Skip to main content

Setup & Configuration

Enable Arrow‑based metric ingestion using Micrometer and DazzleDuck SQL Server.

Requirements​

  • Java 21+
  • Micrometer 1.10+
  • Running DazzleDuck SQL Server (HTTP ingestion enabled)

How It Works​

This module provides a Micrometer MeterRegistry that periodically publishes metric snapshots as Apache Arrow rows and ingests them into DazzleDuck SQL Server over HTTP.

Registry creation and wiring is configuration‑driven.

Registry Creation​

Create and register the Arrow‑backed registry using the factory:

MeterRegistry registry = MetricsRegistryFactory.create();

Metrics.addRegistry(registry);

This returns a CompositeMeterRegistry that contains an ArrowMicroMeterRegistry instance.

Configuration Source​

All settings are loaded from Typesafe Config:

dazzleduck_micrometer { ... }

Configuration can be provided via:

  • application.conf
  • Environment variables
  • JVM system properties

Example Configuration​

dazzleduck_micrometer {
  application_id   = "orders-service"
  application_name = "orders"
  application_host = "host-1"

  http {
    base_url = "http://localhost:8081"
    target_path = "/v1/ingest"
    username = "admin"
    password = "admin"
    http_client_timeout_ms = 5000
  }

  max_in_memory_bytes = 10485760   # 10 MB
  max_on_disk_bytes   = 1073741824 # 1 GB

  min_batch_size = 1048576 # 1 MB
  max_batch_size = 16777216 # 16 MB
  max_send_interval_ms = 1000

  retry_count = 3
  retry_interval_ms = 1000
  transformations = []
  partition_by = []
}

Publishing Model​

  • Metrics are published on a fixed step interval (default: 10s)
  • Each publish cycle sends a snapshot of all meters
  • Publishing is non‑blocking and failure‑tolerant

Graceful Shutdown​

Always close the registry on application shutdown:

registry.close();

This ensures:

  • Final metric flush
  • Sender shutdown
  • Resource cleanup