Migrating to LAVA V2¶
Understand the objectives of the current test job.
Revise the Test Plan.
Use protocols and/or device configuration for any operation which is more than typing commands into a shell. If the V1 job required MultiNode to be able to install custom software necessary to interact with the device, these commands should be executed using the LXC protocol which removes the need for MultiNode.
Test Jobs which used MultiNode in V1 in this way will also include calls to the MultiNode API and these will need to be removed from the test definitions.
Use comments - the pipeline uses YAML submissions, so describe what the job is trying to do and why in comments - just start the line with
Simplify the test job to make best use of the new functionality.
Master images are not necessary - the test writer can implement fallbacks to known working images for specific tasks.
Start with Gold standard test jobs
Look for V2 jobs running on the same or similar device types
Bundles no longer exist and there is no submit action. Results are reported live as the test job runs. Results can be downloaded or exported using the results API.
If the results of your V1 jobs were being retrieved or processed by another service, ensure that the calls are adapted to the new results API.
Filters and Image Reports (including Image Reports 2.0) have been replaced by Queries and Charts.
Certain device-types have changed. In particular, to start with the simplest V2 test job, you will typically be submitting a job to the
qemudevice-type instead of
Specific information and guides¶
Migrating to LAVA V2 is not a trivial process. It is recommended to start by understanding how simple test jobs run, including starting from submitting your first V2 test job.
- Submitting your first job
- Test Writer use cases
- Writing Tests
- Gold standard test jobs
- Introduction to Results in LAVA
- Understanding the Pipeline - an administrator guide
- Device type templates
- Device dictionary
- Information sources
- Extra device configuration
- Extra dispatcher configuration
- Requirements for a LAVA device
- Adding support for a device of a known type
- Obtaining configuration of a known device
- Creating a new device entry for a known device type
- Creating a device dictionary for the device
- Updating a device dictionary
Understanding test job objectives¶
The migration to a pipeline job is not a simple case of reformatting JSON. There is no direct equivalence between the old JSON and the pipeline YAML. It is imperative that the test writer fully understands the reasons why the old JSON job does the things it does before starting to migrate that job to the pipeline. The migrated test job needs to provide equivalent validation to the V1 test job but that does not mean that the test job needs to work in the same or even a similar manner.
Is there pipeline support for the test device type?
How does the test job deploy the test files?
Does the V1 test job use a
master image? This support has been replaced in V2. In general, devices simply boot directly from the bootloader. Where this is problematic, test jobs can use use Secondary media.
How does the test job boot the device?
Does the V1 test job use
linaro-media-createor download a
hwpack? This support has been removed and the files provided to the device using V2 will need to be in the form which the device can use directly.
Which test definitions can be joined into a single test action, which will need a simple reboot and which might need another deploy step?
Which test definitions use
lava-test-case-attach? These will need to be ported to the Publishing API.
The pipeline support is changing the way that the test jobs run so that some device types change and some multinode jobs can become single node. e.g. the old KVM device type is now QEMU, with the architecture of the QEMU binary specified by the test job. Equally, the availability of LXC and other protocols means that there is often no need for a KVM to run ADB or other components.
Viewing the available support¶
The XML-RPC call
provide the full configuration for a device, including all the deployment and
boot methods supported by that device and device-type on that instance.
The Device type detail page has a Support tab which contains details of the methods supported by devices of this device type.
Not all devices of any one device type necessarily support all methods of that type. e.g. some methods require additional hardware to be fitted. Check the device tags and other jobs running on devices of the same type.
The deployments for a LAVA V2 test job determine how the test files and the LAVA test shell overlay files are delivered to the device prior to boot. This can include making the files available for TFTP or copying the files over to a running device using SCP.
The deployment method chosen must match the boot method requirements and it is up to the test writer to ensure that a valid combination of deployment and boot methods are used.
The boot method dictates how the device is instructed to boot. Depending on the method, this will usually also involve how to deliver the deployed files to the device, e.g. using TFTP, as well as controlling the commands sent to the device to set boot options and commands.
The MultiNode Protocol is common to all device types. Other protocols may require specific configuration for the device type or not be available for all devices or on all instances.