Skip to main content
A Prometheus metric source gives Arklow a view of the series you already collect. You can query a Prometheus-compatible HTTP API or scrape a text exposition endpoint directly.

​
Collection modes

Query Range is the default. In Exposition Scrape mode, the endpoint supplies metric families directly and Arklow skips PromQL evaluation.

​
Credentials

Credentials are optional. For an endpoint that expects authentication, create an HTTP Auth Headers credential containing the headers that must accompany each request, then select it on the metric source.

​
Add Prometheus to Arklow

1

Open the metrics page

Go to Metrics and click New metric source.
2

Pick the type

Set Source Type to Prometheus and give the source a name.
3

Enter the endpoint

Enter the base URL for Query Range, or the service URL for Exposition Scrape.
4

Choose the mode

Select Query Range or Exposition Scrape. For a scrape, set Metrics Path if the endpoint does not use /metrics.
5

Select authentication

Select an HTTP Auth Headers credential if the endpoint requires one. Leave authentication empty for an endpoint that accepts unauthenticated requests.
6

Save

Click Create metric source.

​
Add queries

Open the metric source’s Queries tab and click Add Query. For Query Range, enter a PromQL expression that returns the series relevant to capacity or service health:
Keep the labels you will use to distinguish lanes, such as customer_id, service, or region. Arklow records each distinct label set as a metric series. For Exposition Scrape, enter a non-empty expression such as scrape to activate collection. Arklow treats the value as an activation field and reads all compatible metric families from the configured path. Filtering and aggregation belong at the exporter.

​
Identity tag exclusions

Labels normally participate in series identity. Add volatile labels such as pod or instance to Identity Tag Excludes when a rollout should not create a new logical series for otherwise equivalent telemetry. Exclusions apply only to series identity; the provider request remains unchanged. Keep every tenant or partition label needed to match action tags. A query begins informing an action after you attach it with a metric policy.
1

Open the action

Open the action definition and select its Metrics tab.
2

Attach metrics

Click Attach Metrics, or Add Policy when the action already has a policy. Select the Prometheus source and query.
3

Choose a mode

Use Discover to consider the query’s returned series. Use Manual to limit consideration to series matching the selector you provide.
4

Align tag names

In Discover mode, add Tag Bindings when an action tag and Prometheus label carry the same identity under different names, such as customer and customer_id. Manual policies use the metric-side selector instead.
5

Create the policy

Set Max Series to a deliberate bound, then click Create Policy.
Manual selectors map label names to accepted values: