Configuration Design Draft

The following shows configuration layering only; fields are not finally frozen.

Swell

api_version: wavemaker/v1

swell:
  name: staging-login-capacity
  seed: 20261104
  duration: 20m

spectrum:
  arrival:
    model: ramp
    from_per_second: 1
    to_per_second: 20
    duration: 10m

  concurrency:
    max: 5000
    saturation_policy: reject_new

  lifetime:
    distribution: fixed
    duration: 15m

  behavior:
    groups:
      - name: normal-player
        ratio: 1.0
        scenario: login-and-idle

execution:
  backends:
    - id: rust-worker-a
      type: native
      replicas: 2

    - id: kbengine-bots
      type: engine_bot
      adapter: kbengine
      replicas: 1

basin:
  ref: staging
  observability:
    required: true
    provider: prometheus
    collection_interval: 1s

report:
  format:
    - html
    - parquet

Note: this example only demonstrates the configuration structure. The semantics of arrival and concurrency.max need to be frozen after BLOCKER-1 is closed.

Engine Bot Backend

backend:
  id: kbengine-bots
  type: engine_bot
  adapter: kbengine

process:
  executable: /opt/kbengine/bin/bots
  working_directory: /opt/kbengine
  restart_policy: on-failure

bridge:
  transport: unix_domain_socket
  endpoint: /run/wavemaker/kbengine-bots.sock
  protocol_version: 1

engine:
  config_path: /opt/kbengine/assets
  bot_script_path: /opt/kbengine/assets/scripts/bots

capability_policy:
  require:
    - process.start
    - process.stop
    - bots.create_batch
    - observability.bot_count
  allow_unknown:
    - bots.destroy_individual
    - scheduling.exact_arrival