Labeled kube-state-metrics for globally evaluated services
Labeled kube-state-metrics for globally evaluated services
This kube-state-metrics recording rules for globally evaluated
services. This will add :labeled
version of that that contain our
known tier
& type
labels.
For things like code-suggestions (running in the ai-assist
cluster)
this will be better than relying on the global type
label that is
also on the metrics, because it will allow us to separate monitoring
infrastructure (prometheus
) from the application for saturation
metrics. Currently the the metrics from prometheus pods running in
ai-assist
all have the type="code_suggestions"
label, in the
:labeled
variant this will be overwritten with type="monitoring"
.
This shows the series that will get a different type
label in the :labeled
version: https://thanos-query.ops.gitlab.net/graph?g0.expr=count%20by%20(label_type%2C%20type)(kube_pod_labels%7Btype%3D%22code_suggestions%22%7D)&g0.tab=1&g0.stacked=0&g0.range_input=1h&g0.max_source_resolution=0s&g0.deduplicate=1&g0.partial_response=0&g0.store_matches=%5B%5D
For https://gitlab.com/gitlab-com/gl-infra/reliability/-/issues/23724