Advice about how infrastructure should be run is worth more when the person giving it runs some. This is the tooling behind the practice — monitoring, alerting, configuration management and asset tracking, all of it in daily use.
Agent and agentless metric collection across hosts and services, with thresholds, trend history and dependency-aware triggers. Retained history is what turns "it feels slow lately" into a date and a cause.
Answers Is this normal for a Tuesday, and when did it change?
Independent endpoint checks sitting outside the systems they watch, with notification routing so an outage reaches a person rather than a dashboard nobody is looking at.
Answers Is it down, for how long, and who has been told?
Linux configuration held declaratively in git. Changes arrive as commits, which means they can be reviewed, diffed and reverted — and a host can be rebuilt from source rather than reconstructed from memory.
Answers What changed, who changed it, and can we put it back?
A single inventory of record for hardware, licences and assignments, including lifecycle state. Asset data that lives in a spreadsheet stops being true within about a month.
Answers What do we own, where is it, and who has it?
Four rules that decide how anything gets added to the platform.
Instrumentation goes in when a service is built, not after its first outage. A metric only becomes useful once it has history behind it.
Availability checks run independently of the systems they monitor, so a failure cannot quietly take its own alerting down with it.
If a change is not in git it did not happen. Reproducible rebuilds matter more than clever one-off fixes.
Assets, licences and assignments are tracked in one place, because reconciling three partial lists is how licence spend and stale accounts go unnoticed.