- Change verify_prod_deployment from 'when: manual' to 'when: on_success'
- Production deployment verification will now run automatically on master branch
- Ensures production deployments are verified just like dev deployments
- Maintains safety with allow_failure: false
- Add PrometheusMetricsIntegrationTest class with 7 comprehensive tests
- Test Prometheus server startup, custom metrics availability, and port conflict handling
- Test metrics increment, different labels, histogram metrics, and production mode
- Use random ports (9470-9570) to avoid conflicts between tests
- Make tests lenient about custom metrics timing (collection delays)
- Update OpenTelemetry configuration to handle MeterProvider conflicts gracefully
- Update documentation to clarify production vs development Prometheus usage
- Ensure metrics are properly exported via OTLP in production
- Verify comprehensive test coverage for CI environments
All 34 OpenTelemetry tests pass successfully.
- Fix commit_date parsing from git logs (was showing 'unknown')
- Add deployment_date field using manage.py modification time
- Improve git log parsing to handle author names with spaces
- Both dates now show in proper format: YYYY-MM-DD HH:MM:SS +timezone
The endpoint now returns:
- commit_date: When the commit was made (from git logs)
- deployment_date: When the code was deployed (from file mtime)
- Comment out DjangoInstrumentor().instrument() to test if it's causing duplicate Host headers
- Remove unused DjangoInstrumentor import
- Keep other instrumentation (database, HTTP client) enabled
- This is a temporary test to isolate the Host header duplication issue
- Add _ot_initialized flag to prevent multiple setup calls
- Make setup_opentelemetry() idempotent
- Handle 'Address in use' error gracefully for Prometheus server
- Prevent OpenTelemetry setup failures due to multiple initialization
- Always initialize OpenTelemetry in Django settings (instrumentation always enabled)
- Remove ENABLE_OPENTELEMETRY feature flag from config.py
- Simplify views.py OpenTelemetry imports to always use real implementation
- Export control now handled by OTEL_EXPORT_ENABLED environment variable only
This ensures OpenTelemetry is properly initialized during Django startup
and the Prometheus metrics server starts correctly.
Since containers restart on content changes, remove TTL-based cache expiration
and cache version information indefinitely. This provides maximum performance
benefit while maintaining correctness.
- Remove timestamp-based cache expiration logic
- Cache version info indefinitely until container restart
- Simplify caching function by removing TTL complexity
- Maintain optimized git log file reading
- Add _start_prometheus_server method to start HTTP server
- Register PrometheusMetricReader collector with prometheus_client REGISTRY
- Parse endpoint to extract host and port for HTTP server
- This fixes the issue where metrics endpoint was not accessible
- Add proper cache expiration with 5-minute TTL
- Optimize git log file reading to avoid loading entire file
- Read only last chunk (1024 bytes) instead of all lines
- Add shorter TTL (30s) for error cases to allow retry
- Improve error handling with UnicodeDecodeError support
- Maintain backward compatibility and security (no subprocess calls)
This fixes the 30-second response time issue by implementing efficient
caching and optimized file I/O operations.
- Add OpenIDErrorHandlingTestCase with 8 test methods
- Test OpenID discovery failures, confirmation failures, and cancellations
- Test template inheritance (openid/failure.html extends error.html)
- Test direct error.html template rendering with authenticated/anonymous users
- Use unittest.mock to simulate OpenID failures without external dependencies
- Verify error messages are properly displayed to users
- Ensure comprehensive coverage of error.html template through OpenID scenarios
All tests pass and maintain existing test coverage.
