Metrics Explorer

The Metrics Explorer is where you browse and query metrics data. Click Metrics in the sidebar to open it at /metrics. The mode switch at the top has three modes:

• Explore: the browse grid, a visual overview of every metric family in your organization.
• Visualize: the PromQL query editor with chart visualization and dashboard controls. Old /metrics/editor links redirect to this mode.
• Favorites: the metrics you starred in Explore.

Browse grid

The browse grid gives you a fast, visual overview of all metric families. Each card shows the metric name, its type (via a color-coded badge), the current charted value, and a mini sparkline. Use Grid or Rows to change the layout.

Metric cards

Cards are fetched lazily as you scroll. A card queries its default aggregation only when it enters the viewport, so opening the grid on a large org never fires hundreds of queries at once.

Card defaults by type:

You can override a card's function for individual metrics via the gear (⚙) icon on the card, which opens Configure function. Pick a variant, such as Average, Rate (sum) or Percentiles, and click Apply, or click Reset to default. The override is saved per-browser and persists across sessions.

Star a metric (the ★ on a card) to add it to your favorites. The Favorites mode shows only starred metrics, and Convert to dashboard adds them all to a dashboard, one panel each.

Click Drill down on a card to open the metric's detail view. It shows a larger chart with Compare to (1 hour, 1 day or 1 week ago), Forecast (Linear or Smoothed trend) and, for histograms, Exemplars. Below the chart are three tabs:

• Breakdown: one chart per label. Drill down into a label to see its values, then Add to filter or Exclude a value.
• Related: other metrics that share labels with this one.
• Used in: the dashboards, alerts, SLOs and scheduled pipelines that query this metric.

Use Open in Visualize to continue in the query editor. In Enterprise Edition, Logs & traces opens the logs and traces of the metric's service. Back to metrics returns to the grid.

Filtering

The left rail exposes three filter panels:

• Prefix — group by the first segment of the metric name. Useful in orgs that follow naming conventions like node_cpu_*, node_memory_*, http_request_*.
• Suffix — group by the last underscore-delimited segment. Surfaces common units (_bytes, _seconds, _total) alongside domain suffixes.
• Type — filter by metric type: Counter, Gauge, Histogram, Summary, or Other.

The Search metrics box filters cards by name or help text. The sort control orders cards A–Z, Z–A, or by usage in Dashboards or Alerts (most used first). With data (selected by default) hides cards whose queries returned no samples, keeping the grid clean of registered-but-unused metrics; select All to show every metric. A count below the grid shows how many cards are hidden.

Use Views to save the current search, filters, sort, layout and time range as a named view, and to reopen, update or delete views. Saved views are shared with everyone in the organization.

Label filter chips

Click Filter to add a label matcher: select a label, an operator (= exact value, != any other value, =~ regex), and a value. Filters are combined with AND. Cards are then narrowed to metrics whose streams carry every selected label, so a status=~"5.." filter limits the grid to metrics that actually have a status label.

Query editor

The Visualize mode is where you build queries, visualize results, and configure dashboard panels. It opens with the same layout as a dashboard panel editor.

Stream selector

The stream dropdown shows metric names with type badges — a single letter (C, G, H, S, O) on a color-coded background:

The badge colors match the same palette used in the Metrics Explorer browse grid, so a type looks the same wherever you see it.

Builder mode

The builder lets you construct PromQL queries step by step without writing raw PromQL. Select Builder from the mode toggle.

When you pick a metric, the builder auto-seeds a default query based on its type: sum(rate(counter[4m])) for a counter, avg(gauge) for a gauge, histogram_quantile(...) for a histogram. You can then add or remove operations (rate, sum, avg, topk, etc.) and label filters.

Operations are organized into groups: Rate & range, Aggregation, Math, and more. Each operation is rendered as a draggable chip. Drag to reorder; the query string updates live.

The builder writes the full PromQL into the query input, so you can switch to Code mode at any time to see or edit the raw query.

Code mode

Select Code to write PromQL directly in the editor. The editor provides syntax highlighting, autocompletion for metric names and label keys, and inline error indicators.

Press Ctrl+Enter (or Cmd+Enter on macOS) to run the query from the editor without clicking Apply. The editor flushes any pending input before running, so the query that executes is always what is on screen.

Date and time range

The time picker at the top of the editor supports relative ranges (Past 15 Minutes, Past 6 Hours, etc.) and absolute custom ranges. For dashboards (where auto-apply is off), the picker's trigger label shows the range that is actually in force, not your pending selection — so switching the picker to "Past 6 Hours" without clicking Apply leaves the trigger reading the previous range.

Click Run query to run the query against the selected range. The refresh interval dropdown schedules automatic re-runs.

Visualizations

The editor supports the same chart types available for dashboard panels: line, bar, area, scatter, table, heatmap, and more. Switch the chart type from the dropdown above the chart.

Add to Dashboard

After you have a query and chart you want to keep, click Add to Dashboard. Choose an existing dashboard and tab, or create a new one. The panel is inserted with the current query, chart type, time range, and PromQL builder state preserved.

Sharing a metrics view

The Share button in the editor toolbar copies a URL that encodes your current panel configuration — chart type, query, stream, time range, and refresh interval. Anyone opening the link sees the same chart and data. The share URL carries the full state as a base64 blob in the metrics_data query parameter.

The editor also accepts lightweight deep-link parameters (chart_type, query_type, stream_name, query) for programmatic linking from alerts, dashboards, and other pages.

Choosing a stream type auto-selects PromQL

When you create a new panel and set the stream type to metrics, the query mode automatically switches to PromQL. This saves you one click on exactly the new panels that need it. A panel that already has a written SQL query is never converted — so opening an existing SQL panel for editing keeps its query type intact, and SQL over metrics remains one click away.

Next steps

• Dashboards: add metrics panels to dashboards
• Downsampling: reduce storage for long-term metrics
• Prometheus ingestion: send metrics via remote-write
• Alerts: create alerts on metric thresholds