A weblog about software, mostly.
Every two years I find myself re-reading the autovacuum documentation, and every two years I think "I should write this down." This time I did. A practical walkthrough of how autovacuum actually behaves under load, including some surprising things I didn't know after a decade of running Postgres in production.
The official Go documentation is generally excellent, but the guidance around context.Context has aged unevenly. I've been on enough code reviews to know that the original 2014 conventions have drifted in three different directions. Here's how I think about it now, and where I'd push back on the orthodoxy.
An unpopular take. Most of the observability stack at most companies is a way for engineers to feel productive while not actually getting better signal. I think we should be replacing dashboards with runbooks, and SLOs with conversations. Comments welcome, especially from people who disagree.
I moved out of Amsterdam to a small village in Limburg three years ago and switched to fully distributed work. People keep asking how it's going. The short answer is that I'd do it again, but with very different expectations about the trade-offs. The longer answer involves about a thousand words on the social cost of fewer hallway conversations.