on call
Take the pager
An incident simulator, a database of my work you can query, and your own session measured as you read. All of it runs in your browser and nothing leaves it.
01 · the shift
One shift, five pages
The failure modes are ones I've been paged for: an OOMKilled JVM, a poison message stuck on a Kafka partition, an ArgoCD reconciler quietly undoing someone's manual scale. The service names are made up. Reading the evidence costs a little error budget and guessing costs a lot. It takes about five minutes.
- When the spike starts at a deploy boundary, stop reading logs and start reading the deploy history.
- Latency that hits every service at once is rarely in any of them. Check the shared dependencies.
02 · query
Ask the database
A SQLite database of my work, compiled to WebAssembly and running in your tab. Pick a question, read the SQL that answers it, then edit it and run your own. The box is guarded by the same validator that refuses model-generated SQL in Clarity.
ask something
the query that answers it
Questions are mine. Database, query and result are real, and nothing leaves your browser.
03 · measure
Your session, measured live
The same numbers I'd put on a dashboard, taken from this page as you read it. Each one says what it means and what would count as bad.
How long the server took to start answering you.
not measurable here
How long before anything at all appeared on screen.
not measurable here
When the main content finished rendering.
not measurable here
How much the page jumped around while loading. 0 is perfectly still.
not measurable here
Measured in your browser on this visit. Thresholds are the Core Web Vitals definitions. The site is static files, so a slow first byte means a long route to you.
the build serving you this page