Performance Monitoring

GlitchTip captures transaction and span data from your applications, helping you identify slow endpoints and performance regressions.

How It Works

When tracing is enabled in your Sentry SDK, GlitchTip receives transactions (representing top-level operations like HTTP requests) and spans (representing child operations like database queries or API calls within a transaction). These are grouped by endpoint or operation for aggregate analysis.

Configuring Tracing

Enable tracing in your Sentry SDK by setting tracesSampleRate:

import sentry_sdk

sentry_sdk.init(
    dsn="YOUR_DSN",
    traces_sample_rate=0.1,  # capture 10% of transactions
)

Sentry.init({
  dsn: "YOUR_DSN",
  tracesSampleRate: 0.1,
});

Set tracesSampleRate to a value between 0.0 and 1.0. Use a lower rate in production to reduce overhead and data volume.

See your platform's SDK documentation for framework-specific tracing setup.

Viewing Performance Data

Navigate to the Performance page in GlitchTip to see transaction groups with aggregate stats including average duration, p50, p95, request count, and error count.

Click a transaction group to see its span breakdown — which child operations (database queries, HTTP calls, template rendering) are consuming the most time.

Cold Storage

When DuckDB cold storage is enabled, transaction and span data is archived to Parquet files. This enables advanced queries like span grouping across transactions and N+1 query detection.

AI-Assisted Analysis

With the MCP server enabled, AI assistants can query your performance data to:

• Find slow endpoints and database queries
• Detect N+1 query patterns
• Analyze performance trends over time
• Identify regressions between releases