Vector

High-performance observability data pipeline for collecting, transforming, and routing logs, metrics, and events. Built on Vector with a pre-built musl static binary on Alpine Linux.

Quick Start

# Pull the image
docker pull ghcr.io/oorabona/vector:latest

# Run with default demo config (generates sample logs → console)
docker run -d --name vector -p 8686:8686 ghcr.io/oorabona/vector:latest

# Run with custom config
docker run -d --name vector \
  -v /path/to/vector.yaml:/etc/vector/vector.yaml:ro \
  -p 8686:8686 \
  ghcr.io/oorabona/vector:latest

# Monitor pipeline performance
docker exec vector vector top

Build

# Build with latest upstream version
./make build vector

# Build with specific version
./make build vector 0.53.0

Build Args

Configuration

Vector uses a YAML configuration file at /etc/vector/vector.yaml. The default config runs a demo pipeline that generates sample JSON logs and prints them to the console.

For production, mount your own configuration:

# vector.yaml — collect Docker logs, enrich, send to PostgreSQL
api:
  enabled: true
  address: "0.0.0.0:8686"

sources:
  docker_logs:
    type: docker_logs

transforms:
  enrich:
    type: remap
    inputs: ["docker_logs"]
    source: |
      # Parse structured fields and add metadata
      .host = get_hostname!()
      .environment = get_env_var("VECTOR_ENV") ?? "production"
      .processed_at = now()

sinks:
  postgresql:
    type: postgres
    inputs: ["enrich"]
    endpoint: "postgresql://vector:password@postgres:5432/observability"
    table: "logs"
    encoding:
      codec: json

PostgreSQL Sink (Vendor-Free Observability)

Vector pairs with our postgres:full image (TimescaleDB + pgvector + ParadeDB + Citus) for a complete vendor-free observability stack:

• Logs → PostgreSQL with TimescaleDB hypertables for time-series queries
• Metrics → PostgreSQL with continuous aggregates for efficient rollups
• Search → ParadeDB BM25 index for full-text search over logs

See examples/docker-compose.yaml for a ready-to-run stack.

Ports

Additional ports depend on your configured sources (e.g., 514 for syslog, 9000 for Prometheus scrape, 4317 for OpenTelemetry).

Volumes

Health Check

Built-in health check via Vector API:

GET http://localhost:8686/health → {"ok": true}

Management

# Reload configuration without restart (graceful)
docker kill --signal=HUP vector

# View real-time pipeline metrics
docker exec vector vector top

# Validate configuration before deploying
docker exec vector vector validate /etc/vector/vector.yaml

# View logs
docker logs -f vector

Architecture

Sources (inputs)          Transforms (VRL)            Sinks (outputs)
┌─────────────┐          ┌──────────────────┐        ┌──────────────┐
│ docker_logs  │───┐      │ remap (enrich)   │───┐    │ PostgreSQL   │
│ syslog       │───┤─────>│ filter (route)   │───┤───>│ S3 / GCS     │
│ file         │───┤      │ aggregate        │───┤    │ Elasticsearch│
│ prometheus   │───┘      │ dedupe           │───┘    │ console      │
└─────────────┘          └──────────────────┘        └──────────────┘

Runs as non-root user vector (uid 1000). Pre-built musl-linked static binary downloaded from GitHub releases — supports x86_64 and aarch64.

Security Considerations

• Runs as non-root user vector (uid 1000)
• No shell access required — single static binary
• Mount config read-only (:ro)
• Use environment variables or secrets for sink credentials
• Buffer data at /var/lib/vector/ — mount a volume for persistence and to avoid data loss on restart
• TLS supported for all network sources and sinks

Dependencies

Links

• Vector Documentation
• Vector Configuration Reference
• VRL (Vector Remap Language)
• Available Sources
• Available Sinks
• Vector Blog