New Relic Infrastructure Agent

The agent performs holistic host monitoring by gathering a wide array of system telemetry into a single data stream. This includes:

• System inventory: OS version, kernel, CPU architecture, network interfaces.
• Resource metrics: CPU, memory, disk I/O, and network utilization.
• Process-level details: Running services, resource consumption per process.
• Application metrics: From integrated technologies like NGINX, PostgreSQL, or Redis via plugins. This unified collection eliminates the need for multiple, disparate monitoring agents, simplifying the observability stack.

The agent's functionality is extended through a plugin architecture. Users can install integrations (pre-built plugins) for common services or develop custom plugins using the agent's SDK. Key aspects include:

• NRI-Flex: A powerful plugin that allows custom data collection by executing scripts (Bash, Python) or parsing log files, converting the output into New Relic metrics and events.
• On-Host Integrations: Direct monitoring for technologies like Docker, Kubernetes (via kube-state-metrics), AWS/Azure/GCP cloud services (using the cloud provider's metadata service).
• Dynamic Discovery: Some plugins can automatically discover and monitor new instances of a service as they come online.

To optimize network usage and backend costs, the agent performs local data reduction before transmission. Capabilities include:

• Metric aggregation: Rolling up high-frequency samples into lower-resolution aggregates.
• Event filtering: Dropping or sampling low-priority events based on configurable rules.
• Attribute manipulation: Adding, removing, or renaming tags (key-value pairs) on metrics and events for better organization and cost control.
• Local buffering: Data is cached on disk during network outages, ensuring at-least-once delivery when connectivity is restored.

The agent is engineered for production environments with a focus on security and reliability.

• Secure Communication: All data is transmitted to New Relic's platform over TLS-encrypted HTTPS connections.
• License Key Management: The agent authenticates using a unique license key, which can be managed via environment variables or configuration files.
• Resource Isolation: Runs with constrained system privileges and includes configurable resource limits (CPU, memory) to prevent the agent from impacting host performance.
• Health Self-Monitoring: The agent reports its own health and performance metrics, allowing SREs to monitor the monitor.

Agent behavior is controlled through a central, declarative YAML configuration file (newrelic-infra.yml). This enables infrastructure-as-code practices and automated deployment.

• Environment-Aware: Configuration can be templated and injected using environment variables for different deployment stages (dev, staging, prod).
• Orchestration-Friendly: Configuration management tools (Ansible, Chef, Puppet) or container orchestration platforms (Kubernetes ConfigMaps) can be used to manage and distribute agent configuration at scale.
• Dynamic Updates: Most configuration changes can be applied via a agent service restart without reinstalling the agent binary.

The agent supports multiple deployment models to fit diverse infrastructure.

• Bare Metal & Virtual Machines: Installed as a native system package (RPM, DEB, MSI).
• Containerized Environments: Run as a sidecar container alongside application containers or deployed as a DaemonSet in Kubernetes to ensure one agent per cluster node.
• Immutable Infrastructure: The agent can be baked into machine images (AMI, Docker image) for consistent, version-controlled deployments.
• Orchestrated Scaling: In Kubernetes, the DaemonSet automatically deploys the agent to new nodes as the cluster scales.