Writing Tests

The LAVA dispatcher currently supports tests being run in a number of ways:

The most common style of test currently in use is the Lava-Test Test Definition 1.0 using lava-test-case. During the deploy action, an overlay is added to the DUT filesystem including the test writer commands and LAVA helper scripts. The test action uses a helper script to execute the test writer commands and other helper scripts are used to report test results back to the dispatcher, wrapping results in special marker text to allow for easy identification. The dispatcher parses out those test results and reports them alongside the test job log output. Test Shell Definitions contain individual inline commands or references to repositories to deploy custom scripts using a variety of programming languages, according to the support available on the DUT.

Introduction to the LAVA Test Developer Guide

This guide aims to enable users to be able to

  1. Submit desired jobs/tests on targets deployed where the LAVA server is located and report results.

  2. Understand how test job files need to be written so that jobs get submitted properly.

  3. Understand the options for running the test operation. No one test method can suit all test operations or all devices.

  4. Understand how test shell definitions need to be written, depending on how the test should be executed.

Pay particular attention to sections on:

Guide Contents

Assumptions at the start of this guide

  1. The desired board is already configured for use with a LAVA Server instance.

  2. A user account (username, password, email address) is already created by a LAVA administrator on your behalf, with permissions to submit jobs.

  3. lavacli is already installed on your test system and a suitable authentication token has been added.

  4. You are familiar with submitting jobs written by someone else, including viewing the logs file for a job, viewing the definition used for that job and accessing the complete log.

To install lavacli, see lavacli.

To authenticate lavacli, see Authentication Tokens.

To find out more about submitting tests written by someone else, see Submitting your first job.

To find out more about viewing job details, see Job Submission.

Checking device availability

Use the LAVA scheduler to view the device types and devices available in your LAVA instance. The main scheduler status page shows data for each device type as well as the currently active jobs. Also check the Devices pages:

  • All Devices - includes retired devices to which jobs cannot be submitted.

  • All Active Devices - lists only devices to which jobs can be submitted

  • All Devices Health - limited to just the latest health status of each device.

  • My Devices - available from your profile menu by clicking on your name once signed into the instance.

For a MultiNode job, you may need to check more than one device type.

LAVA looks at the device health when working out if a particular device is available for a new job:

  • Good, Unknown - jobs can be submitted OK.

  • Restricted - only specific users may submit jobs.

  • Retired - this device is not available; jobs will be rejected if all devices of this type are retired.

Finding an image to run on the device

Typically, the easiest thing to do here is to start with an image which is already in use in LAVA. You can find one of these images by checking the device type in LAVA and viewing some of the jobs for devices of this type from the table on that page. e.g. for QEMU devices on validation.linaro.org:


Actions to be run for a LAVA test

There are three important sets of actions that will normally be run for a LAVA test:

  1. Deploy: The actions needed to set up a device to boot a test image. Each device type may support a range of different deployment methods.

  2. Boot: The steps to follow to start the test image on the device. Each device type may support a range of different boot methods.

  3. Test: Run the lava test definition, running the specified tests. All methods use the test action. Syntax varies according to the method chosen.

Example of Lava Test

This example will use syntax for the Lava-Test Test Definition 1.0 as well as covering device tags and checksums which may be useful for all test jobs.

Deploying a pre-built QEMU image

  - deploy:
        minutes: 5
      to: tmpfs
            image_arg: -drive format=raw,file={rootfs}
            url: https://files.lavasoftware.org/components/lava/standard/debian/stretch/amd64/2/stretch.img.gz
            compression: gz

Using device tags

A device tag marks a specified device as having specific hardware capabilities which other devices of the same device type may not. To test these capabilities, a test job can specify a list of tags which the device must support. If no devices exist which match all of the required tags, the job submission will fail. If devices support a wider range of tags than required in the test job (or the test job requires no tags), any of those devices can be used for the test job.


Test jobs which use device tag support can only be submitted to instances which have those tags defined and assigned to the requested boards. In your LAVA instance, check the device information to see what tags are used.

When writing a normal single-node test job, the desired tags should be listed as a top level list of strings in the job definition, i.e. at the same level as job_name, timeouts, metadata and device_type:

# Your first LAVA JOB definition for an x86_64 QEMU
device_type: qemu
job_name: QEMU pipeline, first job

- tap_device
- virtual_io

    minutes: 15
    minutes: 5
priority: medium
visibility: public

# context allows specific values to be overridden or included
  # tell the qemu template which architecture is being tested
  # the template uses that to ensure that qemu-system-x86_64 is executed.
  arch: amd64

  # please change these fields when modifying this job for your own tests.
  docs-source: first-job
  docs-filename: qemu-pipeline-first-job.yaml

For MultiNode test jobs, the tags are defined as part of the MultiNode protocol block:

        device_type: qemu
          arch: amd64
        count: 1
        # In this example, only one role in the group uses tags
        - tap_device
        - virtual_io
        device_type: qemu
          arch: amd64
        count: 1
      seconds: 60

Device tags are only relevant during scheduling of the test job and have no meaning to the dispatcher once the job is running.

Using checksums

If an MD5 or SHA256 checksum is provided alongside the URL of the file to be used in a test job, the downloaded content will be checked against the provided checksum. The test job will fail as Incomplete if the checksum fails to match.

Avoid using URLs which include shortcuts like latest when providing the checksum. Specify the full URL to ensure consistency between tests.

Using Lava-Test Test Definition 1.0

The Lava-Test Test Definition 1.0 action provides a way to employ a black-box approach to testing on the target device. Its format is:

- test:
    failure_retry: 3
    name: kvm-basic-singlenode
      minutes: 5
        - repository:
                format: Lava-Test Test Definition 1.0
                name: smoke-tests-basic
                description: "Basic system test command for Linaro Ubuntu images"
                    - printenv
          from: inline
          name: env-dut-inline
          path: inline/env-dut.yaml
        - repository: git://git.linaro.org/lava-team/lava-functional-tests.git
          from: git
          path: lava-test-shell/smoke-tests-basic.yaml
          name: smoke-tests
        - repository: https://git.linaro.org/lava-team/lava-functional-tests.git
          from: git
          path: lava-test-shell/single-node/singlenode03.yaml
          name: singlenode-advanced

The definitions list here may contain multiple test definition URLs. These will all be run sequentially in one run on the test device, and it will not be rebooted between the definitions.

See also

lava_test_shell developer documentation