Adding your first devices

Django administration interface

The django admin interface is a core component of the django framework. Elements can be added for particular implementations but, fundamentally, the operation of the interface is the same as other django sites. The appearance of the menu is determined by the version of django installed on your system. The style changed substantially in django 1.9, so images of the interface itself are not included here.

Changes within the administration interface and changes made as a superuser through the UI are tracked through the History elements of objects in the database. When viewing a specific element (for example a single Test Job or a single Device), click on the History link to view all changes relating to that element. There is also a link back to the UI view of the same element.


the organisation, layout and content of the django administration interface is subject to change with upgrades to django itself and these changes are outside the control of LAVA.

Start with a known device type

It is tempting to jump straight in with your one-off special device which nobody else has managed to automate yet but a fresh install needs to be tested with a known working configuration. Setting up known working devices and learning how to modify the first job is essential to deciding how to best configure a new device. It is also recommended to setup another known device type which is similar to the device you want to add as there are different steps required for certain types of device.

This first QEMU device can be configured on the existing worker which is always available on the master. Subsequent devices can be added to other workers and devices can be shuffled between workers, subject to limitations of physical connections.


QEMU is always recommended as the first device to be set up on any LAVA instance for a few reasons:

  1. QEMU requires no external hardware or software configuration (until a network bridge becomes desirable)
  2. QEMU requires only a minimal device dictionary.
  3. Test images for use with QEMU are readily available and relatively easy to modify.

Create a Device Type

Prior to adding any devices, admins should add suitable device types to the database. The device type name should match a jinja2 template file in:


If an existing template does not exist, a new template will need to be created.

See also

Device types

You can then either use the web admin interface or the lava-server command line to add device types.

Using the admin interface

In order to use the web admin interface, log in to the LAVA instance and click on your username to see the Profile menu.


The django administrative interface is accessed from the Administration link in the profile menu.

  1. Scroll down to the group labelled LAVA_SCHEDULER_APP.
  2. Click on Device types

Just before you add the device type, take a look at the available elements of a device type:

  • Name
  • Has health check
  • Display
  • Owners only
  • Health check frequency
  • Architecture name
  • Processor name
  • CPU model name
  • List of cores
  • Bit count

The only value needed for the QEMU device type is the Name, just check that Display is the default: enabled. Now Save.

Using the command line

On the command line, you can add device types (for instance QEMU and panda) using:

lava-server manage device-types add qemu panda

It’s also possible to add all known device types at the same time with:

lava-server manage device-types add '*'

Descriptive fields like Architecture name, Processor name, CPU model name, List of cores and Bit count cannot be set on the command line.

Using the command line interface it’s also possible to list all known device types:

lava-server manage device-types list --all

Create a device in the database

Using the admin interface

  • Navigate back to LAVA_SCHEDULER_APP and select Devices and Add Device.
  • Select the QEMU device type from the list.
  • Give your device a name
  • Select the worker from the list.
  • Set the Device owner (typically one of the superusers).
  • Your first device should be public.
  • Ensure that the device is enabled as a Pipeline device.

Using the command line

Using the command line interface it’s also possible to list all known device types:

lava-server manage add-device --list

On the command line, you can add device types (for instance a QEMU type device with a hostname qemu01) using:

lava-server manage add-device --device-type qemu qemu01

See lava-server manage help add-device for more options, including initial states of the device in the database.

Adding a dictionary to the first QEMU device

For the first device, a simple device dictionary can be used to provide the device-specific details on top of the template:

{% extends 'qemu.jinja2' %}
{% set mac_addr = '52:54:00:12:34:59' %}
{% set memory = '1024' %}
  • The device dictionary must extend an existing template.

  • The architecture (arch value) is not set in this device dictionary. This allows this device to run test jobs using files for any architecture supported by QEMU.

  • The MAC address needs to differ for each device of this type across all instances on the same subnet.

  • The available memory for the virtual machine is set in megabytes. This can be changed later to balance the requirements of test jobs with performance on the worker.

  • Line ordering within the device dictionary is irrelevant, although it is common to put the extends line first when storing the dictionary as a file.

The template itself lives in:


This dictionary does not include a setting to use a tap device which means that this device would not support a hacking session inside the virtual machine. Setting up a bridge is out of scope for this documentation.

See also

Creating a device dictionary for the device to export and modify a device dictionary, Updating a device dictionary to import a device dictionary into the database for use with a new or existing device, Checking your templates for help with types of devices other than QEMU and Device type templates for help with how the device dictionary works with the device-type templates.

Once updated, the device dictionary is added to the Device view in the administrative interface under the Advanced Properties section at the base of the page.

Adding other devices of known device-types

The core principles remain the same as for QEMU, the main differences are in the way that the device dictionary is needed to provide a wider range of settings covering power control, serial connections, network details and other values.

See also

Writing Health Checks for devices - each time a new device type is added to an instance, a health check test job needs to be defined.

Check existing instances

Templates usually exist for known device types because an existing instance is using the template. Often, that instance will be Linaro’s central validation lab in Cambridge which is accessible via .

The contents of the device dictionary for particular devices are visible to anyone with access to that device type, using the device detail page. Details of the jinja2 files used to update the device dictionary on Linaro instances is also held in git:

The structure of the device dictionary files will be similar for each device of the same type but the values will change. An example for a beaglebone-black device looks like:

{% extends 'beaglebone-black.jinja2' %}
{% set connection_command = 'telnet localhost 7101' %}
{% set hard_reset_command = '/usr/local/lab-scripts/snmp_pdu_control --hostname pdu15 --command reboot --port 11' %}
{% set power_off_command = '/usr/local/lab-scripts/snmp_pdu_control --hostname pdu15 --command off --port 11' %}
{% set power_on_command = '/usr/local/lab-scripts/snmp_pdu_control --hostname pdu15 --command on --port 11' %}


It is recommended to keep the device dictionary jinja files under version control. The templates are configuration files, so if you modify the default templates, those need to be under version control as well.

Migrating V1 devices to V2 devices

If you have a working V1 configuration, this can be migrated to the V2 requirements. Devices can support both models during the migration, admins can choose to make some devices exclusive to V2 at any time before the V1 code support is removed.

Adding new device types


This is the most complex part and it can be a lot of work (sometimes several months) to integrate a completely new device into LAVA. V2 offers a different and wider range of support to V1 but some devices will need new support to be written within lava-dispatcher. It is not always possible to automate a new device, depending on how the device connects to LAVA, how the device is powered and whether the software on the device allows the device to be controlled remotely.

The integration process is different for every new device. Therefore, this documentation can only provide hints about such devices, based on experience within the LAVA software and lab teams. Please talk to us before starting on the integration of a new device using the Mailing lists. Include full details of the type of device, the bootloader specifications, hardware support and anything you have done so far to automate the device. Sometimes, the supplied bootloader must be modified to allow automation. Some devices need electrical modifications or specialised hardware to be automated.

Integrating a new device type will involve some level of development work, the device type templates are more than configuration. Testing new device type templates requires setting up a developer workflow and running unit tests as well as running test jobs on a LAVA instance. If the new device type involves a new boot or deployment method, there will also need to be changes in the lava-dispatcher codebase. New elements of the test job submissions and device configuration may also need changes to the schema in lava-server. Some new device types will be a lot easier than others - for example UBoot tends to have a reasonably consistent interface across multiple devices, so changes for a new UBoot device could be as little as setting variables after extending the base-uboot.jinja2 template.

LAVA encourages new device type templates to be contributed upstream as a community contribution to LAVA.


The LAVA software and lab teams have built up a set hints relating to the integration of new device-types. The further a device deviates from one or more of these hints, the harder it will become to automate such a device. Always remember that the way that the device is supported must scale to large labs which already contain a range of other devices, each with their own issues. It is not acceptable to add a new device-type which is incompatible with devices which are already supported or which imposes restrictions on how many devices of any type can be used in any one lab.

  • serial port - LAVA expects to automate devices by interacting with the serial port immediately after power is applied to the device. The bootloader must interact with the serial port. If a serial port is not available on the device, suitable additional hardware must be provided before integration can begin. All messages about the boot process must be visible using the serial port and the serial port should remain usable for the duration of all test jobs on the device.

  • UBoot - if the device supports UBoot then this is a useful beginning. However, the build of UBoot on the device can hinder integration due to the wide range of configuration options and behavioural changes available inside a patched UBoot build. Generally, the more components of UBoot that are disabled or removed from a vendor build, the harder it will be to integrate. If you are able to fully script a UBoot process from interrupting the bootloader to booting a kernel of your own choice, this will greatly assist in integrating the device into LAVA.

    1. Configuration - ensure that the UBoot build supports a string which can be used to interrupt UBoot and that once interrupted, the prompt is set to a usable string like => or uboot# etc. Make sure that the configuration supports TFTP using commands sent over the serial port. The timeout for interrupting the boot process must be configurable.
  • Android - LAVA relies on ADB and fastboot to control an Android device. Support for ADB must be enabled in every image running on the device or LAVA will lose the ability to access, reboot or deploy to the device. The fastboot serial number must be unique and modifiable by the admin in case an existing device is already using that number. The device needs a reliable way to enter fastboot mode from power on. Typically, if the boot partition is erased, this will force the bootloader into fastboot mode. The device needs to implement the fastboot boot <boot img> command, so that the test image can be loaded directly into memory and executed.

  • Battery Power - devices which have internal batteries become difficult to reliably automate for a few issues, unless the battery can be permanently removed:

    1. forced reboots become impossible without electrical modification of the device to temporarily take the battery out of circuit. This means that it is much easier to cause the device to go offline because of a broken kernel build or broken image.
    2. recharging can be an issue - devices may not behave normally when held in fastboot mode or with a broken kernel build or image deployed to the system. This can cause the device to fail to keep charge in the battery or fail to recharge the battery, despite having power available.
  • Serial power leaks - some devices are capable of drawing power over the serial line used to control the device, despite the actual power supply being disconnected. Sometimes this requires a period of time to discharge capacitors on the board (fixable by adding a sleep in the power_off_command). Sometimes this power leak can cause the device to latch into a particular bootloader mode or other state which prevents the automation from proceeding.

    See also

    Power Commands

  • Networking

    1. Ethernet - all devices using ethernet interfaces in LAVA must have a unique MAC address on each interface. The MAC address must be persistent across reboots but also be modifiable by admins in case it does conflict with some other board in the same lab. No assumptions about fixed IP addresses, address ranges or pre-defined routes. If more than one interface is available, the boot process must be configurable to always use the same interface every time the device is booted.
    2. WiFi - is not currently supported as a method of booting devices.