Documentation can be repetitive, boring, costly, incomplete and error prone. When the next engineer in your organisation comes along to generate more documentation on a similar subject, often, they start again and waste all prior investment of effort whilst possibly blissfully unaware it even existed. Not a wise investment in time and money.
Scope: there are many systems that need documenting and each have their own requirements, tools and peculiarities. This post assumes an infrastructure slant on documentation standards/requirements.
What about all those once loved documents that have been created, and served a useful purpose at one stage of their lives, however, over the years have fallen into disrepair and neglect. The systems they once accurately curated have evolved, adapted or disappeared… people forgot the documentation even existed as they made changes to the infrastructure but hurriedly move onto the next urgent thing.
How about the documentation that nobody ever looks at… denies its existence. Documentation that sits in a documentation store, one of many documentation stores with new and old versions scattered far and wide.
What about the comfortable colleague who doesn’t want to excerpt a little effort to learn how to do something in their job… documentation may as well not exist.
Sometimes it’s easy to wonder why bother with good documentation when far too often it feels like a wasted cause!
Why good documentation matters
- Customer may want to see the documentation for the solution you’ve built, or, manage for them.
- Our intrepid explorer colleague who likes to know how things work, and, takes the initiative and time to learn for themselves very much appreciates the labors of good documentation.
- May actually help diagnose a problem or issue faster.
- Whilst painstakingly writing the documentation, perhaps the author catches mistakes in the documentation or design.
Is there a better way? Maybe.
- Documentation re-use (gold templates)
- Let’s not waste so much effort creating the documentation… re-use as much as possible from templates.
- Documentation source control
- Collaborate with others to build the awesome gold template and scripts to auto populate
- Documentation auto population
- Much time is wasted by engineers transposing table information from an environment to a document.
- Non-searchable copy and pastes of tables (images) into the document make indexing and searching impossible
- Post change documentation automatic update
- A customer view vs an internal view
- A customer may want more detail as to what each component does, whilst, a colleague dealing with similar solutions all day long doesn’t care for the vendor marketing material (marchitecture)or technology descriptions.
- Automatic/schedule documentation updates
- Infrastructure as Code
- To some degree can be self documenting. Even better if it contains good comments.
So many options exist when it comes to documentation. I looked at a couple with a view to knocking something up quickly:
- Seemed like a good option which the possibility of styling a pages PDF output.
- Didn’t invest the time in working out how to integrate external data sources into a page, but, seemed do-able via Dynamic Content Macro.
- Whilst powerful and very popular among many technical professions for technical documentation, it seemed like there was a high learning curve.
- reStructuredText & sphinx
- Was my preferred option for a while there, however, ran into issues generating PDF documents and I didn’t want to spend the time troubleshooting.
- Really liked the sphinx HTML output generated from python script docstring comments
- AsciiDoc (language) & AsciiDoctor (processor)
- Settled on AsciiDoctor as it did everything I wanted and seemed to just work, especially with asciidoctor-pdf.
- The swiss army knife for converting between various documentation types.
Source controlled gold template structure
The directory structure is important to consider as it would enable the ability to collaborate, house chapter based text, refer to images and hold information gathering scripts. Below is a possible directory structure to nicely split out static text, images and dynamic content.
│ └── includes
│ └── chapter_01
│ ├── host_info.csv
│ └── mem_info.csv
│ ├── includes
│ │ ├── chapter_01
│ │ │ └── sample-diagram.png
│ ├── document-footer-image.png
│ └── document-title-image.png
└── scripts (maintain scripts in a separate repository and import as required)
└── query_hosts.ps1 (example)
Further consideration would be required for sharing content between different gold template repositories. In fact, the above structure worked well for generating PDF documents out of the source, however, when editing the AsciiDoc code with Atom or Visual Code Studio with various markup tools enabled, the HTML view present was broken due to image paths.
Great, gold templates in git. What now?
Once you have a new project that needs documenting, clone your gold template and create a new repository to host the documentation you’re about to create.
Modify the new repository to meet your needs and execute some of the helper scripts to gather required environment data, commit and push your changes to the new repository and now build/compile your documentation.
Modify the new repository documentation as desired, to meet your requirements.
Seems like a lot of effort. Why not use a word processor as we normally do?
How about day two? Your documentation and environment align at the time of creation (hopefully), what happens weeks, months, years into existence? Hopefully the documentation is not gathering dust!
If it is, setup an automated scheduled task to gather the current state information for you and compile the up-to-date documentation.
Perhaps your monitoring system provides you with estimated capacity exhaustion information… maybe that’s useful in your documentation if sharing with a customer on a regular basis. (Perhaps better in some sort of reporting document)
What if you had an “infrastructure as code” setup and stored the code in a code repository. How nice would it be to update documentation when configuration is updated in your code repository, perhaps as a webhook or CI/CD pipeline.
Documentation as a Service
How about you don’t want all your engineers to install the documentation tool chain required to produce the final version of the documentation.
Consider a small web app which accepts the following inputs:
- Code repo URL
- Code repo username & password
- Email address, or, target location (document store)
Process flow for web app: Submit params -> Clone Repo -> Build/Process Documentation -> Preview, email or upload final document. Display errors if any.
I thought there would have been more discussion around documentation as code, or, automatically generating documentation. Perhaps documenting infrastructure automatically is not as glamorous as other topics?
Maybe it’s some what redundant if you have self documenting infrastructure as code? But, probably falls short as this only covers the “what” of a solution design and not the “why”.