Skip to content

Latest commit

 

History

History
95 lines (62 loc) · 2.75 KB

performance.md

File metadata and controls

95 lines (62 loc) · 2.75 KB

DNS-collector - Performance tuning

Overview

DNS-collector can handle high-volume DNS traffic with proper tuning. This guide helps you optimize performance for large-scale deployments and understand the performance implications of different configuration choices.

Performance Monitoring

Built-in Metrics

DNS-collector provides comprehensive performance metrics through Prometheus endpoints:

global:
  telemetry:
    enabled: true
    web-listen: ":9165"
    web-path: "/metrics"
    prometheus-prefix: "dnscollector"

Key Performance Metrics

Throughput Metrics

  • dnscollector_worker_ingress_traffic_total
    Total number of DNS messages received by each worker. Reflects successful handoff to the next internal worker only, not external delivery guarantees. Message loss due to temporary output unavailability is tracked via dnscollector_worker_discarded_traffic_total.

  • dnscollector_worker_egress_traffic_total
    Total number of DNS messages successfully passed to the next stage by each worker.
    This does not guarantee delivery to external systems.

  • dnscollector_worker_discarded_traffic_total
    Counts messages dropped when worker output channels are full or when messages are intentionally discarded to avoid blocking the pipeline or when external system are temporary unavailability of the output.

  • dnscollector_policy_forwarded_total
    Total number of DNS messages forwarded by policy rules.

  • dnscollector_policy_dropped_total
    Total number of DNS messages dropped by policy rules.

Runtime & System Metrics (Go runtime)

The following metrics are automatically exposed by the Prometheus Go client and are not specific to dnscollector:

  • go_goroutines
    Current number of active goroutines.

  • go_memstats_alloc_bytes
    Number of bytes allocated and still in use.

  • go_memstats_heap_alloc_bytes
    Heap memory currently allocated.

  • process_cpu_seconds_total
    Total user and system CPU time spent in seconds.

Grafana Dashboard

A pre-built Grafana dashboard is available for comprehensive monitoring:

# Import the dashboard JSON
curl -O https://raw.githubusercontent.com/dmachard/DNS-collector/main/docs/dashboards/grafana_exporter.json

Performance Dashboard

Buffer Optimization

Understanding Buffers

All collectors and loggers use buffered channels for data flow. Buffer sizing is critical for high-throughput scenarios.

Buffer Configuration

global:
  worker:
    buffer-size: 8192    # Default size
    # For high traffic, consider: 16384, 32768, or 65536

Buffer Full Warning

If you see this warning, increase your buffer size:

logger[elastic] buffer is full, 7855 packet(s) dropped