens before the feature flags are set. It is the unglamorous act of ensuring a service has robust logging before it handles a single real request. It is implementing metrics and health checks not as an afterthought, but as the very first endpoints. It is scripting the backup and restore process before a single byte of user data is written to disk.

This pressure feels like a delay. It feels like we are not making progress towards the beautiful shape we envision. But the potter knows that this pressure is not an obstacle to creation; it is the prerequisite for it. A centered piece of clay is responsive. The slightest touch from the potter’s finger translates into a smooth, controlled change. An uncentered piece fights back every step of the way.

When your systems are centered, operations become a gentle guidance rather than a frantic struggle. A spike in latency is immediately visible in the metrics you built first. A failing health check gracefully redirects traffic. A corrupted database is a minor inconvenience, restored from the backups you practiced on day zero. The system spins true, and your hands, though still required on the wheel, can work with confidence and grace. The chaos is subdued not by fighting each wobble as it appears, but by having done the hard, quiet work of finding the center before the storm ever arrived.

Notes & further reading

A few pages I came back to while writing this: