Blue green deploy data simulation

blue-green-deploy-data-simulation

Simulate a blue-green deployment as a finite state machine with phases: idle → deploying-green → warming-green → smoke-testing → cutover → draining-blue → green-active (and a rollback branch from any post-cutover phase back to blue-active). Drive each phase with a tick function that advances a shared clock and emits instance-level events; phase durations should be configurable with sensible defaults (warm-up 30–60s, smoke-test 15–30s, cutover 3–10s, drain 30–120s). Keep a single activeColor field plus a trafficSplit number in [0,1] — during cutover, interpolate trafficSplit with an ease-in-out curve rather than a linear ramp so the animation feels realistic, and derive each side's RPS as totalRPS * split (or 1-split).

Generate per-instance health deterministically from a seeded RNG so the same simulation scenario replays identically: during warming-green, instances flip from starting → healthy at staggered intervals; during smoke-testing, inject a configurable failure probability (default 5%) on one instance to exercise the rollback path; during draining-blue, decay each blue instance's in-flight request count from its pre-cutover level to zero over the drain window. Emit both the aggregate state (for the router view) and a per-instance event log (for the timeline strip) from the same tick so visualizations stay in lockstep.

Expose scenario presets: happy-path, smoke-test-failure-rollback, slow-drain (long-lived websocket connections), partial-warmup-failure (1 of N green instances fails to become healthy), and double-cutover (rapid re-deploy). Each preset is a config object, not a hand-coded sequence, so the same tick engine produces all of them — this keeps the simulator honest and lets users toggle scenarios without reloading. Persist the seed and config in the URL so scenarios are shareable.