Wireshift

Notes on networks, infrastructure and the small tools that keep them running.

Keeping DNS changes boring

Short TTLs, one source of truth for records and a plan before every apply. Most outages we have seen started with a hand-edited zone.

Idempotent server setup, revisited

Why a second run that changes nothing is the most useful test a playbook can pass, and how we check it in CI.

Measuring latency the honest way

Averages hide the interesting part. A few notes on percentiles, jitter and sampling intervals for small monitoring setups.