Skip to main content

Infra

The Infra tab shows a live view of host or Kubernetes node resource usage next to your APM workflows, so you can tell an application regression apart from a saturated machine.

The tab appears only when enabled for a deployment with the required OpenTelemetry metrics.

Infra tab with CPU, memory, network, and filesystem charts above the instances table

Hosts or Kubernetes Nodes​

An administrator configures the page in one of two modes:

  • Hosts identifies machines by host name.
  • Kubernetes identifies cluster nodes by node name.

Use the instance selector in the topbar to show one or several discovered hosts or nodes. Environment and configured telemetry-service restrictions also apply.

Fixed Live Window​

Infra uses a fixed recent window rather than the application's general time picker. Depending on plugin settings, the window is 5, 10, or 15 minutes. The page clearly reports when no instances have sent data during that interval.

Charts​

ChartWhat it shows
CPU UsageCPU utilization percentage per host or node
Memory UsageMemory utilization percentage
Network Usage (receive)Receive throughput in bytes per second
Filesystem UsageFilesystem utilization percentage

Multiple instances are plotted as separate series. The charts do not share a crosshair because each resource uses a different unit or scale.

CPU Usage, Memory Usage, Network Usage, and Filesystem Usage charts plotted per instance

Instances Table​

The table shows the latest recorded CPU, memory, network receive, and filesystem value for every visible host or node. Select any metric heading to sort; missing measurements stay below recorded values in either direction.

Create Infrastructure Alerts​

Each chart menu can open a Grafana alert-rule draft for the selected host or node scope. The query, environment, data source, metric, and infra mode are pre-filled. Review the threshold before saving. Availability depends on your Grafana permissions and alerting configuration.

Use Cases​

Decide Whether to Blame the Machine​

  1. Confirm the application regression in Services.
  2. Open Infra and select the instances you know run that service. Infra lists hosts or nodes, not services, so the mapping is yours to make.
  3. If CPU, memory, and filesystem are all unremarkable, the cause is in the application path.

Catch a Saturating Node​

  1. Sort the instances table by CPU, then by memory.
  2. Watch whether one instance sits above the rest across the live window.
  3. Open the chart menu to draft an alert for that scope.

Check a Host After a Deploy​

  1. Select the instance that took the new build.
  2. Watch memory across the live window for a step change that does not settle.
  3. Cross-check the service's error rate in Services.

FAQ​

Why doesn't Infra follow the time range I set?​

Infra uses a fixed live window instead of the application time picker, either 5, 10, or 15 minutes depending on plugin settings. It reports clearly when no instance has sent data during that interval. Every other APM tab follows the time picker.

Can I see which host or node runs a given service?​

Not from this tab. Infra lists hosts or nodes, selected with the instance selector, and does not map services onto them. Pick the instances you know run the service, or use k8X when you need workloads resolved to nodes.

Why don't the charts share a crosshair?​

Each chart uses a different unit or scale, so a shared crosshair would imply a comparison the values do not support. Compare instances within a chart rather than across them.

Last updated on
Was this page helpful?