NVIDIA Jetson

Plugin: go.d.plugin Module: jetson

Overview

Monitor GPU and memory controller utilization and clock frequencies on NVIDIA Jetson devices.

You get:

• GPU utilization.
• GPU clock frequency, as one value or, on devices that report it, one value per graphics processing cluster (GPC).
• External memory controller (EMC) bandwidth utilization and clock frequency. The CPU and GPU share this memory, so high EMC utilization can reveal memory-bandwidth-bound workloads.

CPU, memory, and swap usage come from Netdata's standard Linux system monitoring; this collector adds the Jetson GPU and EMC readings those charts do not cover.

The collector runs NVIDIA's tegrastats --interval 1000 on the local host and reads the line of readings it prints every second. One tegrastats process runs for as long as the job runs, under the Netdata service account (usually netdata); the collector never elevates its privileges.

It stops only the tegrastats process it started; other instances on the host, such as one you run in a terminal, are not affected.

This collector is only supported on the following platforms:

• Linux

This collector only supports collecting metrics from a single instance of this integration.

The Netdata service account must be able to run tegrastats. Because the collector never elevates privileges, a reading that your Jetson Linux release shows only to root is not collected.

Default Behavior

Auto-Detection

The stock configuration runs one job named localhost, which looks for tegrastats in Netdata's PATH and starts collecting when it is found. If tegrastats is missing when Netdata starts, the job does not start; restart Netdata after making it available.

Limits

• The readings depend on the Jetson module, the Jetson Linux release, and permissions. For example, tegrastats on Jetson Thor can report GPU clocks without GPU utilization. A reading that tegrastats does not report leaves a gap; it is never charted as zero.
• If tegrastats stops printing readings, the charts show a gap after three seconds instead of repeating the last values.

Performance Impact

Each job keeps one tegrastats --interval 1000 process running, the same load as running that command yourself. Netdata parses one line of its output per second. A longer update_every reduces how often Netdata stores readings, not how often tegrastats samples.

Setup

You can configure the jetson collector in two ways:

Method	Best for	How to
UI	Fast setup without editing files	Go to Nodes → Configure this node → Collectors → Jobs, search for jetson, then click + to add a job.
File	If you prefer configuring via file, or need to automate deployments (e.g., with Ansible)	Edit go.d/jetson.conf and add a job.

important

UI configuration requires paid Netdata Cloud plan.

Prerequisites

Make tegrastats available to Netdata

The collector needs Netdata installed natively on the Jetson device and the tegrastats utility that NVIDIA Jetson Linux (installed with JetPack) provides. Netdata running in a container is not supported.

Check that the Netdata service account can run it:

sudo -u netdata tegrastats --interval 1000

Each line should contain a GR3D_FREQ (GPU) or EMC_FREQ (memory controller) reading; stop the command with Ctrl+C. Readings missing from this output are also missing from Netdata.

If the command is not found, restore tegrastats with NVIDIA's tegrastats deployment instructions. Netdata searches its own PATH, which adds /sbin, /usr/sbin, /usr/local/bin, and /usr/local/sbin to the PATH it starts with. If tegrastats is in another directory, set PATH in the [environment variables] section of netdata.conf to a list that includes it, then restart Netdata.

Configuration

Options

The option can be set globally or per job. One job covers the whole Jetson device.

Config options

Option	Description	Default	Required
update_every	Data collection interval, in seconds. tegrastats samples once per second regardless; a longer interval charts the latest reading, not an average.	1	no

via UI

Configure the jetson collector from the Netdata web interface:

1. Go to Nodes.
2. Select the node where you want the jetson data-collection job to run and click the ⚙ (Configure this node). That node will run the data collection.
3. The Collectors → Jobs view opens by default.
4. In the Search box, type jetson (or scroll the list) to locate the jetson collector.
5. Click the + next to the jetson collector to add a new job.
6. Fill in the job fields, then click Test to verify the configuration and Submit to save.
   • Test validates the provided settings and checks the collector's startup prerequisites. Successful validation does not guarantee that every metric will be available during collection.
   • If it fails, an error message appears with details (for example, connection refused, timeout, or command execution errors), so you can adjust and retest.

via File

The configuration file name for this integration is go.d/jetson.conf.

The file format is YAML. Generally, the structure is:

update_every: 1
autodetection_retry: 0
jobs:
  - name: some_name1
  - name: some_name2

You can edit the configuration file using the edit-config script from the Netdata config directory.

cd /etc/netdata 2>/dev/null || cd /opt/netdata/etc/netdata
sudo ./edit-config go.d/jetson.conf

Examples

Basic

The stock configuration, with one job for the local Jetson device.

Config

jobs:
  - name: localhost

Alerts

There are no alerts configured by default for this integration.

Metrics

Metrics grouped by scope.

The scope defines the instance that the metric belongs to. An instance is uniquely identified by a set of labels.

Charts show only readings present in the latest tegrastats line. A missing or invalid reading leaves a gap; a reported zero is charted as zero.

Per NVIDIA Jetson instance

The Jetson device's integrated GPU and external memory controller (EMC).

This scope has no labels.

Metrics:

Metric	Description	Dimensions	Unit
jetson.gpu_utilization	GPU Utilization	utilization	percentage
jetson.gpu_frequency	GPU Frequency	frequency	MHz
jetson.emc_utilization	Memory Controller Bandwidth Utilization at Current Frequency	utilization	percentage
jetson.emc_frequency	Memory Controller Frequency	frequency	MHz

Per GPU graphics processing cluster

One graphics processing cluster (GPC) of the GPU, present when tegrastats reports a clock for each GPC.

Labels:

Label	Description
gpc	Position of the cluster in the tegrastats GPU clock list, starting at 0.

Metrics:

Metric	Description	Dimensions	Unit
jetson.gpu_gpc_frequency	GPU Graphics Processing Cluster Frequency	frequency	MHz

Troubleshooting

Diagnostics

Debug Mode

Important: Debug mode is not supported for data collection jobs created via the UI using the Dyncfg feature.

To troubleshoot issues with the jetson collector, run the go.d.plugin with the debug option enabled. The output should give you clues as to why the collector isn't working.

• Navigate to the plugins.d directory, usually at /usr/libexec/netdata/plugins.d/. If that's not the case on your system, open netdata.conf and look for the plugins setting under [directories].

  cd /usr/libexec/netdata/plugins.d/

• Switch to the netdata user.

  sudo -u netdata -s

• Run the go.d.plugin to debug the collector:

  ./go.d.plugin -d -m jetson

  To debug a specific job:

  ./go.d.plugin -d -m jetson -j jobName

Getting Logs

If you're encountering problems with the jetson collector, follow these steps to retrieve logs and identify potential issues:

• Run the command specific to your system (systemd, non-systemd, or Docker container).
• Examine the output for any warnings or error messages that might indicate issues. These messages should provide clues about the root cause of the problem.

System with systemd

Use the following command to view logs generated since the last Netdata service restart:

journalctl _SYSTEMD_INVOCATION_ID="$(systemctl show --value --property=InvocationID netdata)" --namespace=netdata --grep jetson

System without systemd

Locate the collector log file, typically at /var/log/netdata/collector.log, and use grep to filter for collector's name:

grep jetson /var/log/netdata/collector.log

Note: This method shows logs from all restarts. Focus on the latest entries for troubleshooting current issues.

Docker Container

If your Netdata runs in a Docker container named "netdata" (replace if different), use this command:

docker logs netdata 2>&1 | grep jetson

Known Errors

tegrastats executable not found in PATH

Cause

No executable named tegrastats is in Netdata's PATH. This is expected on Linux hosts that are not Jetson devices, and when Netdata runs in a container.

Fix

On a Jetson device, make tegrastats available as described in Prerequisites, then restart Netdata. On other hosts, ignore the message or set jetson: no in the modules section of go.d.conf.

start tegrastats: error

Cause

Netdata could not start tegrastats through its unprivileged command helper, nd-run.

Fix

Read the error text. If nd-run is missing or cannot be executed, repair or reinstall Netdata.

tegrastats source unavailable: error

Cause

The running tegrastats exited or stopped printing readings. Netdata keeps restarting it, at most 30 seconds apart.

Fix

Run the check in Prerequisites and compare its output with the error text. tegrastats exited: exit status N means the utility failed; tegrastats stopped producing records means it kept running without printing complete lines.

no fresh tegrastats sample

Cause

Netdata has not received a complete tegrastats line in the last three seconds, usually because tegrastats is starting, has stopped, or is being restarted.

Fix

If the error persists, look for tegrastats source unavailable messages in the collector log (see Diagnostics) and run the check in Prerequisites.

tegrastats sample contains no supported GPU or EMC readings

Cause

The latest tegrastats line has neither a valid GR3D_FREQ (GPU) nor EMC_FREQ (memory controller) reading. Which readings tegrastats prints depends on the Jetson module, the Jetson Linux release, and the account that runs it.

Fix

Run the check in Prerequisites. If its output has no GR3D_FREQ or EMC_FREQ readings, this collector has nothing to chart on this device.