Using LAVA with KernelCI

How to use LAVA with KernelCI

Any LAVA lab with a publicly available API can be added to Or, if you have a private instance of KernelCI, you can set it up locally too. The KernelCI core tool kci_test can be used to generate test job definitions and submit them. Then the kernelci-backend can receive callback notification HTTP requests directly from LAVA and add the test results to the database.

Setting up a LAVA lab

A distributed LAVA lab is composed of a main server with a web frontend, and a number of dispatchers which have a direct connection to the devices running tests. For KernelCI development purposes, typically a self-contained LAVA lab can be set up with a server and a single dispatcher with at least one QEMU device instance.

LAVA can be installed natively using Debian packages as per the LAVA documentation. Another convenient way to do this is using the KernelCI LAVA Docker containers.

Steps to add a LAVA lab to

Once you have a LAVA lab up and running, follow these steps to start running tests:

  1. Create a kernel-ci user as per the LAVA documentation
  2. Create a LAVA API authentication token as per the LAVA documentation for the kernel-ci user. It will be used for submitting jobs using kci_test, either manually or automatically via Jenkins.
  3. Create another LAVA API authentication token for the kernel-ci user with the description set to kernel-ci-bisection-webhook for automated bisection via Jenkins.
  4. Create a GitHub pull request to add your lab definition to lab-configs.yaml.
  5. Optionally, create another GitHub pull request to add device definitions in test-configs.yaml if your lab has new device types not already covered by
  6. A maintainer will get in touch with you to provide the LAVA API tokens and give you some KernelCI backend API tokens for your lab. They will be used for LAVA callback notifications.
  7. Store the backend API tokens for in LAVA for the kernel-ci user using kernel-ci-callback and kernel-ci-callback-staging respectively as their descriptions.

Your lab will then start being used on to verify it’s all working as expected. Jobs will be scheduled automatically every 8h, and the results will start appearing on the web dashboard and in email reports. Then it will normally get enabled in production to run tests for all the trees and test suites enabled for your lab (or everything by default).

Last modified August 4, 2021