Skip to content

Latest commit

 

History

History
196 lines (148 loc) · 7.28 KB

README.md

File metadata and controls

196 lines (148 loc) · 7.28 KB
Raw

lite-lava-docker-compose

This is a Linaro LITE customized fork of LAVA dockerized setup. Instructions for it are available in the wiki.

Below is the original README.

docker-compose

docker-compose file to setup an instance of lava-server and/or lava-dispatcher from scratch. In this setup, every service will be running in a separate container.

Usage

Requirements

In order to use this docker-compose file, you need:

  • docker.io
  • docker-compose

You can install the dependencies using:

apt install docker.io docker-compose

Installing

You just need to fetch the sources:

git clone https://git.lavasoftware.org/lava/pkg/docker-compose
cd docker-compose

Using it

In order to start the containers, run:

docker-compose build
docker-compose up

docker-compose will spawn a container for each services:

  • PostgreSQL
  • apache2
  • gunicorn (aka lava-server-gunicorn)
  • lava-master
  • lava-logs
  • lava-publisher
  • lava-dispatcher
  • ser2net
  • tftpd
  • dispatcher-webserver
  • ganesha-nfs

All the services will be connected to each other.

docker-compose will also create some volumes for:

  • device dictionaries
  • health-checks
  • job outputs
  • PostgreSQL data
  • dispatcher httpd
  • dispatcher tftpd

Backup/restore of the database

Backup

Create the sql backup file:

docker exec --user postgres docker-compose_db_1 bash -c \
            "pg_dump --username=lavaserver lavaserver > /tmp/lavaserver.sql"

Then retrieve it on host:

docker cp docker-compose_db_1:/tmp/lavaserver.sql .

Restore

In order to restore an instance from a backup, first delete the db container and its associated volume:

docker-compose stop
docker container rm docker-compose_db_1; docker volume rm lava-server-pgdata

Then place the .sql (or .sql.gz) file in initdb.d/ folder:

docker-compose -f docker-compose.yaml -f docker-compose-restore-backup.yaml up -d

The file docker-compose-restore-backup.yaml mounts an extra volume in postgres container, in /docker-entrypoint-initdb.d References:

Standalone dispatcher container

Configuration (simple, for QEMU purposes)

All configuration is stored in .env file. Some of the steps are required whilst others are optional.

  • Change DC_LAVA_MASTER_HOSTNAME and DC_LAVA_LOGS_HOSTNAME to <server_name> which points to the running LAVA master instance.
  • (optional) set DC_LAVA_MASTER_ENCRYPT to --encrypt if the master instance is using encryption for master-slave communication.
  • (optional) Create certificates on the slave. sudo /usr/share/lava-dispatcher/create_certificate.py foo_slave_1 This can be done in three ways:
    • by running "docker exec -it docker-compose_lava-dispatcher_1 bash" (for this to work you'd need to build and run the containers first - see below).
    • by using published container images: docker run -v $PWD:/tmp/certs --rm lavasoftware/lava-dispatcher /usr/share/lava-common/create_certificate.py --directory /tmp/certs foo_slave_1
    • alternatively you can create the certificates on system which has LAVA packages already installed.
  • (optional) Copy public certificate from master and the private slave certificate created in previous step to directory dispatcher/certs/ of this project. Currently the key names should be the default ones (master.key and slave.key_secret).
  • Execute make lava-dispatcher; at this point multiple containers should be up and running and the worker should connect to the LAVA server instance of your choosing.
  • Add a new device and set its' device template (alternatively you can update existing device to use this new worker) Example QEMU device template: 
    {% extends 'qemu.jinja2' %}
{% set mac_addr = 'DF:AD:BE:EF:33:02' %}
{% set memory = 1024 %}
    You can do this via XMLRPC, lavacli or REST API (if using version 2020.01 and higher).
  • (optional) If the lab where this container runs is behind a proxy or you require any specific worker environment settings, you will need to update the proxy settings by setting the worker environment You can do this via this XMLRPC API call. In case the worker sits behind a proxy, you will also need to set SOCKS_PROXY=--socks-proxy <address>:port in the .env configuration file Furthermore, you will need to add a proxy settings to the .env file for docker resource downloads (http_proxy, ftp_proxy and https_proxy environment variable).

Note: If the master instance is behind a firewall, you will need to create a port forwarding so that ports 5555 and 5556 are open to the public.

Configuration (advanced, for physical DUT purposes)

If you're setting up a standalone dispatcher container, make sure you go through the above configuration first, it is mandatory for this step. In order to run test jobs on physical devices we will need a couple of additional setup steps:

  • PDU control:
    • The dispatcher docker container will already download pdu scripts from lava-lab repo which you can use in device configuration but if you use custom PDU scripts you need to provide them and copy them into dispatcher/power-control directory; they will be copied into /root/power-control path in the container.
    • If you need SSH keys for PDU control, copy the private key to the dispatcher/ssh directory and the public key on to the PDU
    • SSH config - if there's a need for a specific SSH configuration (like tunnel passthrough, proxy, strict host checking, kexalgorithm etc), create the config file with relevant settings and copy it into dispatcher/ssh dir; it will be copied to /root/.ssh directory on the dispatcher container.
  • ser2net config - update ser2net/ser2net.config with the corresponding serial port and device settings
  • Update/add device dictionary with power commands and connection command
  • Add dispatcher_ip setting to the dispatcher configuration. Alternatively you can use REST API if you are using version 2020.01 or higher:
    • dispatcher_ip: <docker host ip address>
  • Disable/stop rpcbind service on host machine if it's running - docker service nfs will need port 111 available on the host.

Running

In order to start the containers, run:

docker-compose build lava-dispatcher
docker-compose up lava-dispatcher

or, alternatively:

make lava-dispatcher