Advanced Installation Topics

The basic installation guide should be a good start for most users installing LAVA. For more advanced users, here is much more information and recommendations for administrators.

Requirements to Consider Before Installing LAVA


Be careful with laptop installations, particularly if you are using health checks. It is all too easy for a health check to take the device offline just because the laptop was suspended or without an internet connection at the relevant moment.

Laptops also have limitations on device availability but are routinely used as development platforms and can support QEMU devices without problems.

Virtual Machines

LAVA installations inside a virtual machine (or container) have particular constraints. A QEMU device or container may suffer from being executed within the constraints of the existing virtualization and other devices may need USB device nodes to be passed through to the VM. Depending on the VM, it is also possible that storage space for the logs may become an issue.


Consider the expected load on the master and each of the workers:

  • The workload on the master primarily depends on:

    1. the visibility of the instance,

    2. the number of users,

    3. the average number of jobs in the queue and

    4. the total number of devices attached across all the workers connected to this master.

  • The workload on the worker involves a range of tasks, scaling with the number of devices attached to the worker:

    1. doing a lot of synchronous I/O,

    2. decompression of large files

    3. serving large files over TFTP or HTTP and

    4. git clone operations.

An ARMv7 device can serve as a small master or worker, but SATA support is strongly recommended along with at least 2GB of RAM.


LAVA expects to be the primary virtual host configured on the master. This has improved with V2 but unless your instance is V2-only, you may experience problems or require additional configuration to use LAVA as a virtual host.

Other infrastructure

LAVA will need other services to be available, either using separate tools on the same machines or as separate hardware. This list is not exhaustive.

Remote power control

Automated power control using a PDU is one of the most common issues to be solved when setting up a new LAVA lab. Hardware can be difficult to obtain and configuring the remote power control can require custom scripting. There is no single perfect device for all use cases, and a wide variety of possible solutions may exist to cover your needs. Take the time to research the issues and ask on the lava-users mailing list if you need guidance.

Serial console support

Once more than a handful of devices are attached to a worker it will often become necessary to have a separate unit to handle the serial connectivity, turning serial ports into TCP ports. Bespoke serial console servers can be expensive; alternatives include ARMv7 boards with ser2net installed but the USB and ethernet support needs to be reliable for this to work well.

Network switches

Simple unmanaged switches will work for small LAVA labs but managed switches are essential to use VLANd support in LAVA test jobs and will also be important for medium to large LAVA labs.

Power supply

Use of a UPS will allow the entire lab to cope with power interruptions. Depending on the budget, this could be a small UPS capable of supporting the master and the worker for 10 minutes, or it could be a combination of larger UPS units and a generator.


The master is not the correct place to be building or storing build artefacts. In a busy lab, the extra load may cause issues when workers download large files at job startup. Development builds and creation of files to support the LAVA test should happen on a suitably powerful machine to meet the performance expectations of the CI loop and the developers.

Shelving and racks

While it may be tempting to set up a lab on a desk or test bench, this can very quickly degenerate into a tangled mess as more devices are added. On top of the test devices, switches and other infrastructure, there will be a lot of power cables, network cables and serial cables. For even a small lab of a handful of devices, a set of shelves or a wall-mounted rack is going to make things a lot easier to manage.


For sending user notifications via email there is additional configuration to be done. There needs to be a SMTP server or a mail relay set up (which is out of scope for this document) and LAVA configured to use it.

Create a new YAML configuration file for LAVA like


and put in your configuration there. A minimal configuration will most probably look like this:


See django email settings for a full list of all supported configuration, including TLS and authentication.

Don’t forget to reload LAVA webserver to set configuration active:

$ systemctl reload lava-server-gunicorn.service

Other installation notes

LAVA server branding support

The instance name, icon, link, alt text, bug URL and source code URL of the LAVA link on each page can be changed in the settings /etc/lava-server/settings.conf (JSON syntax):

"INSTANCE_NAME": "default",
"BRANDING_ALT": "Example site",

Admins can include a sentence describing the purpose of the instance to give more detail than is available via the instance name. This will be added in a paragraph on the home page under “About the {{instance_name}} LAVA instance”:

"BRANDING_MESSAGE": "Example site for local testing",
"INSTANCE_NAME": "dev-box",

If the stylesheet or icon is available under the django static files location, this location can be specified instead of a URL:

"BRANDING_CSS": "path/to/style.css",
"BRANDING_ICON": "path/to/image.png",

There are limits to the size of the image, approximately 32x32 pixels, to avoid overlap.

The favicon is configurable via the Apache configuration:

Alias /favicon.ico /usr/share/lava-server/static/lava_server/images/logo.png

Unattended upgrades

Debian provides a package called unattended-upgrades which can be installed to automatically install security (and other) updates on Debian systems. This service is recommended for LAVA instances, but is not part of LAVA itself.

If you plan to use unattended-upgrades, it is a good idea to set up monitoring on your systems, for example by also installing apt-listchanges and configuring email for administrator use. Ensure that the master and all workers are similarly configured, to avoid potential problems with skew in package versions.

Example changes


The default installation of unattended-upgrades enables automatic upgrades for all security updates:

Unattended-Upgrade::Origins-Pattern {


Optionally add automatic updates from the LAVA repositories if those are in use:

Unattended-Upgrade::Origins-Pattern {


Other repositories can be added to the upgrade by checking the output of apt-cache policy, e.g.:

release v=8.1,o=Linaro,a=unstable,n=sid,l=Lava,c=main

Relates to an origin (o) of Linaro and a label (l) of Lava.

When configuring unattended upgrades for the master or any worker which still supports LAVA V1, PostgreSQL will need to be added to the Package-Blacklist. Although services like PostgreSQL do get security updates and these updates are important to apply, unattended-upgrades does not currently restart other services which are dependent on the service being upgraded. Admins still need to watch for security updates to PostgreSQL and apply the update manually, restarting services like lava-master, lava-server and vland afterwards:

Unattended-Upgrade::Package-Blacklist {

Email notifications also need to be configured.

Unattended-Upgrade::Mail "";

Unattended-Upgrade::MailOnlyOnError "true";

With these changes to /etc/apt/apt.conf.d/50unattended-upgrades, the rest of the setup is as described on the Debian wiki.

Configuring event notifications

Event notifications must be configured before being enabled.

  • All changes need to be configured in /etc/lava-server/settings.conf (JSON syntax).

  • Ensure that the EVENT_TOPIC is changed to a string which the receivers of the events can use for filtering.

    • Instances in the Cambridge lab use a convention which is similar to that used by DBus or Java, simply reversing the domain name for the instance (e.g. org.linaro.validation)

  • Ensure that the EVENT_SOCKET is visible to the receivers - change the default port of 5500 if required.

  • Enable event notifications by setting EVENT_NOTIFICATION to true

When changing the configuration, you should restart the corresponding services:

$ sudo service lava-publisher restart
$ sudo service lava-master restart
$ sudo service lava-scheduler restart
$ sudo service lava-server-gunicorn restart

The default values for the event notification settings are:

"EVENT_TOPIC": "org.linaro.validation",
"INTERNAL_EVENT_SOCKET": "ipc:///tmp/",
"EVENT_SOCKET": "tcp://*:5500",

The INTERNAL_EVENT_SOCKET does not usually need to be changed.

Services which will receive these events must be able to connect to the EVENT_SOCKET. Depending on your local configuration, this may involve opening the specified port on a firewall.

Events and network reliability

With the default configuration, LAVA will publish events to the EVENT_SOCKET only, using a zmq PUB socket. This type of socket is efficient for publishing messages to a large audience. However, in case of a network breakage, the connection may be lost and events may be missed.

For more reliable event publication on an unreliable network (like the Internet) with a small set of known listeners, you can also use EVENT_ADDITIONAL_SOCKETS. The publisher will connect to each of the endpoints in this list using a zmq PUSH socket. These sockets are configured to keep a large queue of messages for each of the endpoints, and will retry to deliver those messages as necessary. No messages will be lost until the queue overflows.

PostgreSQL Port configuration

In the majority of cases, there is no need to change the PostgreSQL port number from the default of 5432. If a change is required, edit /etc/lava-server/instance.conf and then restart the LAVA master and UI daemons:

$ sudo service lava-server-gunicorn restart
$ sudo service lava-master restart

Configuring the LAVA UI

Initial settings for a LAVA instance change over time as the requirements change and dependencies improve internal security implementations. Most instances will need some adjustment to the apache configuration for the main LAVA UI in /etc/apache2/sites-available/lava-server.conf and LAVA does not attempt to patch this file once admins have made changes. Admins therefore need to subscribe to the lava-announce mailing list and make changes using separate configuration management.

Gunicorn3 bind addresses

Work is beginning to extend the Docker support to have different parts of LAVA in different containers. Some parts of this are easier to implement than others, so the support will arrive in stages.

gunicorn3 supports changing the bind address which will allow to run the lava-server-gunicorn service to run alone in a container while having the reverse proxy in another container. The bind address and other gunicorn3 options can be changed by editing: /etc/lava-server/lava-server-gunicorn

Apache proxy configuration

Django requires the allowed hosts to be explicitly set in the LAVA settings, as a list of hostnames or IP addresses which LAVA is allowed to use. If this is wrongly configured, the UI will raise a HTTP500 and you will get information in the output of lava-server manage check --deploy or in /var/log/lava-server/django.log. For example, /etc/lava-server/settings.conf for contains:


It is also important to enable ProxyPreserveHost in /etc/apache2/sites-available/lava-server.conf:

ProxyPreserveHost On

In some situations, you may also need to set USE_X_FORWARDED_HOST to True in /etc/lava-server/settings.conf

Apache headers

Browser caching can be improved by enabling mod_header in Apache to allow LAVA to send appropriate cache control headers as well as mod proxy and mod proxy_http:

$ sudo a2enmod header
$ sudo a2enmod expires
$ sudo service apache2 restart

Banning badly behaved bots

Despite setting robots.txt, LAVA instances can sometimes come under high load due to badly behaved web crawling bots. Typically, this will show up as an unusually slow LAVA UI and large apache log files showing a lot of unusual activity. For example, recursively retrieving every query string variation for every table or trying to access every possible URL without being signed in.

To control these bots, the DISALLOWED_USER_AGENTS setting can be extended. By default, LAVA blocks yandex, SemrushBot, bing and WOW64. The list can be extended in /etc/lava-server/settings.conf. If you do extend the list, please let us know by subscribing to the lava-devel mailing list and posting your updated list.

Tracking errors

In order to track server errors, admins can enable sentry error tacking. In /etc/lava-server/settings.conf add:

"SENTRY_DSN": "https://<public_key>@<server>/<project_id>"

When any of the services is crashing, an error will be recorded along with some metadata.

Configuring default table length

The LAVA UI mainly consists of tables. The length of each table can be configured by the user right above it (“Show 10..100 entries”). A default value for the table length can be set in /etc/lava-server/settings.conf:


Configuring submitter full name

The LAVA main job page by default will show user name of submitter. Admin could set if to show full name of submitter in /etc/lava-server/settings.conf:



As LAVA is based on Django framework, localisation is available. To adjust date and time to local format (including DST adjustment), configure settings accordingly in a new config file, e.g. /etc/lava-server/settings.d/02-l10n.yaml:

TIME_ZONE: "Europe/Berlin"

Without this, time will be american format and all times displayed in UTC.

See django documentation for details and a list of possible TIME_ZONE values.

It is possible to set LANGUAGE_CODE to another value than the default en-us, but this will only affect administration interface as LAVA does not support translation to other languages.

Controlling the Django Admin Interface

Some instances may need to allow selected users to be Django superusers to provide access to the Django Admin Interface. Some of the features of the interface need very careful handling, especially the deletion of database objects.

Deleting database objects is always a problem and needs careful consideration after looking at all the relevant logs. There are complex inter-relationships not just in the UI but also in the scheduler, logging support and publishing support. UI errors and scheduling errors can be caused by inappropriate deletion of database objects and recovery from these situations can be complex and may involve a complete restoration of the instance from backups.

Admins can choose to disable access to the Delete button in critical areas of the admin interface by adding a setting to /etc/lava-server/settings.conf:


This disables the Delete button and the delete action for selected database objects, particularly Device, TestJob, TestCase, TestSuite and TestSet. None of these objects should be deleted in the admin interface (helpers exist to delete using the command line interface, with suitable safeguards).

Restart the lava-server-gunicorn service each time /etc/lava-server/settings.conf is modified.

Configuring log file display

By default, test logs will be formatted and displayed for easy viewing in the browser and this should work fine for the majority of users. However, if your test jobs are creating very large test logs it can cause problems when trying to display them. Depending on network and client configuration, this might show up as timeouts when viewing or maybe error codes like “502 Proxy Error”. If this is a problem, there is an option to control the maximum size of test log that will be displayed; any log larger than this will instead just be offered for direct download.

Edit /etc/lava-server/settings.conf (JSON syntax) to set the limit, in MiB. For example:


will limit the maximum display size to 10 MiB. To find the right size for your needs, check on the sizes of the output.yaml log files on your lava-master server.

Some test jobs are generating a lot of test cases. For such test jobs, rendering the log page could be really slow while the server query the database for every test case ids. In order to improve the page rendering speed, if more than TESTCASE_COUNT_LIMIT (10000 by default) test cases exists for a job, the server will not query the test case table. The logs will still be visible, but without the links to the result pages.

You can change the default value by editing /etc/lava-server/settings.conf:


Extending the schema white list

Since LAVA 2019.04, the keys that can be used in the job definition context dictionary is restricted. Admins can extend this white list by updating EXTRA_CONTEXT_VARIABLES in the settings:

Add to /etc/lava-server/settings.conf:

"EXTRA_CONTEXT_VARIABLES": ["custom_variable1", "variable_2"],