coalesce-labs/trace-analysis

Trace Analysis

Investigate distributed traces, transaction performance, and slow requests using Sentry.

Usage

/trace-analysis <trace-id-or-query>

Examples:
  /trace-analysis "a4d1aae7216b47ff8117cf4e09ce9d0a"
  /trace-analysis "slow API requests to /checkout"
  /trace-analysis "traces with >5 second response time"
  /trace-analysis "performance issues in payment service"

What This Analyzes

Trace Components

• Transaction spans (API calls, DB queries, external services)
• Timing breakdown per span
• Parent-child span relationships
• Span operations and descriptions

Performance Metrics

• Total transaction duration
• Time spent in each service
• Database query performance
• External API latency
• Network overhead

Bottleneck Identification

• Slowest spans in trace
• Sequential vs parallel operations
• N+1 query detection
• Inefficient operations

Example Analyses

Specific Trace Investigation

/trace-analysis "Analyze trace abc123def456: where's the bottleneck?"

Performance Pattern

/trace-analysis "Why are checkout API requests slow today?"

Service Comparison

/trace-analysis "Compare performance of payment service vs order service"

Database Performance

/trace-analysis "Find traces with slow database queries in user service"

Output Format

Analysis includes:

Trace Overview:

• Transaction name and operation
• Total duration
• Timestamp
• Environment and release

Span Breakdown:

Transaction: POST /api/checkout (2.4s)
├─ Authentication (45ms)
├─ Database Query: SELECT users (120ms)
├─ External API: Payment Gateway (1.8s) ⚠️ SLOW
├─ Database Query: INSERT orders (230ms)
└─ Email Service (180ms)

Performance Insights:

• Slowest operations
• Time distribution (pie chart/percentages)
• Parallel vs sequential execution
• Optimization opportunities

Recommendations:

• Cache frequently accessed data
• Optimize specific queries
• Implement async processing
• Add timeouts for external calls

Advanced Analysis

Multi-Trace Patterns

/trace-analysis "Find common bottlenecks across all slow checkout traces today"

Service Dependencies

/trace-analysis "Map service call chain for failed transactions"

Error Correlation

/trace-analysis "Traces that resulted in errors: what went wrong before?"

Integration Opportunities

With Error Debugging

# Enable debugging plugin (if not already)
/plugin enable catalyst-debugging

# Combine trace and error analysis
> "Show me the trace for the transaction that caused error ISSUE-456"

With Code Changes

After identifying bottleneck:

/catalyst-dev:create-plan "Optimize the slow payment gateway call identified in trace analysis"

Performance Optimization Workflow

1. Identify Slow Transactions

/trace-analysis "transactions with >2s response time in last hour"

2. Analyze Bottlenecks

> "Drill into the slowest trace: which span is the problem?"

3. Root Cause

> "Why is the database query taking 800ms?"

4. Implement Fix

/catalyst-dev:create-plan "Add database index for user lookups based on trace analysis"

5. Verify Improvement

> "After deploy, compare trace durations before and after"

Tips

1. Start with aggregates - "slow checkouts" before diving into specific traces
2. Look for patterns - One slow trace might be an outlier, many indicate systemic issue
3. Check external dependencies - Third-party APIs often cause slowdowns
4. Consider concurrency - Sequential operations that could be parallel
5. Database queries - N+1 queries, missing indexes, inefficient queries

Context Cost

Plugin uses ~20k tokens. Disable after analysis:

/plugin disable catalyst-debugging

---

See also: /catalyst-debugging:debug-production-error, /catalyst-debugging:error-impact-analysis