Metrics Cardinality Explorer

Metrics cardinality is the number of unique label combinations a metric has — each combination is a distinct time series. The Cardinality Explorer doesn’t run a query against that data; it lets you explore the shape of the label space itself, so you can find a high-cardinality label before you write a query against it, not after a query times out. See Cardinality for the underlying concept.

Access the Cardinality Explorer

Click Metrics in the top navigation, then Cardinality from the dropdown. The screen opens on three columns — Metrics, Labels, and Values — and a time range control (default: the last 5 minutes) independent of the Explorer’s.

Metrics, Labels, and Values: how the three panes cross-filter

The three columns narrow each other as you select:

Metrics

Every metric matching the current filters, with its type (counter, gauge, or histogram) and series count. Search by name, or click a metric’s …​ menu to Copy name, Open in Metrics Explorer, or Select it — selecting narrows the Labels column to only the labels that metric actually carries.

Labels

Every label present on the metrics currently in view, with how many distinct values each one has. Search by name; check Show series count in the column header to also show how many series carry each label — this triggers an additional, more expensive query, so leave it off unless you need it. Click a label to drill into its values.

Values

Populated once you’ve clicked into a label — every distinct value for that label, with a search box, an Operator dropdown (=, !=, =~ regex match, !~ regex exclude), and a checkbox per value. Each value also has its own …​ menu: Copy value, Open in Metrics Explorer, or Add to filter.

Drilling into a label shows a breadcrumb (← Back to labels list) and that label’s own row — its value count, series count, and series percentage of the current selection. Drilling into a metric shows the equivalent ← Back to metrics list breadcrumb.

Building a filter

Check one or more values in the Values column, or use a single value’s …​ menu → Add to filter for a quick one-value filter, then click Add filter (top right of the Values column) to apply the selection. The active filter appears as a chip in the Filters row above all three columns (for example name : any value), narrows the Metrics and Labels columns to match, and can be removed with the chip’s ×. Add filters on more than one label to combine them — each applied label narrows the result further.

Handing off to the Explorer

Open in Metrics Explorer (top right of the screen, or on any metric’s or value’s …​ menu) opens the Metrics Explorer with the current selection — metric, or metric plus label filter — already built as a live query, so you can move from "this is a high-cardinality label" straight to "here’s what its series actually look like" without re-entering anything.

Example: finding what’s driving a metric’s cardinality

  1. Open the Cardinality Explorer and check Show series count in the Labels column.

  2. Select a metric you suspect is expensive — its Labels column now shows only that metric’s own labels, each with a value count and series count.

  3. Sort by series count (or just scan) to find the label contributing the most series. A label whose values look like build hashes, timestamps, or other high-cardinality identifiers rather than a small fixed set is usually the culprit.

  4. Click that label to see its values, and Open in Metrics Explorer to confirm what the resulting series actually look like.

  5. If the label is genuinely unbounded and not useful to filter or group by, drop it at ingestion with a metric shaping rule rather than trying to filter it out of every query afterward.

See also