The Potter's Centered Wheel: On the Balance of a True Spin

There’s a world of difference between a service that is merely up and one that is truly healthy. We see the successful HTTP 200, the green checkmark on our dashboard, and we breathe a sigh of relief. The system is alive. But a veteran potter would look at this and see only the first, most basic condition of their craft: the lump of clay is on the wheel. It is ‘up’. The real work, the art of creating something stable and functional, has only just begun.

The true measure of a potter’s skill isn't in getting the wheel to spin; it's in the act of centering. This is the delicate, tactile process of applying pressure to the spinning clay, feeling for the wobbles and inconsistencies, and coaxing the entire mass into a perfectly symmetrical rotation around a single, fixed axis. A pot thrown off-center will look misshapen, its walls will be uneven, and, most critically, it will be fundamentally weak. It might hold water for a time, but the slightest stress will cause it to fail. It's a service that responds to pings but collapses under a modest load.

Our health checks are often the equivalent of checking if the clay is on the wheel. We probe a single endpoint, maybe check a database connection, and call it a day. But what about the subtle wobbles? The slow queries that haven't yet timed out? The memory leak that’s only noticeable under a specific, infrequent workflow? These are the imbalances that our simplistic health probes completely miss. The system appears centered from one narrow angle, but it's spinning with a dangerous, hidden eccentricity.

Feeling for the Wobble

A potter doesn't rely on a single, binary gauge. They use their hands, their eyes, and their intuition. They feel the vibration through their fingertips. In our world, this translates to a richer, more tactile form of observability. We need to move beyond the simple up/down and feel the system’s vibration through detailed metrics: not just average response time, but the 95th and 99th percentiles that reveal the long tail of user frustration. We need to listen for the sound of imbalance in our logs and traces, correlating a slow API call in one service with a subtle queue buildup in another.

Centering is also an active, continuous process, not a one-time event. A potter must constantly make micro-adjustments as the clay rises and changes shape. Similarly, declaring a service ‘healthy’ after a deployment is a dangerous fantasy. True health is a persistent state maintained through continuous observation and adjustment. Our automated systems must be capable of detecting the nascent wobble—the slight increase in error rates, the gentle creep of latency—and applying corrective counter-pressure before the entire piece flies off the wheel.

Ultimately, the potter’s goal isn't just a stable lump of clay; it's to create something of utility and beauty. Our goal isn't just a service that’s technically running. It’s a service that performs reliably, predictably, and robustly for the people who depend on it. By borrowing the potter’s mindset, we learn that reliability isn't a status to be checked off, but a delicate, dynamic balance to be constantly maintained. It’s the hard-won skill of feeling for the wobble and having the patience to center the spin, ensuring that what we build is not merely up, but truly sound.

Notes & further reading

A few pages I came back to while writing this: