Metrics

Coolify displays resource history collected by Sentinel. The global Metrics page summarizes usage across your team's servers, each server has detailed charts, and supported applications and standalone databases have their own CPU and memory charts.

The short version

Enable metrics on an eligible server to collect CPU, memory, disk, network, and load samples through Sentinel. Sampling and retention are configured per server, and charts show only the history still stored on that server.

Resource metrics have limitations

Dedicated resource charts are not available for Docker Compose applications or one-click service deployments. Use the fleet container table for available bulk metrics. The resource container must also be running before Coolify can display its metrics.

What metrics are available

Server CPUShows the percentage of total server CPU currently in use over the selected period.

Server memoryShows the percentage of total server memory currently in use over the selected period.

Application containersShows CPU percentage and memory usage in MB for a running, non-Compose application container.

Standalone databasesShows CPU percentage and memory usage in MB for a running standalone database container.

Before you enable metrics

Use a validated, non-Swarm server that is not configured as a build server.
Enable and sync Sentinel on that server.
Allow enough disk space on the connected server for the sampling rate and retention period you configure.

Enable metrics

Metrics are enabled at the server level and use the Sentinel container already assigned to that server.

Select the server

Open Servers and select the eligible connected server.

Open the Metrics page

Open Metrics and select Enable Metrics.

Wait for Sentinel to restart

Coolify saves the setting and restarts Sentinel with the metrics collector enabled. The collector then begins storing server samples and samples for supported running resources.

Confirm that metrics are available

Wait for the CPU Usage and Memory Usage charts to begin showing data.

If Sentinel is disabled, the Metrics page directs you to enable it first. Enabling metrics starts Sentinel when the server is eligible.

View fleet metrics

Open Metrics from the main navigation. Select All servers or one server, then choose 24 hours, 7 days, or 30 days. On the 24-hour range, Live Refresh can reload data every 60 seconds; the browser remembers your preference.

The fleet view includes:

Current server availability, CPU, memory, disk, network, and container totals.
CPU, memory, disk, network-throughput, and load-average trend charts.
A Servers table with sortable name and usage columns. Offline servers stay at the bottom.
Hottest containers, ranked by CPU, memory, disk, or network usage. Select a ranking metric and sort the available columns to investigate busy containers.

CPU is an approximate average of reporting servers' percentages rather than a CPU-core-weighted total. Memory and disk percentages use the combined used and total capacities; network throughput sums reporting servers. Offline servers do not contribute to usage totals. Missing or unavailable data should not be read as zero usage.

When no servers have metrics enabled, the page can offer Enable metrics for eligible servers you can update. They must have Sentinel enabled, be functional, and support metrics. Wait about a minute for collection after enabling.

The fleet container table uses Sentinel's bulk metrics. It can include containers that do not have a dedicated Coolify resource chart. Older Sentinel versions may lack these endpoints.

View server and resource metrics

Server and resource charts use different memory units.

Server metrics

Open Servers > [server] > Metrics.

CPU Usage is the percentage of total server CPU in use.
Memory Usage is the percentage of total server memory in use.
Disk usage shows root-filesystem usage, rather than a sum of mounted disks.
Network throughput shows received and transmitted traffic rates.
Load average shows system load. Load is not a CPU percentage; interpret it against the server's CPU capacity.

Disk, network, and load charts require a compatible Sentinel version. Missing optional metrics do not prevent CPU and memory charts from loading.

The individual server and resource Metrics pages offer these time ranges:

5 minutes and 10 minutes, with live refresh
30 minutes
1 hour
12 hours
1 week
30 days

The live ranges ask Coolify for new chart data every five seconds. The Sentinel sampling rate determines how often a new metric point is available.

Read pressure indicators

The fleet usage tables highlight CPU and memory at 70% and disk usage at 80%. Values of 90% or higher receive the highest pressure highlight. These are visual triage thresholds; review trends and workload requirements before changing infrastructure.

Sampling and retention

Open Servers > [server] > Metrics to change these values:

Collection rate controls the collection interval. The default is 10, and the minimum is 1.
History retention controls local retention. The default is 7, and the minimum is 1.
Push interval controls health-report frequency. The default is 60 seconds, and the minimum is 10.

Save the settings to apply them. Chart tooltips show both your local time and UTC.

Sentinel stores this history under /data/coolify/sentinel on the connected server. A lower collection interval creates more points and uses more disk space.

The chart can offer a longer range than the retained data. For example, a 30-day chart shows only the available history when History retention remains at its seven-day default.

Disable metrics

Open the server Metrics page and select Disable Metrics. Coolify restarts Sentinel with the collector disabled.

Disabling metrics does not disable Sentinel itself. Server heartbeats, container state, health information, and disk usage reporting can continue without CPU and memory history.