- 06 Jul, 2022 3 commits
-
-
-
Logging formatter that outputs json Configuration options: * extra_keys: List[str] - extra keys to output to generated json Example config: ``` formatters: json_formatter: (): lava_common.logging_utils.json_formatter extra_keys: - sql - duration - alias ``` -
New configuration priotity: 1. Passed arguments such as `--log-file` 2. LOGGING variable The systemd files had to be changed to allow LOGGING configuration be the default. The `--log-file` and `--level` arguments can be passed with no arguments which is equal to not passing them at all.
-
- 30 Jun, 2022 37 commits
-
-
-
Generally we're standardizing on ansible as a tool to deploy $things; Whether that's to kubernetes or otherwise. Switch CI to deploy the staging setup via the lava-server-staging playbook in the ansible core recipes. Signed-off-by:
Sjoerd Simons <sjoerd@collabora.com> -
To allow qemu devices both inside and outside to operate mosty identically, especially with respect to netdevice configuration, the docker network environment should match the host network environment. To achieve that tell docker to set the network to host and add the NET_ADMIN capability. On top of that to support the qemu bridge netdevices add the tun device from the host. Fixes: #548 Signed-off-by:
Sjoerd Simons <sjoerd@collabora.com>
Upstream: https://git.lavasoftware.org/lava/lava/-/merge_requests/1778 -
Match messages like: 'Error code <number>' Signed-off-by:
Ricardo Cañuelo <ricardo.canuelo@collabora.com> -
This commit adds error message detection in the final bootloader-commands action. Closes: #308
-
Improves performance because it avoids extra queries for different device types. Comparison: Before: 10010766 function calls (9764524 primitive calls) in 4.330 seconds After: 9476569 function calls (9237052 primitive calls) in 3.910 seconds Upstream MR: https://git.lavasoftware.org/lava/lava/-/merge_requests/1769
-
-
-
-
-
Breaking changes to query parameters and JSON object
-
-
Add list of error message patterns for the Depthcharge boot method. Allow users to redefine the errors through depthcharge_error_messages. Signed-off-by:
Laura Nao <laura.nao@collabora.com> -
Paweł Wieczorek authored
This patch is intended to prevent CI pipeline flakiness caused by the changes from the container image publisher. Script environment is expected to be Alpine hence the specific base image. DinD service does not need to be bound to a specific base image. Signed-off-by:Paweł Wieczorek <pawiecz@collabora.com> -
Paweł Wieczorek authored
By default "docker-compose up" picks up any changes in service configuration by stopping and recreating the containers. This behaviour caused recreation of LAVA server and database when worker token was passed to the dispatcher configuration even though these services are not affected by this change. Signed-off-by:Paweł Wieczorek <pawiecz@collabora.com> -
Paweł Wieczorek authored
Compose plugin is now available by default on the DinD image [0]. Version was bumped from 1.29.2 to 2.5.0. Default field separator has been changed from '_' to '-'. [0] https://github.com/docker-library/docker/commit/9aaa75893c70431f1302b0608af47ee26d67edf2 Signed-off-by:
Paweł Wieczorek <pawiecz@collabora.com> -
Add device type template for the Acer CB514-2H spherion Chromebook. Signed-off-by:Laura Nao <laura.nao@collabora.com> -
Creating volume as a separate step is not required [1]. This way tracking lavacli volume name so that only single volume is created is no longer necessary. No additional ones would be created in case of a typo. [1] https://docs.docker.com/storage/volumes/#start-a-container-with-a-volume Signed-off-by:
Paweł Wieczorek <pawel.wieczorek@collabora.com> -
Paweł Wieczorek authored
Entrypoint script for dispatcher ("start-worker.sh") expected "LOG_LEVEL" instead of "LOGLEVEL" in its environment. This patch also fixes "LOG_LEVEL" flag: "lava-worker" accepts "--level". Signed-off-by:Paweł Wieczorek <pawiecz@collabora.com> -
Paweł Wieczorek authored
Having more than one hundred devices running jobs at the same time and pushing log statements generates high traffic to the server. Increasing logger time limit could help to reduce the number of HTTP requests. Signed-off-by:Paweł Wieczorek <pawiecz@collabora.com> -
Paweł Wieczorek authored
This patch is a temporary work-around (famous last words). It should be replaced with a restarting policy once LAVA services are separated in Collabora deployment. Signed-off-by:Paweł Wieczorek <pawiecz@collabora.com> -
Paweł Wieczorek authored
From LAVA release 2022.01 onward requirements of package "lava-dispatcher-host" cannot be resolved on Debian Bullseye without backports. Affected requirement is "python3-bpfcc" (>= 0.21) which is available in Bullseye as 0.18.0+ds-2 (and as 0.22.0+ds-2~bpo11+1 in backports) as of writing this patch. Patch that introduced this requirement: 180cb3c1 APT repository on lavasoftware.org supplies updated "python3-bpfcc" (0.21.0+ds-7~bpo11+1~lava1). Signed-off-by:
Paweł Wieczorek <pawiecz@collabora.com> -
With this and adjusted gitlab variables changes in staging will be published in a the staging lava repository published at: deb https://repositories.collabora.co.uk/lava/staging buster main While the production repository will still publish to the sysadmin debian repository Signed-off-by:
Sjoerd Simons <sjoerd@collabora.com> -
-
The base docker image we use configures dpkg to not install most files to /usr/share/doc/, which makes sense in general. However, it is from there that LAVA obtains the files it uses to show for its 'Help' section, so we override the configuration to include the files installed by lava-server-doc.
-
Only deploy it from the staging branch though; As that will get an actual dispatcher attached and our development cluster isn't big enough for a lot of deployments...
-
At startup we can simply add all of the device types that are supported, rather than requiring users to go through that additional step when adding a new device.
-
Rather than sending the logs to a file in /var/log, make them all go through stderr (the default used by logging.StreamHandler), so that it can all be seen by using docker logs.
-
At startup, we now create the directories used by configuration for when the setup is not using an overlaid docker volume, but an empty host directory. The commands are idempotent so should always work.
-
-
For testing publishing docker images from all branches is quite useful; So make the naming be specific rather then one general name
-
-
The new publish stage builds only the lava-server image and publishes it to the registry.
-
This tests our containerized lava-server based on the Debian packages by firing it up, setting up a dispatcher container and a lavacli container alongside it and then: - creating a super user - adding the qemu device type and a qtest device of that type - using lavacli to set the device dictionary and health test - running a regular job - making sure both the health check and the regular job succeeded
-
This is a very basic definition for just the LAVA service that can be used along with an override file to properly wire the dispatcher configuration directories and the database. A sample override to start the server with a containerized db is also included, along with a sample env file to specify all the required configuration through the environment.
-
The Dockerfile uses the debian packages as the basis, with a few files patched so that it can work without PostgreSQL on the same system, giving us more flexibility on how the db is set up.
-
Rémi Duraffort authored
Signed-off-by:Rémi Duraffort <remi.duraffort@linaro.org>
-