Imagine you are building a model village on a large table. You could glue every house down as you go and hope you remember where things were. Or you could keep a drawing of the village, a notebook listing what you actually placed, and a habit of glancing at the table to see that nothing has moved. InfraOps works a lot like the second approach, only the village is made of servers, networks and storage in the cloud. The primer below goes one small step at a time and finishes with a short word list for later.
Why a village needs three records
In the Ops Atlas, a public reference about operations topics, the central picture is that infrastructure managed through code has three copies of the truth. The drawing is your code, written to say what you want. The notebook is the state, a file the tooling uses to remember what it has built. The table itself is reality, the real resources running at the cloud provider. A beginner can get a long way by remembering only this trio and asking, whenever something feels odd, which of the three has changed.
When the table and the drawing disagree
Suppose a friend moves a house while you are away. The drawing still shows the old position. In technical language, that is drift. The Atlas also describes three other mismatches: a change that was approved but never put in place, an item built without ever appearing in your records, and an item removed from the table while the notebook still lists it. Four small mismatches, four different ways to notice them.
A tiny point trips people up on day one. A report that says nothing is different covers only the things in your notebook. A house someone placed on the table without telling you will not be mentioned. That is why careful teams also go looking for strays now and then.
The plan before the apply
Before the tooling changes anything, it can show a preview called a plan. Reading a plan is the single most useful beginner skill, because it lists what will be created, changed or destroyed. The Atlas warns against skipping that review on the grounds that the code is its own documentation. The code tells you what you hope for; the plan tells you what is about to happen. There is also a practical wrinkle: a plan is tied to a particular version of the state, so once one plan is applied, other plans prepared earlier stop being valid.
Keeping the village small enough to manage
Beginners are sometimes tempted to put everything in one drawing. The Atlas quotes advice from Google Cloud to keep a single state to around a few dozen resources and not go beyond 100. Smaller pieces mean quicker previews and a smaller area affected by a mistake. If you are reading this from Japan or anywhere that supports large, busy services, you will see the same principle in the way big systems are divided into regions and accounts.
A very short history
None of this is brand new. The Atlas traces the idea back to a configuration tool from 1993, then to the first cloud template service in 2011, and to the arrival of Terraform, announced in 2014. A formal definition of infrastructure as code followed in 2016, along with a book that is now in its third edition. Knowing the dates helps you see that you are joining a mature craft, not chasing a fashion.
Mini-glossary
- Code (intent): the written description of what you want to exist.
- State: the tooling’s record of what it believes it created.
- Reality: what the provider is actually running right now.
- Plan: a preview of the changes before they happen.
- Apply: the step that makes the changes real.
- Drift: a difference caused by someone altering a resource outside the code.
- Module: a reusable bundle of code, like a standard house design.
- Golden image: a prebuilt machine template, refreshed on a schedule.
- Blast radius: how much could be affected if one change goes wrong.
- Import: bringing an existing resource under the control of the code.
Where to go from here
Beginners in Japan often ask what the wider field looks like before they choose a learning path. Two pieces on this site help with that: one on the likely direction of DevOps work across Japan and abroad, and a companion collection of fifty-plus terms every newcomer should know. They give context for the words above and let you see how infrastructure discipline fits among them.
A gentle first exercise
Open a free cloud account or a local sandbox and create one small resource, such as a storage bucket, through code. Read the plan carefully before you apply. Then go to the provider console, change one setting by hand, and run the plan again. Seeing a difference appear is the best lesson in drift you can get, and it costs almost nothing.
When you want to go deeper, the long article at the Ops Atlas page about InfraOps covers the topics in this primer with diagrams and more careful definitions. It also notes that the term is informal, with no certification behind it, so enjoy the learning without worrying about a badge.
That is all the village needs: a drawing, a notebook and a regular look at the table. Everything else in the field is a more careful version of those three habits.









