Configure Profiler Reaction

The Profiler Reaction Reaction A component that receives query result changes and takes action on them. Learn more collects and reports performance metrics for queries Continuous Query A query that runs continuously, maintaining an always-current result set as data changes. Learn more . Use it to monitor query performance, identify bottlenecks, and optimize your Drasi Server Drasi Server A standalone Data Change Processing server running as a process or Docker container. Learn more deployment.

Basic Configuration

reactions:
  - kind: profiler
    id: query-profiler
    queries: [my-query]

Configuration Reference

Metrics Collected

The profiler tracks:

Window Configuration

The windowSize controls how many recent results are used for metric calculations:

reactions:
  - kind: profiler
    id: detailed-profiler
    queries: [high-volume-query]
    windowSize: 1000  # Use last 1000 results for metrics

Report Interval

Control how often metrics are reported to the Drasi Server logs:

reactions:
  - kind: profiler
    id: frequent-reports
    queries: [critical-query]
    reportIntervalSecs: 10  # Report every 10 seconds

Sample Output

Profiler output in logs:

[query-profiler] ========== Profiling Report ==========
[query-profiler] Source→Query: ...
[query-profiler] Query Processing: ...
[query-profiler] Query→Reaction: ...
[query-profiler] Reaction Processing: ...
[query-profiler] Total End-to-End: ...
[query-profiler] ======================================

Use Cases

Development Profiling

Quick feedback during development:

reactions:
  - kind: profiler
    id: dev-profiler
    queries: [test-query]
    windowSize: 50
    reportIntervalSecs: 10

Production Monitoring

Long-term production monitoring:

reactions:
  - kind: profiler
    id: production-profiler
    queries: [order-query, inventory-query, customer-query]
    windowSize: 500
    reportIntervalSecs: 60

Performance Testing

Detailed metrics during load tests:

reactions:
  - kind: profiler
    id: load-test-profiler
    queries: [load-test-query]
    windowSize: 2000
    reportIntervalSecs: 5

Multi-Query Profiling

Profile multiple queries:

reactions:
  - kind: profiler
    id: all-queries-profiler
    queries:
      - orders
      - inventory
      - customers
      - analytics
    windowSize: 200
    reportIntervalSecs: 30

Complete Example

host: 0.0.0.0
port: 8080
logLevel: info

sources:
  - kind: postgres
    id: main-db
    host: ${DB_HOST}
    database: production
    user: ${DB_USER}
    password: ${DB_PASSWORD}
    tables:
      - public.orders
      - public.inventory

queries:
  - id: order-updates
    query: "MATCH (o:orders) RETURN o"
    sources:
      - sourceId: main-db

  - id: low-inventory
    query: |
      MATCH (i:inventory)
      WHERE i.quantity < i.reorder_point
      RETURN i
    sources:
      - sourceId: main-db

reactions:
  # Production reactions
  - kind: http
    id: order-webhook
    queries: [order-updates]
    baseUrl: https://api.example.com

  - kind: sse
    id: inventory-dashboard
    queries: [low-inventory]
    port: 8081

  # Profiler for monitoring
  - kind: profiler
    id: query-profiler
    queries: [order-updates, low-inventory]
    windowSize: 200
    reportIntervalSecs: 60

Combining with Log Reaction

Use both profiler and log reactions for comprehensive debugging:

reactions:
  # Debug logging
  - kind: log
    id: debug-log
    queries: [my-query]

  # Performance metrics
  - kind: profiler
    id: profiler
    queries: [my-query]
    windowSize: 100
    reportIntervalSecs: 30

Interpreting Metrics

Throughput

• Low throughput: May indicate source issues or query complexity
• Variable throughput: Normal for event-driven systems
• Sudden drops: Check source connectivity

Latency

• High P99: Occasional slow events (may be acceptable)
• High P50: Consistently slow processing (investigate)
• High variance: Query complexity varies with data

Errors

• Any errors: Investigate immediately
• Increasing errors: Check source connectivity
• Pattern-based: May indicate specific data issues

Performance Tuning Guide

Based on profiler metrics:

High Latency

1. Check query complexity
2. Review source configuration
3. Consider query optimization
4. Increase server resources

Low Throughput

1. Verify source is producing events
2. Check query subscriptions
3. Review reaction configurations
4. Monitor system resources

Memory Issues

1. Reduce windowSize
2. Increase report interval
3. Profile fewer queries

Troubleshooting

No Reports

• Verify profiler is started
• Check query is receiving events
• Review log level configuration

Inaccurate Metrics

• Increase windowSize for stability
• Wait for enough events to accumulate
• Check for event bursts skewing metrics

High Overhead

• Increase reportIntervalSecs
• Reduce windowSize
• Profile only critical queries

Documentation resources

Next Steps

• Operations Guide - Production operations
• Troubleshooting - Debug issues