Always include annotations in the Deployment object

See merge request lava/lava-helm-charts!8

Checksum in annotations ensures that the pod restart when an object is
updated. Previosuly, checksums were not included when there were no other
annotations on the object, so they had no effect.
Instead, always include annotations with at least the checksums, so they
are be used regardless.

Fixes: 45bed51a

 ("Ensure the pod restarts when the K8s secret with the config is updated")

Signed-off-by: Pablo Vigo <pvigo@collabora.com>

Document LAVA server deployment using Helm chart

See merge request lava/lava-helm-charts!3

Ensure Pod Restarts upon Object Update

See merge request lava/lava-helm-charts!7

Pablo Vigo Mas authored 1 year ago and Pablo Vigo Mas committed 1 year ago

Annotate objects with a hashsum of a part of the object. This ensures
the pod restarts automatically whenever the secret is updated.

Signed-off-by: Pablo Vigo <pvigo@collabora.com>

Avoid failing when a job compression schedule is not set

See merge request lava/lava-helm-charts!6

Add a default `compression` value so that rendering does not fail with
an exception when not set by the caller:

    Error: Failed to render chart: exit status 1:
    Error: UPGRADE FAILED: template: lava-server/templates/cronjob.yaml:11:14:
    executing "lava-server/templates/cronjob.yaml"
    at <.Values.compress.schedule>:
    nil pointer evaluating interface {}.schedule

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

Convert job cleanup CronJob to a template and add TestJob output compression CronJob

See merge request lava/lava-helm-charts!5

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>

This setting can still be overridden by appending '--set
image.repository="selected/image"' to the Helm command.

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>

Allow specifying pod PriorityClass

See merge request lava/lava-helm-charts!4

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>

Improve various bit of the helm chart

See merge request lava/lava-helm-charts!1

The lava service expects to manage the underlying volume directly and
does migration on the database. So move to a statefulset rather then a
deployment to avoid having multiple lava containers running at the same
time with potentially conflicting database schema requirements

Signed-off-by: Sjoerd Simons <sjoerd@collabora.com>

To clean up test job outputs (not just the database) the cleanup job
needs access to the PV as well. So mount that in the normal location.

As the PVC is RWO make sure the cleanup job gets scheduled on the same
node as the lava pod otherwise it won't be able to attach to the volume

Signed-off-by: Sjoerd Simons <sjoerd@collabora.com>

The lava pods don't interact with the kubernetes API themselves so no
need for a service account

Signed-off-by: Sjoerd Simons <sjoerd@collabora.com>

Drop SWP file from initial import

See merge request lava/lava-helm-charts!2

Signed-off-by: Paweł Wieczorek <pawiecz@collabora.com>