System Overview

System Overview shows how much memory, CPU and storage the platform and each of its services used over time, and how each service is doing now. Use it to see whether a period had resource problems, and which service caused them.

Read usage for a period

1. Open Observability > System Overview.
2. In the toolbar above the chart, pick a metric: RSS, RSS + Cache, CPU or Storage.
3. Set Period · to Last 1h, Last 24h, Last 7d or Last 30d, or choose Custom and pick a range.
4. Set Interval · to change how much time each point on the chart covers.
5. Read the chart. With All Services selected, its header reads Total System with the period, for example Total System · Last 7d, and it draws Max, Avg and Min lines by default. A triangle under a point marks a Detected spike. Click a legend item under the chart to hide that line, and click it again to show it. The legend also shows Total Memory with the platform's capacity.

The circular-arrow button at the right of the toolbar resets the period to Last 7d, clears any custom range and sets Interval back to its default. It is greyed out when nothing differs from those defaults.

Find the service behind a spike

1. On All Services, click a point on the chart to pin that moment. Clear selection appears in the chart card.
2. Read the cards above the chart. They list the top 10 services at that moment, highest first, with each service's value. With Storage, they show disk, compressed and efficiency.
3. Cards for AI services also show what those services did at that moment: counts of messages, tokens, conversations, artifacts, reports and presentations, and unique users or Minds. Hover a badge to see its name, such as Conversation count or Token usage.
4. A service whose value rose sharply at that moment carries a ring on its card.
5. To unpin, click the same point again, press Escape, or click Clear selection.

Check the state of services

• The list on the left shows one row per service. All Services is pinned at the top, and its dot pulses while it is selected.
• To find a service, type in Search.... The refresh button reloads the list and spins while it works. The list keeps its scroll position, and selecting a service scrolls it into view.
• Each row shows a sparkline of the service's recent values, its current value, its share of the total as a percent, and an uptime dot. A row with incidents in the period also shows their count, for example 3 inc, in amber.
• Rows for the platform's own services carry a small badge next to the name.
• Hover a row to see its State, Uptime, Ran for, Started, Finished, Restarts, Last exit code and OOM killed. Only the fields that apply to the service appear.
• With Storage, the same hover also shows the service's storage volume: Mount type, Status and References. The status reads Mounted · Active, Mounted · Inactive or Not mounted · Inactive.
• All Services shows a dot and a count such as 3/4 up. The dot is red when a service is down, and amber when a service is unhealthy, restarting or paused.

The state labels are Running, Running · healthy, Running · starting up, Running · unhealthy, Restarting, Paused, Created, never started, Killed out of memory, Dead, Exited with code and the number, One-shot task completed and Stopped. Red marks states that need attention now, such as Killed out of memory, Dead and a non-zero exit. Amber marks degraded states, such as Running · unhealthy. A finished one-shot task stays grey, because that is its normal end.

Choosing a service shows that service's own chart and incident list; both are described in Service Overview.

Permissions

To open this tab, your role needs the View System Overview permission. Without it, the tab is hidden. The System Overview Viewer preset grants it. The Observability Viewer and Observability Administrator presets include it too. All three sit in the Observability category. See the permission reference and roles.

Limits and edge cases

• A custom period can span up to one year, so Custom is the way to look further back than Last 30d. A longer range is not applied, and you see Date range too large with the message "Custom date range cannot exceed 1 year." Clearing the custom range returns you to the period you used before.
• While the chart draws, it shows Loading chart. If the period and interval have no samples, it shows No data available with "No system metrics found for the selected time range and interval." If it cannot load, it shows Failed to load data with the reason. See Observability troubleshooting.
• If the service cards cannot load, they show Failed to load service breakdown. Refresh the services list.
• When the window is too narrow for them, the service cards are hidden. Widen the window to see them again.
• Changing the period, interval, custom range or metric redraws the chart and clears a selected point and any hover.
• Changing the metric while you are zoomed into an incident takes you back to that service's incident list.
• Your metric, period, interval and selected service are remembered in this browser only. They are not saved to your account.

Questions, answered

What is the difference between RSS and RSS + Cache?

RSS shows the memory a service holds. RSS + Cache adds cache memory on top of it.

How do I find which service caused a memory spike?

On All Services, click the spike on the chart. The cards above the chart then list each service's value at that moment, highest first.

What do the small badges on a service card mean?

They count what the service did at the pinned moment: messages, tokens, conversations, artifacts, reports and presentations, and unique users or Minds. Hover a badge to see its name, such as Conversation count or Token usage.

A service shows "Killed out of memory". What happened?

The service ran out of memory and was stopped. Hover its row in the services list: OOM killed shows Yes, and Restarts appears if the service has restarted.

Why does the chart say "No data available"?

The platform recorded no samples for that period and interval. Widen the period or choose another interval.

Why does the interval change when I change the period?

Each period offers its own intervals. When you change the period, the page sets Interval · back to the default for the new period.

What does "3/4 up" mean on All Services?

Three of the four services that report a state are running and not unhealthy. A running service that is unhealthy does not count as up. Finished one-shot tasks are left out of the count.

Why are the service cards missing?

The window is too narrow for them, so they are hidden. Widen the window and they return.

Why did my selected point disappear?

Changing the period, interval, custom range or metric clears it. Click a point again to pin a new moment.

Is there a keyboard shortcut for the metric or period?

Yes. Alt+M switches to the next metric. Alt+T cycles the period through Last 1h, Last 24h, Last 7d and Last 30d and clears any custom range. Press Escape to clear a selected point.