All Products
Search

This Product

    Managed Service for Prometheus:Explore metrics at the environment level

    Document Center

    Managed Service for Prometheus:Explore metrics at the environment level

    Last Updated:Jun 25, 2026

    You can use PromQL queries to verify successful data ingestion into a target environment.

    Prerequisites

    You have activated Application Real-Time Monitoring Service. For more information, see Activate ARMS.

    Procedure

    1. Log on to the Cloud Monitor console.

    2. In the left-side navigation pane, click Integration Management.

    3. On the Integrated Environments tab, click the name of your target environment to open its details page.

    4. On the environment details page, click the Metrics Explorer tab.

    5. In the Explore section, set the metric, labels, and options, and then click Run query. This uses PromQL syntax to query data from the default Prometheus storage instance for the environment to verify successful data ingestion.

      image

    Common PromQL query examples

    On the Metrics Explorer page, you can run PromQL queries to explore your metric data. The following section provides examples for common use cases.

    Query Pod-level P95/P99 latency

    When you use the histogram_quantile function to query P95 or P99 latency metrics for Pods in an ACK cluster, you must include both the le label and any required dimensional labels (such as namespace and pod) in the by clause of the sum(rate()) aggregation. If these labels are omitted, the query results will not include Pod-level data.

    To query the P95 latency for each Pod:

    histogram_quantile(0.95, sum(rate(http_request_duration_seconds_bucket[5m])) by (le, namespace, pod))

    To query the P99 latency for each Pod:

    histogram_quantile(0.99, sum(rate(http_request_duration_seconds_bucket[5m])) by (le, namespace, pod))

    If the le label is missing from the by clause, the histogram_quantile function cannot correctly calculate the quantile. If dimensional labels such as namespace or pod are missing, the query returns only node-level aggregated data and cannot be broken down by Pod.

    If a search for http_request_duration_seconds_bucket in Metrics Explorer returns data, but adding a filter such as namespace yields no results, check the following:

    • Verify that the Prometheus instance is correctly associated with the target ACK cluster.

    • Verify that the label values (for namespace and pod) match the actual values in the cluster.

    For a list of basic metrics supported for container clusters, see Basic metrics for container clusters.