The card was added in the mattermost messages mostly for debugging (to
do a full state). However with more then 100 devices in lava it ends
up bigger then mattermost will accept so drop it.

The cleanup function should only cleanup (or well mark) jobs that have
been read by the metrics. That's impossible to do with the cleanup
function in the JobCache itself (at least not without introducing
interesting locking rules.

So instead move it to the JobCacheReader which will guarantee that a
specific view gets cleaned up instead (as the reader lock is being
held). This doesn't do the actual cleanup anymore though; just schedules
the job as expired for the next update run to actually drop it later.

By dropping the device reader between each iteration we open a window
for races where the underlying data might updated inbetween. Fix by
simply holding the same reader while generating all metrics.

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

Retrieving jobs is much slower than retrieving devices from Lava,
which has the effect of causing device status updates to be delayed,
and of limiting the quantity of queries that can be made in any
reasonable amount of time.

Cache maintainance code originally from Ed Smith
<ed.smith@collabora.com>, but heavily refactored so simplification.

Prepare for seperating out the job cache into it's own object; Leaving
the current cache to only cache devices, so match the name (even though
at the moment that's still a lie)

This was previously a degenerate HashMap that only contained the
value 'true' for the keys that existed.

Trivial simplifications of existing code.

Gitlab looks for app.gitlab.com/{app,env} annotations in various objects
including deployments; Add support for that to the chart

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

This is only used to inline image snapshots into alerts, which we don't
need at all. It's also smashing titanium's CPUs to bits.

Reports the current running time in seconds the jobs have been processing.

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>

This gives some basic instruction and information about the project.

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>

A new report has been added:

lava_queue_size{device_type=DEVICE_TYPE,submitter=SUBMITTER,priority=PRIORITY}

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>

Major update of tokio.

Update warp, prometheus and env_logger to their latest available version.

Signed-off-by: Corentin Noël <corentin.noel@collabora.com>

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

The lava frontend talks about devices and workers; So follow that
jargon. A specific worker in all metrics should be referred to by the
worker label, while a device always by a device label. That way we can
merge timeseries without requiring label renaming

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

When either lava or mattermost throw errors warn and retry later rather
then throwing up an error.

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

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

Make the polling for the mattermost monitoring have a configurable
interval and accept data that's no older then a quarter of the interval.

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