Define the production target for load testing

This lab builds k6 streaming load test for chat completion gateway. The human in the loop is capacity planner before Black Friday traffic 3× baseline. Scope is intentionally narrower than “make AI reliable”: you will prove one oracle — find saturation knee near 180 RPS; beyond that queue depth explodes — and one invariant — abort if error_rate > 1% or p95_ttft > 1.5s; test cost ≤ $50. Record non-goals in your notes so a later change cannot silently expand authority. The incident mnemonic for the chapter is LOAD-FALSE-GREEN-33; design as if that ticket is already written and you are filling evidence.

Turn the oracle into a table: input fixture, expected observable, prohibited side effect, owner, latency/cost ceiling. Separate model taste from software correctness — transport, auth, parsing, and termination must be deterministic even when generated text varies. Primary metric family: saturation_rps, p95_ttft_ms, test_usd. Averages without a denominator or revision label do not gate release. Fake external dependencies in unit tests; live calls wait until fakes pass.

Before implementation, encode a failing check that would have caught load generator retries hide server shedding — false green. That failure is the pedagogical north star for later pages: contracts reject it, happy path never performs it, validation asserts it, failure-handling contains it, observability detects it, security-ops prevents privilege tricks around it, and mastery replays it in a drill. Endpoint under study: POST /v1/chat/completions (stream).

Capacity note for planners: estimate peak demand on POST /v1/chat/completions (stream) and the cost ceiling for a failed retry storm. Write the abort conditions — unbounded spend, cross-tenant leakage, or inability to roll back — before you enjoy the first green test. Prefer synthetic fixtures shaped like production over anonymized production dumps you cannot share in class. When you are tempted to widen scope, re-read the oracle (find saturation knee near 180 RPS; beyond that queue depth explodes) and cut features that do not serve it. The teaching outcome is judgment under constraints: capacity planner before Black Friday traffic 3× baseline gets a trustworthy control, not a kitchen-sink framework. Keep the language of release decisions: promote, hold, or roll back — never “see if it gets better.”

Decide what will live in version control on day one: fixtures, contract markdown, and a failing test name. Write the cost ceiling as a hard number with currency and period. If the lab involves clusters, name the non-prod context you will use and forbid prod kubecontexts in scripts. Capture the baseline metric once before changing code so later gains are comparative. Refuse tools that hide the request path behind magic macros until the oracle is green on fakes. Your README section for this page should be five lines or fewer and still falsifiable. In this chapter the product is k6 streaming load test for chat completion gateway, the human stakeholder is capacity planner before Black Friday traffic 3× baseline, and the incident id you design against is LOAD-FALSE-GREEN-33. Re-state the oracle in your notes — find saturation knee near 180 RPS; beyond that queue depth explodes — and keep the invariant visible: abort if error_rate > 1% or p95_ttft > 1.5s; test cost ≤ $50. Track saturation_rps, p95_ttft_ms, test_usd as the scoreboard. Surface under change control: POST /v1/chat/completions (stream). If you only have forty minutes, finish the fixture for load generator retries hide server shedding — false green before polishing UI. Promotion language stays ternary: promote, hold, or roll back based on evidence, not hope.