Edge Agent Collectors

Collectors are the data-gathering components of the Edge Agent. Each collector is responsible for a specific type of telemetry data. All collectors run independently and can be individually enabled, disabled, or configured.

Built-in Collectors

System Collector

Collects host-level system metrics.

Metrics collected:

Default interval: 10 seconds

Process Collector

Collects per-process resource usage for the top N processes by CPU and memory consumption.

Metrics collected:

Default interval: 30 seconds

Disk Collector

Collects disk usage and I/O performance metrics.

Metrics collected:

Default interval: 30 seconds

Network Collector

Collects network interface traffic and connection metrics.

Metrics collected:

Default interval: 10 seconds

Log Collector

Tails log files and forwards them to the Tenant Plane with structured field extraction.

Features:

• Glob-pattern file discovery (e.g., /var/log/*.log)
• Automatic multiline detection for stack traces and structured logs
• Built-in parsers for syslog, JSON, Apache/Nginx access logs, and common application frameworks
• Custom parser support via Grok patterns
• File rotation handling (detects renamed/truncated files)
• Bookmark persistence (resumes from last read position after agent restart)

Custom Collectors

You can extend the agent with custom collectors by placing scripts in /etc/atlasai/collectors.d/:

#!/bin/bash
# /etc/atlasai/collectors.d/my-app-health.sh
# Runs every collection interval and outputs metrics in a simple format
 
response=$(curl -s http://localhost:3000/health)
status=$(echo "$response" | jq -r '.status')
 
if [ "$status" = "healthy" ]; then
  echo "my_app.health 1"
else
  echo "my_app.health 0"
fi

Custom collector scripts must:

• Be executable (chmod +x)
• Output one metric per line in the format: metric_name value
• Complete within 10 seconds (configurable timeout)