Security + multi-tenant: API_AUTH_TOKEN auth, POST /stop, and tenant metrics

[cbaugus]

Summary

• API_AUTH_TOKEN bearer token auth on POST /config and POST /stop — when the env var is set both endpoints require Authorization: Bearer <token>; 401 on mismatch. Backwards-compatible: when unset, endpoints are open (closes Security: Authenticate POST /config endpoint #91)
• POST /stop endpoint — stops all workers, transitions node to idle, returns JSON summary (closes Add POST /stop endpoint for test cancellation #93)
• Optional tenant label on Prometheus metrics — set metadata.tenant in the YAML config and every request metric gets a tenant label so per-tenant request counts are queryable for billing
• Tenant-scoped stop — POST /stop {"tenant": "acme"} returns 409 Conflict if a different tenant's test is active, preventing one client from stopping another's test
• Nomad HCL job files and Consul KV YAML config example added

What changed

src/yaml_config.rs

• YamlMetadata gains pub tenant: Option<String> — optional, fully backwards-compatible

src/metrics.rs

Five metrics now carry a "tenant" label alongside "region":

• REQUEST_TOTAL — primary billing counter
• REQUEST_STATUS_CODES
• CONCURRENT_REQUESTS
• REQUEST_DURATION_SECONDS
• REQUEST_ERRORS_BY_CATEGORY
• SCENARIO_REQUESTS_TOTAL

src/worker.rs

• WorkerConfig and ScenarioWorkerConfig gain pub tenant: String
• All metric recording call sites updated to pass tenant

src/main.rs

• TestState tracks tenant: Option<String> for the active run
• Config-watcher extracts tenant from YAML and passes it to workers
• Metrics updater reads active tenant from TestState; resets RPS delta on tenant change
• GET /health exposes "tenant": null | "acme"
• POST /config — auth check before body consumption
• POST /stop — optional {"tenant": "..."} body; 409 if tenant mismatch; clears tenant on stop

Auth behaviour

API_AUTH_TOKEN set?	Request has correct header?	Result
No	N/A	Open
Yes	Yes	Proceeds
Yes	No / missing	401

POST /stop tenant scoping

Body	Active tenant	Result
(empty)	any	Stop all workers
{"tenant":"acme"}	"acme"	Stop
{"tenant":"acme"}	"other" or none	409 Conflict

YAML example

metadata:
  name: "API smoke test"
  tenant: "acme-corp"     # optional — adds tenant label to all metrics
config:
  baseUrl: "https://api.acme.com"
  workers: 50
  duration: "30m"
load:
  model: "rps"
  target: 500

Prometheus query for billing

# Total requests for a specific tenant
sum(rust_loadtest_requests_total{tenant="acme-corp"})

# Per-tenant RPS
sum(rate(rust_loadtest_requests_total{tenant="acme-corp"}[1m]))

Test plan

• CI passes on dev
• POST /config with tenant YAML → GET /health shows "tenant": "acme"
• Prometheus shows tenant="acme" label on requests_total
• POST /stop {"tenant":"acme"} stops correct test
• POST /stop {"tenant":"other"} returns 409 when different tenant active
• POST /stop (no body) stops regardless of tenant

Related issues

• Closes Security: Authenticate POST /config endpoint #91 — Authenticate POST /config
• Closes Add POST /stop endpoint for test cancellation #93 — POST /stop endpoint

🤖 Generated with Claude Code