end-to-end-life-cycle.md

The majority of scenario in system tests are based on EndToEndScenario class, itself based on _DockerScenario class. One of the major constraints of those scenario is to precisely control the startup sequence, to ensure that no flakyness is induced by a service not healthy when a depending one is rely on it.

Once the infra is ready to be tested, the test execution is divided in three sequential steps : setup, wait, then test.

Here is the life cycle map :

flowchart TD
    Proxy --> Agent
    C --> B
    Agent --> B[[Buddies]]
    Agent --> Weblog
    D[Open telemetry collector]
    C[[Postgres, and any other DB engine...]] --> Weblog
    Weblog --> E(Infra ready to be tested):::greenText
    D --> E
    B --> E
    E --> F[Execution of all setup methods]
    F --> G[Grace period]
    G --> H[Execution of all relevants test methods]
    classDef greenText color:#2B2

Loading

Two important note :

1. each container must satisfy to an healtcheck command to be consider as started
2. depending on the scenario, some container may not be started (Buddies, DB engines, OTM collector ...)
3. on replay mode, only the very last step (Execution of all relevants test methods) is executed.