TUSST — the Stroop SigilTUSST
‹ advanced path

Running It in Production

What separates a prototype from infrastructure someone is paged for: observability, honest percentiles, load testing, graceful shutdown and a reliability posture you can defend.

serves: Ownership of production infrastructure7 lessons · ~4.5h
progress0/7
Start track

// lessons

  1. 01

    Counters, Gauges & Histograms

    Pick the right instrument for a question, and see what each one cannot answer.

  2. 02

    Percentiles From Buckets, and Why You Cannot Average Them

    Compute p50/p95/p99 from bucket counts, and merge two instances without lying.

  3. 03

    Structured Logs & a Correlation ID

    Thread one ID through a call chain and reconstruct a single request out of an interleaved stream.

  4. 04

    Backoff, Jitter & a Retry Budget

    Bound retry amplification with a budget instead of hoping the dependency recovers.

  5. 05

    A Circuit Breaker as a State Machine

    Stop sending traffic into a dead dependency, and probe it back to life without a stampede.

  6. 06

    Graceful Shutdown: Drain, Deadline, Force-Close

    Take a pod out of rotation and finish its in-flight work without a deploy dropping requests.

  7. 07

    Little's Law: A Latency Target Is a Concurrency Limit

    Turn a latency SLO into the number of concurrent requests you are allowed to admit.

Every exercise compiles and runs in the Forge sandbox — standard library only, no external crates.