Skip to main content

​
Installation

Install the Visent Telemetry agent on your GPU nodes to begin monitoring performance metrics, resource utilization, and system health. The agent supports Linux, Windows, and containerized environments with automatic GPU discovery and minimal performance overhead. Supports NVIDIA, AMD, and Intel GPUs with both bare metal and cloud deployments.

​
Prerequisites

  • Linux system with GPU drivers installed (NVIDIA, AMD, or Intel)
  • Root access for GPU device access
  • For Kubernetes: Version 1.19+ with GPU nodes

​
Choose Your Installation Method

# Install Visent Agent
curl -fsSL https://install.visent.io | sudo bash

# Start monitoring
sudo VISENT_ENROLL_TOKEN=<your_token> visent-agent run

​
2. Docker

docker run -d \
  --name visent-agent \
  --privileged \
  --pid=host \
  -v /sys:/sys:ro \
  -v /dev:/dev \
  -v /proc:/proc:ro \
  -e VISENT_ENROLL_TOKEN=<your_token> \
  ghcr.io/raiyorahq/visent-agent:latest

​
3. Kubernetes (Helm)

# Add repository
helm repo add visent https://charts.visent.io
helm repo update

# Install
helm install visent-agent visent/visent-agent \
  --namespace visent-system \
  --create-namespace \
  --set cloud.enrollToken=<your_token>

​
4. Manual Download

# Download and install
wget https://visent.io/releases/latest/download/visent-agent-linux-amd64.tar.gz
tar xzf visent-agent-linux-amd64.tar.gz
sudo mv visent-agent-linux-amd64 /usr/local/bin/visent-agent
sudo chmod +x /usr/local/bin/visent-agent

# Start monitoring
sudo VISENT_ENROLL_TOKEN=<your_token> visent-agent run

​
Get Your Enrollment Token

  1. Log in to Visent Cloud
  2. Go to Settings > API Keys
  3. Create a new enrollment token
  4. Copy the token for use in the commands above

​
Verify Installation

Once running, your GPU metrics will appear in the Visent Dashboard within 30 seconds. Check agent status: 
# View logs
sudo journalctl -u visent-agent -f

# For Docker
docker logs visent-agent -f

# For Kubernetes
kubectl logs -n visent-system -l app.kubernetes.io/name=visent-agent -f

​
Configuration Options

​
Export Modes

  • Cloud: Push metrics to Visent Cloud (default)
  • Prometheus: Expose metrics locally on port 9101
  • Both: Enable both modes simultaneously
# Prometheus mode
sudo VISENT_EXPORTER=prometheus visent-agent run

# Both modes
sudo VISENT_EXPORTER=both VISENT_ENROLL_TOKEN=<token> visent-agent run

​
Environment Variables

VariableDefaultDescription
VISENT_ENROLL_TOKEN-Authentication token
VISENT_EXPORTERcloudExport mode
VISENT_INTERVAL10Collection interval (seconds)
VISENT_DEBUGfalseDebug logging

​
Troubleshooting

​
No metrics appearing?

  1. Check GPU drivers: nvidia-smi or rocm-smi
  2. View agent logs for errors
  3. Test with mock collector: COLLECTOR=mock VISENT_DEBUG=1 visent-agent run

​
Permission errors?

  • Ensure agent runs as root
  • For Docker/K8s: verify --privileged flag

​
Network issues?

  • Test connectivity: curl -v https://api.visent.io/health
  • Check firewall rules for HTTPS (port 443)

​
Next Steps