Race weekscene 1 / 26~2 min

The freeze

Monday of race week. Nothing new reaches production until the race is run.

On the Monday of race week, a rule comes into force at Traversal that sounds like superstition and is anything but: nothing new reaches production until the race is run.

The freeze. No new features, no casual improvements, no Friday cleverness, however tested, however small, for seven days. The system that exists on Monday morning is the system that will meet Sunday, and everyone has a week to become deeply, personally acquainted with it.

The word with a change of voice

The place the freeze protects is Concept · lights on your mapproductionThe live system serving real users. Not a copy, not a rehearsal. In prod carries stakes because every action there affects real people and real money in real time, and the word works as culture as much as vocabulary. It marks where carefulness stops being optional., and you have been hearing engineers say the word all story with a small change of voice. Now you know why. Staging holds nothing real. Production holds twenty-five thousand entries, a decade of results, and one Sunday that cannot be run twice. Race day does not raise production’s stakes — race day is what the stakes have been about all along.

The watching

All week, the wall of charts in the ops room does what it always does, with more people looking at it. The discipline of collecting a live system’s vital signs and comparing them to expectations is Concept · lights on your mapmonitoringContinuously collecting signals from a live system, its vital signs, error counts, response times, resource use, and comparing them against expectations, with alerts when reality drifts. It answers the questions somebody thought to ask in advance., and it answers the questions somebody thought to ask in advance. Is the entry service healthy. Are the workers keeping up. Known questions, watched forever.

Its deeper cousin is Concept · lights on your mapobservabilityThe property of a system whose internal state can be understood from its outputs: logs, metrics and traces rich enough that engineers can ask new questions without shipping new code. Monitoring watches for known problems; observability equips you for the ones nobody predicted.: building the system so rich in outward evidence that new questions, the ones nobody predicted, can be answered on the spot. Monitoring is the smoke alarm. Observability is being able to walk into any room of the building and see. Sunday will ask Traversal one question nobody has thought of yet, at the worst possible moment, on a bridge. The whole point of this week is that the answer will be findable.

Monday passes quietly. Tuesday mostly does too, apart from one scheduled act of violence that Thursday’s review will brag about. On Wednesday morning, the outside world declines to respect the freeze.

End of scene

This scene covers: production, monitoring, observability

The advisory

Act III