4.0 KiB

Raw Blame History

Release v1.9.4 - Performance & Reliability Improvements

Release Date: 2026-02-07 Branch: mcpserver-impl Tag: v1.9.4

Summary

Second comprehensive review cycle focusing on performance optimization, error handling reliability, configuration fixes, and observability enhancements.

Performance Improvements

SSE Handler Optimizations

strings.Builder - Replaced fmt.Sprintf with pre-allocated strings.Builder
Pre-allocated buffers - SSE event prefixes and data as byte slices
Reduced allocations - Type-safe context keys, JSON buffer pool
Optimized session cleanup - Batched deletion, reduced lock contention

Database Optimizations

Query improvements in documents, rules, and project stores
Better transaction handling
Reduced memory allocations in hot paths

Reliability Improvements

Error Handling Fixes

Fixed silent failures - GetByID, Count errors now properly returned
Error wrapping - All errors use %w for context preservation
HTTP status codes - Proper 404 for not found, 500 for server errors
Type assertions - Safe type assertions with proper error handling

Panic Recovery

Added panic recovery middleware
Panic metrics tracked by path
Structured logging with stack traces
Graceful error response instead of crash

Configuration Fixes

Environment Variable Naming

Fixed incorrect naming in deployment configs:

MCP_RATE_LIMIT → RATE_LIMIT_MCP
IDE_RATE_LIMIT → RATE_LIMIT_IDE
SESSION_RATE_LIMIT → RATE_LIMIT_SESSION

Feature Flags Added

ENABLE_METRICS - Toggle Prometheus metrics
ENABLE_AUDIT_LOGGING - Toggle audit logging
ENABLE_CACHE - Toggle Redis caching

CORS Configuration

CORS_ALLOWED_ORIGINS - Configure allowed origins
CORS_MAX_AGE - Configure preflight cache duration

Observability Enhancements

New Metrics

PanicsTotal - Track recovered panics by path
DBConnectionsActive - Database connection pool stats
DBQueryDuration - Query latency histograms
SLOCompliance - SLO compliance tracking
ErrorBudgetBurnRate - Error budget consumption
SLIValue - Service Level Indicator values

Request Tracing

Correlation ID middleware for request tracing
Request ID propagation
Structured logging with correlation IDs

API Consistency

Route Ordering

Fixed route registration order:

// Before - search would never match
api.GET("/documents/:id", s.getDocument)
api.GET("/documents/search", s.searchDocuments)

// After - specific routes first
api.GET("/documents/search", s.searchDocuments)
api.GET("/documents/:id", s.getDocument)

Response Format Standardization

Consistent error response format across all endpoints
Proper pagination metadata
Standardized field naming

Files Changed

File	Change
`internal/mcp/server.go`	SSE optimizations, reduced allocations
`internal/web/handlers.go`	Error handling fixes, silent failures fixed
`internal/web/server.go`	Panic recovery, correlation ID middleware
`internal/database/*.go`	Query optimizations
`internal/config/config.go`	Config validation improvements
`internal/metrics/metrics.go`	New panic, DB, SLO metrics
`deploy/k8s-deployment.yaml`	Fixed env var names
`deploy/podman-compose.yml`	Fixed env var names, feature flags

Migration from v1.9.3

Update environment variables:

# Old (incorrect)
MCP_RATE_LIMIT=1000
IDE_RATE_LIMIT=500

# New (correct)
RATE_LIMIT_MCP=1000
RATE_LIMIT_IDE=500

Deploy new version:
```
make build
podman-compose up -d
```

Credits

Comprehensive review and improvements by:

Performance Engineer - SSE and database optimizations
SRE Engineer - Error handling and reliability
API Designer - Route consistency and standardization
Platform Engineer - Configuration management
Observability Engineer - Metrics and tracing

Co-Authored-By: Claude Opus 4.5 noreply@anthropic.com

4.0 KiB Raw Blame History