Querying Metrics

Learn how to query and visualize time-series metrics in Grafana Explore

Querying Metrics in Explore

Explore provides powerful capabilities for querying and analyzing time-series metrics data from sources like Prometheus, Graphite, InfluxDB, and other metrics databases.

Metrics Visualization

When you query metrics data, Explore displays results in multiple formats:

Graph View

The primary visualization for time-series data shows:

• Line, bar, or points visualization styles
• Multiple series overlaid on the same graph
• Time-based X-axis with automatic scaling
• Interactive tooltips showing values at specific times
• Zoom capabilities by selecting a time range

Click and drag on the graph to zoom into a specific time range. The time picker updates automatically.

Table View

The table displays metric data in a structured format:

• Time-indexed rows
• Columns for each series and label
• Sorting and filtering capabilities
• Export to CSV option

Raw Data View

For Prometheus specifically, the Raw Prometheus view shows:

• Native Prometheus response format
• Instant query results
• Metadata about the query execution

Prometheus Queries

Basic Queries

Prometheus is the most common metrics data source in Grafana. Here are essential query patterns:

Instant Vector Queries

Query current values of a metric:

up

This returns the current value of the up metric for all time series.

Range Vector Queries

Query values over a time range:

rate(http_requests_total[5m])

Calculates the per-second rate of HTTP requests over the last 5 minutes.

Query Builder

The Prometheus query builder helps construct queries visually:

1. Select metric

   Start by choosing a metric from the dropdown. Metrics are autocompleted based on what’s available in your Prometheus instance.

2. Add label filters

   Refine your query by adding label matchers:

   http_requests_total{job="api-server", method="GET"}
   
3. Apply functions

   Add functions like rate(), sum(), or histogram_quantile() to transform the data:

   sum(rate(http_requests_total{job="api-server"}[5m])) by (status)
   
4. Add operations

   Combine multiple queries with mathematical operations or aggregations.

Common Query Patterns

Calculate Rate of Change

For counter metrics, use rate() to calculate per-second increase:

rate(node_cpu_seconds_total[5m])

Aggregation

Sum metrics across dimensions:

sum(rate(http_requests_total[5m])) by (handler)

This groups requests by the handler label and sums them.

Percentiles

Calculate percentiles from histogram metrics:

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

This calculates the 95th percentile request duration.

Filtering by Labels

Use label matchers to filter time series:

# Exact match
http_requests_total{status="200"}

# Regex match
http_requests_total{status=~"2.."}

# Not equal
http_requests_total{status!="200"}

# Regex not match
http_requests_total{status!~"5.."}

Advanced Techniques

Offset Queries

Compare current values with historical data:

rate(http_requests_total[5m]) 
  / 
rate(http_requests_total[5m] offset 1w)

This shows the current rate compared to the same time last week.

Multiple Aggregations

Combine different aggregation functions:

avg(rate(http_requests_total[5m])) by (job)
  / 
count(rate(http_requests_total[5m])) by (job)

Subqueries

Perform calculations over aggregated results:

max_over_time(
  sum(rate(http_requests_total[5m]))[1h:5m]
)

This finds the maximum request rate within a 1-hour window.

Graph Customization

Graph Styles

Explore supports multiple visualization styles:

• Lines - Default style showing connected data points
• Bars - Vertical bars representing each data point
• Points - Individual markers for each measurement

Change the style using the graph style selector in the Graph panel header.

Multiple Series Display

By default, Explore shows up to 20 time series. If your query returns more, you’ll see a warning with an option to show all series.

Manage series display:

• Use label filters to reduce the number of series
• Apply aggregations to combine related series
• Click “Show all series” if you need to see everything

Time Controls

Time Picker

Control the time range for your queries:

• Relative time ranges: Last 5m, 15m, 1h, 6h, 24h, etc.
• Absolute time ranges: Specify exact start and end times
• Fiscal year settings: Configure fiscal year start month

Zoom and Navigation

1. Click and drag on the graph to zoom into a time range
2. Use the ← → buttons to shift the time range backward/forward
3. Click Zoom out to return to the previous time range
4. Use Time sync in split view to synchronize time ranges across panes

Auto-refresh

Enable auto-refresh to continuously update metrics:

1. Click the refresh interval dropdown
2. Select an interval (5s, 10s, 30s, 1m, etc.)
3. Queries re-run automatically at the selected interval

Be mindful of query performance when using short auto-refresh intervals, especially with complex queries or large time ranges.

Query Inspector

The Query Inspector provides detailed information about query execution:

1. Open Inspector

   Click the Query inspector button below the query editor.

2. View Stats

   See query execution time, number of series returned, and data transfer size.

3. Review Request

   Examine the exact API request sent to the data source.

4. Inspect Response

   View the raw response data from the data source.

Logs Volume (Metrics from Logs)

When querying log sources that support metrics (like Loki), Explore can show a metrics overlay:

• Histogram showing log volume over time
• Colored by log level (error, warning, info, etc.)
• Click bars to filter to that time range
• Toggle on/off with the Logs volume button

This helps identify when issues occurred before diving into individual log lines.

Exemplars

If your Prometheus instance has exemplars enabled, you can:

1. See exemplar markers on the graph (small diamonds)
2. Click an exemplar to view associated trace
3. Follow the link to the trace in the tracing data source

Exemplars connect metrics to traces, enabling seamless navigation from “what” to “why.”

Exemplars require Prometheus 2.26+ and must be explicitly recorded by your instrumentation.

Exporting and Sharing

Share Query

1. Click the Share button in the toolbar
2. Copy the generated URL
3. Share with team members - the URL contains all query and time range information

Export Data

From the table view:

1. Click the download icon
2. Choose CSV format
3. Data exports with all columns and time formatting

Create Dashboard Panel

Convert your Explore query into a dashboard panel:

1. Click Add to dashboard (when available)
2. Select target dashboard
3. Configure panel options
4. Save the dashboard

Best Practices

Optimize query performance

• Use specific label filters to reduce series cardinality
• Avoid queries that select thousands of time series
• Use appropriate time ranges - longer ranges require more processing
• Apply rate() before aggregation functions like sum()

Use the query builder for exploration

• Start with the visual query builder when exploring unfamiliar metrics
• Switch to code mode for advanced PromQL features
• Use metrics browser to discover available metrics and labels

Leverage split view for correlation

• Compare metrics before and after a deployment
• View related metrics side-by-side (CPU vs. memory, requests vs. errors)
• Use time sync to ensure consistent time ranges

Save frequently used queries

• Use query history to access recent queries
• Star important queries for quick access
• Consider creating dashboards for regularly-reviewed metrics

Troubleshooting

No data returned

• Verify the time range includes data for your metric
• Check label filters aren’t too restrictive
• Confirm the metric exists in your Prometheus instance
• Use the metrics browser to explore available metrics

Query timeout

• Reduce the time range
• Add more specific label filters
• Simplify complex aggregations
• Check Prometheus performance and resource limits

Too many series

• Add label filters to reduce cardinality
• Use aggregation to combine related series
• Consider if you need all the data or just a subset

Next Steps

Querying Logs

Learn how to explore and analyze log data

Distributed Tracing

Investigate traces and service dependencies