Six InfraOps Myths, Retold From Mission Control

Anyone who has followed a satellite launch knows the odd calm of mission control. Rows of screens, a few quiet voices, and a constant comparison between what the plan said would happen and what the spacecraft is actually doing. Server rooms and cloud accounts have a similar rhythm, and the working style behind it is often called InfraOps. It is a loose label, not a regulated profession, but the habits it describes are surprisingly easy to explain with orbits and flight plans. Here are six myths I keep hearing, retold from the control-room side.

Myth 1: “If the checklist is green, we know what is up there”

A flight controller would laugh at this. A green board tells you that the items someone thought to instrument are healthy; it says nothing about debris, a stray object, or a part nobody added to the tracking sheet. The same holds for infrastructure written as code. A tidy preview of changes only covers resources the tool already manages. According to the Ops Atlas entry on InfraOps, a clean plan proves nothing about things that were created by hand outside the code, and its own drift walkthrough only noticed a hand-made firewall rule because an unfamiliar ID appeared in a diff. Quiet is not the same as complete.

Myth 2: “There is one record, and it is right”

Think of three documents on a launch day: the mission plan, the ground station logbook, and the real position of the spacecraft. They agree most of the time, and the whole job lives in the moments they do not. The Atlas describes infrastructure the same way: the code states intent, the tool keeps its own record called state, and the cloud provider holds reality. Four kinds of mismatch follow, namely changes merged but never applied, plain drift, resources nobody codified, and things deleted behind the tool’s back. Naming the gaps is half of closing them.

Myth 3: “Immutable means we never patch”

Orbiters are not repaired with a wrench either, yet they still get updates. When people hear that production machines are replaced rather than modified, some conclude patching has gone away. The Atlas calls that a myth: the patching moves upstream into a scheduled image pipeline, so a fresh build with the fix replaces the old machine instead of anyone logging into it. For a feel of constant scheduled upkeep, see our note on the nearest orbital band.

Myth 4: “A safety lock on deletion protects the data”

Many teams add a guard flag so that a destructive command stops with an error. It is a sensible seatbelt, but the Atlas points out a catch: if someone removes the resource block from the configuration altogether, the guard disappears with it and the protection never fires. The advice is to pair the code-level lock with a deletion lock at the provider. Compare it to a launch abort switch wired into one console only; if that console is unplugged, the switch is gone too. Redundant protection lives in two different places.

Myth 5: “Automatic correction is the grown-up answer to drift”

It sounds appealing: whenever reality wanders from the plan, a robot nudges it back. The Atlas warns that blanket auto-revert is really an apply pipeline that nobody reviews, running with the strongest credentials you own. Its suggested compromise is narrow. Let the robot fix modules that hold no durable data, only when no incident is open, and for everything else raise an alert and assign a person, with a target of roughly 72 hours. Spacecraft operators work much the same way. Small station-keeping burns can be automated, but a dramatic manoeuvre goes through a human who has read the situation. For the physics of why stationary satellites need such nudges at all, see our plain walk through geostationary orbit.

Myth 6: “Bring everything under the code on day one”

Imagine a team inheriting an old ground station with hundreds of undocumented cables. Writing code for all of it at once is tempting and usually a mistake. The Atlas says imported definitions tend to preserve historical accidents, and it notes that cloud documentation itself advises avoiding imports where possible. Its illustrative suggestion for a console-built estate of about 300 resources is to take the ten riskiest first and aim for a coverage jump from nothing to around 80 percent over a quarter, a pace that leaves time to understand what is being captured.

One more habit worth stealing from flight operations is separating phases. Our overview of each stage between liftoff and operations shows how each stage has its own checks, and infrastructure changes benefit from the same staging: review a plan, apply it, then compare outcome to intent.

What to take from the six myths

Strip away the jargon and the story is a familiar one. You hold a plan, you hold a record of the plan, and you hold a view of reality; trouble starts when any one of them is trusted without the others. Teams who do this well keep their records small enough to reason about, review every change before it flies, and measure how much of their estate the code truly covers. They also choose a few honest numbers instead of a flattering one. The Atlas cites a worked case where 82 percent coverage sounds fine until you notice it leaves several hundred resources that nobody reviews.

That is the quiet lesson of both mission control and infrastructure work. Confidence comes from comparing sources and saying out loud which parts of the picture are still blind, not from a single reassuring dashboard.