People who buy and track aircraft components already understand the central idea behind InfraOps, even if they have never heard the word. A part is identified by a number, it carries a serial, its history sits in a record, and an inspector can compare the installed item against that record. Computing teams have spent two decades learning the same lesson about servers and networks. This primer explains the vocabulary in plain terms and keeps a mini-glossary at the end.
Where the term comes from
InfraOps is shorthand for operating infrastructure, and it has no official owner. The Ops Atlas entry on the topic is frank that no standards body or certification defines it. Job ads, team badges and one commercial product all borrow the word, and the Atlas files those under coincidence of naming. Its own usage is narrower: it applies the label to the life cycle of infrastructure that is described in code, where machines, networks and storage are created from written definitions instead of manual clicking.
If you shop for parts, think of the difference between asking a supplier to find you something similar and handing over a drawing with exact specifications. The drawing can be reviewed, repeated and audited. Code does the same for computing.
The three records every beginner should know
A reference such as the Atlas page that frames InfraOps around infrastructure code sets out three separate ledgers that any environment ends up with:
- The written definitions. These state what you want, like a bill of materials.
- The automation’s memory. A file where the tooling notes what it thinks it created, much like a stores database.
- The live environment. This is what actually exists, the shelf in the warehouse.
Good practice keeps all three aligned. Bad days start when someone takes a part off the shelf without updating the ledger, or when the ledger is updated for an item that never arrived. Anyone who has chased a missing line-replaceable unit through a stockroom knows the feeling.
What goes wrong, in four flavours
The Atlas lists four ways the records fall out of step. Each has a stockroom twin.
- Approved but not installed: the change was accepted and nobody carried it out.
- Drift: the live environment changed after the last official build.
- Unmanaged items: something exists that no ledger covers, like an unlabelled part in a bin.
- Removed without a record: the ledger lists an item that was already scrapped.
A detail worth underlining for beginners: comparing the ledger with the definitions will not reveal the third kind. You only find unlabelled items by comparing against reality. That is why the Atlas lists inventory derived from code, and the habit of counting what exists, among the practices inside its scope.
Standard images and the swap-the-unit habit
Computers can be built in two ways. You can repair a server in place, patching it over the years until nobody knows what it contains. Or you can build a new one from a known template, called a golden image, and retire the old one. The second approach is often described as immutable. The Atlas clears up a common misunderstanding here: immutable does not mean never patched. It means the patching happens in the image pipeline, on a schedule, and the replacement is rolled out rather than tinkered with.
This is the same reasoning behind exchanging a serviceable unit for an overhauled one. For a buyer comparing hardware, this walk through aircraft fasteners and hardware shows how much detail gets standardised so that parts stay interchangeable. A server image does that job for software.
Records and standards that apply
Because the subject is partly about knowing what you own, it overlaps with a well-known family of security controls. The Atlas points to the configuration management group in NIST SP 800-53 Rev. 5: a baseline, change records, documented deviations and an inventory. You do not need to memorise control numbers. You only need to see the intention, which will look familiar to anyone who has read an airworthiness record: define the baseline, record the changes, explain the exceptions, and keep an accurate list of what you hold.
Why reviews matter more than speed
Before changing anything, an InfraOps workflow produces a preview, called a plan, showing exactly what will be created, altered or removed. People review the preview. The Atlas puts skipping this step on its list of anti-patterns, especially the excuse that the code explains itself. A purchasing analogy helps: you would not approve a large order from the title of a spreadsheet alone, and you should not approve a change without reading the proposed difference.
A second beginner trap is the order of operations. A preview is perishable. If another change lands first, the earlier preview no longer reflects reality, which is why careful teams generate a fresh one at the moment of apply.
A tiny exercise
Take a small system you know, such as a part-ordering website with a database and a file store. Write three lists on one sheet of paper: what the definitions say, what the hosting dashboard shows, and what your teammates remember. Mark every disagreement. That sheet is a crude audit, and you will learn faster from it than from days of study. If the sheet has surprises, you have just discovered the reason the discipline exists. For a related lesson from the parts world, common mistakes when purchasing control system parts is a useful parallel.
Mini-glossary
- Infrastructure as code: describing servers, networks and storage in written files.
- State: the automation’s remembered list of its creations.
- Plan: a dry preview, read before anything is altered.
- Drift: a difference that appears when the live environment moves away from the record.
- Golden image: a prepared template used to build new machines.
- Blast radius: how much can be affected if one change goes wrong.
- Decommissioning: removing something deliberately and recording that it is gone.
Start small, keep the records honest, and read the full Atlas entry when you are ready for depth. The vocabulary above will carry you through most of it.

