|
| 1 | +# VoltWeave end-to-end tests |
| 2 | + |
| 3 | +This module verifies the V1 platform through the public API Gateway while using |
| 4 | +real local infrastructure and service processes. It is opt-in so a normal |
| 5 | +`mvn test` remains fast. |
| 6 | + |
| 7 | +## Prerequisites |
| 8 | + |
| 9 | +- JDK 21 |
| 10 | +- Maven 3.9+ |
| 11 | +- Docker Desktop with Linux containers |
| 12 | +- Available local CPU and memory for PostgreSQL, Keycloak, Kafka, Mosquitto and |
| 13 | + six Spring Boot processes |
| 14 | + |
| 15 | +No separately installed PostgreSQL, Kafka, Keycloak or Mosquitto is required. |
| 16 | + |
| 17 | +## Run from a clean checkout |
| 18 | + |
| 19 | +From the repository root: |
| 20 | + |
| 21 | +```powershell |
| 22 | +mvn -Pe2e verify |
| 23 | +``` |
| 24 | + |
| 25 | +The root reactor command is the recommended command. It builds every executable |
| 26 | +service JAR and the simulator dependency before Failsafe starts the E2E suite. |
| 27 | + |
| 28 | +For a quicker rerun after those artifacts already exist: |
| 29 | + |
| 30 | +```powershell |
| 31 | +mvn -pl tests/e2e -Pe2e verify |
| 32 | +``` |
| 33 | + |
| 34 | +Running only the module on a clean machine can fail if the service JARs or the |
| 35 | +`simulation-service` artifact have not been built yet. In that case, use the |
| 36 | +root reactor command above. |
| 37 | + |
| 38 | +## Runtime lifecycle |
| 39 | + |
| 40 | +`PlatformEnvironment` performs the following work for each test run: |
| 41 | + |
| 42 | +1. Reserves random host ports. |
| 43 | +2. Creates a unique Docker Compose project and named volumes. |
| 44 | +3. Starts PostgreSQL/TimescaleDB, Keycloak, Kafka and Mosquitto. |
| 45 | +4. Initializes Kafka topics and Mosquitto users/ACLs. |
| 46 | +5. Starts Portfolio, Telemetry, Intelligence, Dispatch, Settlement and Gateway |
| 47 | + as child JVM processes. |
| 48 | +6. Waits for each `/actuator/health` endpoint. |
| 49 | +7. Runs the ordered platform scenarios. |
| 50 | +8. Stops child processes and removes the Compose project and volumes. |
| 51 | + |
| 52 | +The suite never depends on the developer's normal Compose project or fixed |
| 53 | +host ports. A failed run preserves service logs under: |
| 54 | + |
| 55 | +```text |
| 56 | +tests/e2e/target/runtime-<pid>/logs/ |
| 57 | +``` |
| 58 | + |
| 59 | +## Covered journeys |
| 60 | + |
| 61 | +The suite currently proves: |
| 62 | + |
| 63 | +- Keycloak password grant for seeded admin, customer and operator users. |
| 64 | +- Organization membership, site preferences and device provisioning. |
| 65 | +- A provisioned MQTT device publishing telemetry through Kafka into the durable |
| 66 | + authorized device twin. |
| 67 | +- VPP membership, forecast baseline, flexibility snapshot, optimization preview |
| 68 | + and idempotent manual dispatch creation. |
| 69 | +- Provisioning and dispatch replay returning the original resource. |
| 70 | +- Reusing an idempotency key with a different payload returning `409 Conflict`. |
| 71 | +- Cross-tenant site isolation, role boundaries and forged JWT rejection. |
| 72 | +- Scheduled dispatch and prepared command state surviving Dispatch restarts. |
| 73 | +- Re-preparing commands after restart returning the same command IDs. |
| 74 | + |
| 75 | +The dispatch is deliberately scheduled in the future. Preparing it changes the |
| 76 | +workflow to `PREPARING`, but Telemetry does not publish the MQTT command before |
| 77 | +its `validFrom`. Command ACK, timeout and rebalance transitions are covered by |
| 78 | +their focused service integration tests. |
| 79 | + |
| 80 | +## Troubleshooting |
| 81 | + |
| 82 | +### Missing local Maven artifact |
| 83 | + |
| 84 | +If Maven reports a missing `io.voltweave:*:0.1.0-SNAPSHOT` artifact, run: |
| 85 | + |
| 86 | +```powershell |
| 87 | +mvn -Pe2e verify |
| 88 | +``` |
| 89 | + |
| 90 | +Do not start with `-pl tests/e2e` on a clean checkout because `-pl` alone does |
| 91 | +not build sibling reactor modules. |
| 92 | + |
| 93 | +### A service exits during startup |
| 94 | + |
| 95 | +Open its log in the newest runtime directory: |
| 96 | + |
| 97 | +```powershell |
| 98 | +$runtime = Get-ChildItem tests/e2e/target -Directory -Filter 'runtime-*' | |
| 99 | + Sort-Object LastWriteTime -Descending | |
| 100 | + Select-Object -First 1 |
| 101 | +Get-Content "$($runtime.FullName)\logs\dispatch.log" -Tail 100 |
| 102 | +``` |
| 103 | + |
| 104 | +Replace `dispatch.log` with `portfolio.log`, `telemetry.log`, |
| 105 | +`intelligence.log`, `settlement.log` or `gateway.log` as needed. |
| 106 | + |
| 107 | +### Docker resources remain after an interrupted JVM |
| 108 | + |
| 109 | +Normally teardown removes all resources. If the test JVM itself is killed, |
| 110 | +list projects first and remove only the exact E2E project shown: |
| 111 | + |
| 112 | +```powershell |
| 113 | +docker compose ls |
| 114 | +docker compose --project-name voltweave-e2e-<pid> ` |
| 115 | + --env-file tests/e2e/target/runtime-<pid>/compose.env ` |
| 116 | + -f infrastructure/compose/compose.yml down --volumes --remove-orphans |
| 117 | +``` |
| 118 | + |
| 119 | +Never use a broad volume-prune command for this cleanup. |
| 120 | + |
| 121 | +### Port or startup timeout |
| 122 | + |
| 123 | +Retry once after confirming Docker Desktop is healthy. Random ports avoid the |
| 124 | +usual `8080 already in use` issue, while startup failures are reported with the |
| 125 | +last service log lines. If failures repeat, inspect the preserved logs instead |
| 126 | +of increasing timeouts first. |
0 commit comments