Workflows

A high level description of the runner manager workflows.

Webhook events

Webhook events are sent by GitHub on selected events, such as:

• Push to a repository.
• Pull requests is merged.
• A workflow is triggered.

In the case of the runner manager, we are interested in workflow_job events that are triggered when a job is queued, running or completed.

Here's a description of the workflow that is triggered by a workflow_job event:

sequenceDiagram
    actor Developer
    Developer ->> GitHub: Pushes code
    GitHub ->> GitHub: Triggers a GitHub Actions workflow
    GitHub ->> Runner Manager: Sends webhook event for workflow jobs
    alt The job status is queued
        Runner Manager ->> Runner Manager: Checks how many runners are authorized to run simultaneously
        alt The maximum number of runners is not reached
            Runner Manager ->> Backend: Request a new runner
            participant Runner as Self-hosted Runner
            Backend ->> Runner: Creates a new runner
            activate Runner
            Runner ->> GitHub: Connects to GitHub Actions
        else The maximum number of runners is reached
            Runner Manager ->> Runner Manager: Try again later
        end
    else The job status is running
        Runner Manager ->> Backend: Update the runner status and metadata
    else The job status is completed
        Runner Manager ->> Backend: Delete the runner
        Backend ->> Runner: Deletes the runner
        deactivate Runner
    end

Health checks

It will also periodically check the health of the runners and perform actions based on their status.

sequenceDiagram
    participant Runner Manager
    participant Redis
    participant Queue
    participant Job
    participant GitHub
    participant Backend
    Runner Manager ->> Redis: Request the list of runners groups
    loop for every runner group
        Runner Manager ->>  Queue: Create a job to check the group's health
        Job ->> Redis: Request the list of runners for the group
        loop for every runner
            Job ->> GitHub: Get the runner status
            Job ->> Backend: Get the runner status
            alt The runner's time to start has expired
                Job ->> Queue: Create a job to re-create the runner.
            else The runner's time to live has expired
                Job ->> Queue: Create a job to delete the runner.
            end
            Job ->> Redis: Update the runner and group information.
        end
        Job ->> Queue: Wait for all created jobs to finish
        Job ->> Redis: Update the runner group information
        alt minimum number of runners is not reached
            Runner Manager ->> Queue: Create a job to create a new runner
        end
        end