fleet/tools/signoz

Running SigNoz locally with Fleet

SigNoz is an open-source observability platform that provides traces, metrics, and logs in a single UI. This guide explains how to run SigNoz locally for Fleet development with optimized settings for reduced latency.

Prerequisites

• Docker and Docker Compose
• A locally-built Fleet server (see Testing and local development)

Setup

1. Clone the SigNoz repository at a specific release:

git clone --branch v0.110.1 --depth 1 https://github.com/SigNoz/signoz.git
cd signoz/deploy

2. Modify the SigNoz UI port to avoid conflict with Fleet (which uses port 8080):

In docker/docker-compose.yaml, change the signoz service port mapping:

services:
  signoz:
    ports:
      - "8085:8080"  # Changed from 8080:8080 to avoid conflict with Fleet

3. (Optional) For reduced latency during development, modify docker/otel-collector-config.yaml:

processors:
  batch:
    send_batch_size: 10000
    send_batch_max_size: 11000
    timeout: 200ms  # reduced from 10s for dev
  # ...
  signozspanmetrics/delta:
    # ...
    metrics_flush_interval: 5s  # reduced from 60s for dev

4. Start SigNoz:

cd docker
docker compose up -d

Give it a minute for all services to initialize. The SigNoz UI will be available at http://localhost:8085.

Configuring Fleet

Start the Fleet server with OpenTelemetry tracing and logging enabled:

export FLEET_LOGGING_TRACING_ENABLED=true
export FLEET_LOGGING_OTEL_LOGS_ENABLED=true
export OTEL_SERVICE_NAME=fleet
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

./build/fleet serve

Note: All log levels (including debug) are always sent to SigNoz regardless of the --logging_debug flag. That flag only controls stderr output.

Multiple environments

When you point more than one Fleet deployment (for example production, staging, and dev-local) at the same SigNoz instance, keep the service name the same and tell them apart by environment instead.

Leave OTEL_SERVICE_NAME at its default of fleet on every deployment so they stay grouped under one service, then set the environment through OTEL_RESOURCE_ATTRIBUTES. Set both of these attributes to the same value:

• deployment.environment.name is the current OpenTelemetry semantic convention.
• deployment.environment is the older, now-deprecated attribute that SigNoz still keys off (the dashboard environment selector reads this one).

export OTEL_SERVICE_NAME=fleet
export OTEL_RESOURCE_ATTRIBUTES=deployment.environment.name=dev-local,deployment.environment=dev-local

Use a distinct value on each deployment and keep both attributes in sync so the dashboard environment selector resolves to a single deployment.

Low-latency configuration (optional)

For faster feedback during development, you can reduce the batch processing delays on the Fleet side:

# Batch span processor delay (default 5000ms)
export OTEL_BSP_SCHEDULE_DELAY=1000

# Log batch processor settings
export OTEL_BLRP_EXPORT_TIMEOUT=1000
export OTEL_BLRP_SCHEDULE_DELAY=500
export OTEL_BLRP_MAX_EXPORT_BATCH_SIZE=1

./build/fleet serve

Using SigNoz

After starting Fleet with the above configuration, you should start seeing traces, logs, and metrics in SigNoz UI at http://localhost:8085.

Pre-canned dashboards

JSON exports of Fleet-specific SigNoz dashboards live alongside this README. Import them from the SigNoz UI via Dashboards → New dashboard → Import JSON (top-right dropdown).

• database_custom_dashboard.json — MySQL query metrics (RPS, latency, slow queries) derived from db.sql.* instrumentation.
• host_cache_dashboard.json — Redis-backed host lookup cache (LoadHostByNodeKey / LoadHostByOrbitNodeKey). Shows hit rate over time, lookups/sec by result, errors/sec by op, and invalidations/sec by write-path reason. Requires FLEET_REDIS_HOST_CACHE_ENABLED=true (default on).
• http_errors_dashboard.json — Fleet HTTP errors (the "Errors" signal of the RED method / Google SRE Golden Signals). Shows 4XX client errors and 5XX server errors from fleet.http.client_errors / fleet.http.server_errors, broken down by error.type.

Each dashboard includes an environment selector (a dynamic dashboard variable on the deployment.environment resource attribute) so you can scope panels to a single deployment or view all. It defaults to ALL. Fleet always emits deployment.environment (default value default, overridable via OTEL_RESOURCE_ATTRIBUTES), so the selector populates on any instance Fleet reports to.