Stats
Real-time and historical resource dashboards with ring gauges and sparkline charts for every container.
Use this page as your resource dashboard. Redesigned in v4.5.0, it combines ring gauges with sparkline charts so you can spot pressure quickly. With historical stats, you can also look back at trends over the past 7 days.
Overview dashboard
The Stats page shows an overview of running containers first, then lets you drill into per-container panels when you need more detail.
Ring gauges
Ring gauges show instantaneous CPU and memory usage per container. The outer ring shows the current value, and the faded inner arc shows the average over the selected time range. Green stays below 50 percent, amber covers 50 to 80 percent, and red marks anything above that.
Sparkline charts
The four sparkline charts below each gauge row give you a compact history for each metric.
- CPU percent
- Memory in MB (live ranges) or memory percent (historical ranges)
- Network I/O in bytes per second for receive and transmit
- Disk I/O in bytes per second for read and write
Hover a point to see the exact value for that moment.
Time range selector
The selector at the top offers seven ranges split into two modes:
| Range | Mode | Data source |
|---|---|---|
| 1 min, 5 min, 15 min | Live | Real-time stream from the Docker daemon |
| 1 hour, 6 hours, 24 hours, 7 days | Historical | Time series store (5-second raw samples for 1h, 5-minute aggregates for longer ranges) |
Live ranges show data as it arrives from the Docker stats stream. Historical ranges query the time series database, which is populated by the background task scheduler.
Historical ranges require the stats_aggregation and stats_retention_prune background tasks to be enabled. They are on by default. Raw 5-second samples are retained for 1 hour; 5-minute aggregates are retained for 7 days.
Network interfaces and disk devices
Expand a container row when you want to break out network I/O by interface or disk I/O by device. This is useful when a container has multiple network attachments or several bind mounts.
Tips
- Spiky CPU over 1h that flattens out over 24h usually means short-lived batch work, not a leak.
- Memory that climbs over 24h and never drops is a strong leak signal.
- Disk I/O spikes that line up with log growth often mean unrotated logs, so check Logs.
- Use the 7-day range to identify weekly patterns, such as backup jobs that spike resources on a schedule.