Setting up a LAVA instance

This repository contains ansible scripts that help set up a LAVA instance on a Debian 9.4 (stretch) machine.

Installation

# Get the repository
git clone http://git.k-space.ee/madislutter/ansible-lava.git
cd ansible-lava

# Put the IP address of your machine into the hosts file.
vim hosts

# Run the lava-install playbook. This will install the necessary packages 
# from stretch-backports. Not using the stretch-backports would result in 
# installing a deprecated version. It will also configure the apache 
# server to serve the LAVA website.
ansible-playbook lava-install.yml

# Create a superuser account with the help of lava-create-user playbook
ansible-playbook lava-create-user.yml

• The next commands need to be run in the target machine.
• If your server doesn't have HTTPS then you need to add these two lines into /etc/lava-server/settings.conf.

  "CSRF_COOKIE_SECURE": false,
  "SESSION_COOKIE_SECURE": false
  

  After editing the file, restart the django service for the changes to take effect.

  service lava-server-gunicorn restart
  

• Now log in through the web interface and create a token at http://<ip-address>/api/tokens/.

lava-tool auth-add http://<username>@<ip-address>/
# Paste the token you created

# The next section will create a shortcut for lava-tool so that we
# wouldn't have to type out the server address for each command.
lava-tool auth-config --default-user http://<username>@<ip-address>/RPC2/
lava-tool auth-config --endpoint-shortcut local http://<username>@<ip-address>/RPC2/

Testing the setup, adding devices

In order to test the setup, let's add a qemu device and a standard test job that should run if everything is set up correctly. The following commands should be run inside the target machine.

# Add the qemu device type
lava-server manage device-types add qemu
# Add the device itself. The last argument is the device name.
lava-server manage devices add --device-type qemu --worker $(hostname -f) qemu01

• Test jobs can only be submitted to devices that have a device dictionary. So we need to add a device dictionary to the created device. Create a file with the following contents and name it qemu-dict.jinja2.

{% extends 'qemu.jinja2' %}
{% set mac_addr = '52:54:00:12:34:59' %}
{% set memory = '1024' %}

• Now add the device dictionary to the device.

lava-tool device-dictionary --update /qemu-dict.jinja2 local qemu01

• Navigate to this address http://<ip-address>/scheduler/jobsubmit and submit this standard test job.

If the job runs and succeeds then the setup is successful.

Debugging

supermin: failed to find a suitable kernel (host_cpu=x86_64)

If the test job is not able to boot the device and the output contains this error then make sure you have a kernel installed in /boot directory.

Could not access KVM kernel module

If the job is not able to boot the device and the output contains:

Could not access KVM kernel module: No such file or directory
failed to initialize KVM: No such file or directory
Connection closed

then make sure your kernel has the kvm module loaded:

lsmod | grep kvm

Sometimes the module is loaded, but the module files themselves aren't mounted in /lib/modules/$(uname -r). This is especially likely if you are running inside a virtual machine. Run this to see if there is a problem with finding the module files:

modprobe kvm

A simple solution to bypass the kvm problem is to disable kvm in the test job definition. Just add no_kvm: true as seen in this test job.

Noteworthy

Creating a superuser account

If you try to create a superuser account with this command:

lava-server manage users add --superuser

then be aware that users created like this don't actually get access to the Django admin area of the site at http://<ip-address>/admin. A user gets access to that only if the account is created with these commands:

lava-server manage createsuperuser --username <username> --email <<e-mail>
lava-server manage changepassword <username>

---

Use tmux for long-running processes

Installing all the packages for the LAVA setup can take a long time. It's possible you'll get a broken pipe during the installation. For this reason it might make sense to SSH into the machine, start a tmux session there and install the necessary packages inside this session. If you get a broken pipe then you'll be able to reconnect to the machine and attach to the same tmux session.

If you installed the packages outside a tmux session and there was a configuration prompt and you then got a broken pipe then you'll need to do this:

# Find the dpkg job and kill it
ps aux | grep dpkg
kill <pid>

# Finish configuring the packages
dpkg --configure -a