Current Versions
Application Dependencies
Infrastructure Components (Docker Compose)
Kubernetes Base Manifests
Note: Keycloak now uses a pre-optimized GHCR image with
--optimized mode, enabling readOnlyRootFilesystem: true for enhanced security. Health check endpoints are at /authn/health/* paths to match the KC_HTTP_RELATIVE_PATH=/authn configuration for Traefik gateway routing.
Helm Chart Dependencies
Upgrade Summary (2025-10-14)
Critical Updates
-
OpenFGA: v1.5.0 → v1.10.2 (5 minor versions)
- Risk: Medium
- Breaking Changes: None affecting our usage
- Features Added: ReverseExpand improvements, performance optimizations
- Action: Test authorization flows after upgrade
-
Keycloak: 23.0 → 26.4.0 (3 major versions)
- Risk: High
- Breaking Changes: Review Keycloak Upgrading Guide
- Features Added: Passkeys, improved OAuth 2.0 support, federated client auth
- Action: Review realm configurations, test SSO flows
-
OpenTelemetry Collector: 0.91.0 → 0.137.0 (46 versions!)
- Risk: High
- Breaking Changes: Multiple config schema changes
- Action: Review otel-collector.yaml configuration compatibility
Major Updates
-
Jaeger: 1.53.0 → 1.74.0 (21 versions)
- Risk: Medium
- Features: Performance improvements, v2 architecture available
- Action: Test trace visualization
-
PostgreSQL: 15-alpine → 16-alpine (1 major version)
- Risk: Medium
- Breaking Changes: Minor, mostly backward compatible
- Action: Test database migrations, backup before upgrade
-
Prometheus: v2.48.0 → v3.2.1 (Major version bump!)
- Risk: High
- Breaking Changes: Review Prometheus 3.0 Migration Guide
- Features: Native histograms, improved performance
- Action: Test PromQL queries, review dashboard compatibility
-
Grafana: 10.2.3 → 11.5.1 (1 major version)
- Risk: Medium
- Features: Enhanced alerting, new visualizations
- Action: Test dashboards, review plugin compatibility
New Components
-
PostgreSQL StatefulSet (New)
- Now deployed as StatefulSet with persistent storage
- Includes multi-database initialization script
- Shared by OpenFGA and Keycloak
-
OpenFGA Kubernetes Deployment (New)
- High-availability setup with 2 replicas
- Uses PostgreSQL backend (previously memory only)
- Proper health checks and resource limits
Breaking Changes
Keycloak 23.0 → 26.4.0
Potential Issues:- Deprecated features from 23.x may be removed in 26.x
- Admin console UI changes
- Some REST API endpoints may have changed
- Backup Keycloak database:
kubectl exec -it postgres-0 -- pg_dump -U postgres keycloak > keycloak-backup.sql - Review release notes
- Test authentication flows in staging environment
- Update client configurations if needed
OpenTelemetry Collector 0.91.0 → 0.137.0
Configuration Changes:- Review
monitoring/otel-collector/otel-collector.yaml - Some processor configurations may need updates
- New recommended practices for pipeline setup
- Validate config:
docker run otel/opentelemetry-collector-contrib:0.137.0 validate --config=/etc/otel-collector-config.yaml - Review changelog
- Test metric export to Prometheus
- Test trace export to Jaeger
Prometheus v2 → v3
Breaking Changes:- Native histogram format changes
- Some PromQL functions updated
- TSDB format changes (backward compatible for reads)
- Review dashboards for deprecated PromQL functions
- Test alert rules compatibility
- Backup Prometheus data:
kubectl cp prometheus-xxx:/prometheus ./prometheus-backup - Consider running v2 and v3 in parallel initially
Compatibility Testing Checklist
Before Deploying
- Review all release notes and changelogs
- Backup all databases (PostgreSQL, Redis if persistent)
- Export Keycloak realm configurations
- Export Grafana dashboards
- Document current OpenFGA authorization model
After Deploying (Docker Compose)
After Deploying (Kubernetes)
Integration Tests
- Test user authentication (Keycloak)
- Test authorization checks (OpenFGA)
- Test session management (Redis)
- Test LLM API calls with tracing
- Verify traces in Jaeger UI
- Verify metrics in Prometheus
- Verify dashboards in Grafana
- Test health check endpoints
- Verify database connectivity
- Test graceful shutdown/restart
Version Update Policy
Semantic Versioning
We follow this upgrade cadence:- Patch versions (x.y.Z): Apply immediately in development, test, then production
- Minor versions (x.Y.z): Test in staging for 1 week before production
- Major versions (X.y.z): Extended testing (2-4 weeks), create rollback plan
Update Frequency
- Security patches: Within 48 hours
- Minor updates: Monthly
- Major updates: Quarterly or as needed
- Dependency updates: Bi-weekly (automated via Dependabot)
Rollback Procedures
Docker Compose Rollback
Kubernetes Rollback
References
- OpenFGA Releases
- Keycloak Releases
- OpenTelemetry Collector Releases
- Jaeger Releases
- PostgreSQL Release Notes
- Prometheus Releases
- Grafana Releases
Support Matrix
Known Issues
Keycloak 26.4.0
- Some older themes may need updates
- Check realm import/export compatibility
- Review custom extensions for API changes
OpenTelemetry Collector 0.137.0
- Configuration validation may be stricter
- Some deprecated processors removed
- Review pipeline configurations for new best practices
Prometheus 3.2.1
- Native histograms enabled by default (can disable)
- TSDB compaction behavior changed
- Some legacy recording rules may need updates