What is Testflinger?

Testflinger is a system for orchestrating the time-sharing of access to a pool of target machines.

Each Testflinger system consists of:

• a web service (called just Testflinger) that provides an API to request jobs by placing them on a queue
• per machine agents that wait for jobs to placed on queues they can service and then process them

Jobs can be either fully automated scripts that can attempt to complete within the allocated time or interactive shell sessions.

The Testflinger system is particular useful for sharing finite machine resources between different consumers in a predictable fashion.

Typically this has been used for managing a test lab where CI/CD test runs and also exploratory testing by human operators is desired.

Documentation

You can find more information and documentation on the Testflinger Documentation Page.

Content of this repository

A full deployment of testflinger consists of the following components:

• Testflinger Server: The API server, and Web UI
• Testflinger Agent: Requests and processes jobs from associated queues on the server on behalf of a device
• Testflinger Device Connectors: Handles provisioning and other device-specific details for each type of device
• Testflinger CLI: The command-line tool for submitting jobs, checking status of jobs, and retreiving results

This monorepo is organized in a way that is consistant with the components described above:

└── providers
    ├── server
    ├── agent
    ├── device-connectors
    ├── cli
    └── docs

Github actions

If you need to submit a job to a testflinger server through a Github action (instead, for example, of using the command-line tool), you can use the submit action in a CI workflow. Please refer to the inputs field of the action for a complete list of the arguments that the action can receive.

The corresponding step in the workflow would look like this:

    - name: Submit job
      id: submit-job
      uses: canonical/testflinger/.github/actions/submit@v1
      with:
        poll: true
        job-path: ${{ steps.create-job.outputs.job-path }}

This assumes that there is a previous create-job step in the workflow that creates the job file and outputs the path to it, so that it can be used as input to the submit action. Alternatively, you can use the job argument (instead of job-path) to provide the contents of the job inline:

    - name: Submit job
      id: submit-job
      uses: canonical/testflinger/.github/actions/submit@v1
      with:
        poll: true
        job: |
            ...  # inline YAML for Testflinger job

In the latter case, do remember to use escapes for environment variables in the inline text, e.g. \$DEVICE_IP.

The id of the submitted job is returned as an output of the submit action, so you can use it (if you need it) in any of the subsequent steps of the workflow:

    - name: Display results
      run: |
        testflinger results ${{ steps.submit-job.outputs.id }}"

In this example, submit-job is the step where the submit action is used.

If the submitted job is a reservation job (i.e. includes reserve_data) then setting the poll input argument to true results in modified behaviour: the job is polled only until the reservation phase is complete, instead of waiting for the entire job to complete (which happens when the reservation timeout expires or the job is cancelled). There will be no output to record after the reservation so there is little point in polling and idly occupying the runner. However, please do remember to manually cancel the job after you are done with the reserved device.