This post is about “real” virtual machines, not container-based virtualization. So the “you can just stuff it into Docker” reflex doesn’t quite apply here.
Building cloud services means we constantly use virtual machines. In production, type-1 hypervisors — the kind that, simply put, run on bare
metal; on our local machines, type-2, the familiar ones — VirtualBox,
qemu, and possibly Parallels Desktop.
Sadly, despite all the modern build systems available, we still end up using make and writing Makefiles by hand.
I needed to put together a help screen for the build targets in a project’s current Makefile. Out of the box, the ancient utility doesn’t
give you a list of build targets and their descriptions — it was never
meant to.
All of our project documentation is written in Markdown.
It’s a simple, convenient format you can read in raw form without any
post-processing.
I often need to experiment with monitoring settings and error/metric collection across various environments.
Rather than carving out a separate project in our shared Sentry,
I prefer to spin up my own local copy.
The deployment takes a while; on an “old” laptop it can take about ten minutes.
Many teams discuss and maintain ADRs using tools like Confluence and Wiki. We all know perfectly well that this is a “write once, read never”
approach.
ADR — Architecture Decision Record.
I keep running into engineering teams that underestimate ADRs.
asdf is a runtime version manager.
It can replace nvm, rbenv, pyenv, gvm, and a pile of other
version managers with a single tool.
jo is a command-line json builder.
I often see coworkers in chats passing around request snippets or API call examples with the json baked in as escaped strings.