Monitor deployments

Go beyond simple threshold alerts to detect meaningful performance changes and potential issues before they impact people using your product.

How it Works

Cased’s deployment monitoring enables AI analysis to detect anomalies and handle post-deploy observability:

1. Baseline Establishment: Cased analyzes historical metrics to establish normal performance baselines for your applications
2. Real-time Monitoring: During and after deployments, Cased continuously monitors key metrics against these baselines
3. Anomaly Detection: AI algorithms detect significant deviations from normal patterns, not just threshold breaches
4. Contextual Analysis: Cased correlates multiple metrics to provide meaningful insights about deployment health

Monitored Metrics

Resource Metrics

• CPU Utilization: Detects unusual CPU spikes or sustained high usage
• Memory Usage: Monitors memory pressure and potential memory leaks
• Disk I/O: Tracks read/write latency and throughput changes
• Network Traffic: Monitors network utilization and packet loss

Application Metrics

• Error Rates: Detects increases in 4xx and 5xx HTTP errors
• Response Times: Monitors API latency and response time degradation
• Throughput: Tracks request volume and processing rates
• Queue Depths: Monitors message queue backlogs and processing delays

Infrastructure Metrics

• Load Balancer Health: Monitors healthy/unhealthy host counts
• Database Performance: Tracks query latency and connection counts
• Cache Hit Rates: Monitors cache performance and effectiveness
• Container Health: Tracks container restart rates and resource limits

Anomaly Detection Thresholds

Cased uses intelligent thresholds based on statistical analysis:

Severity Levels

• Medium (1.25x baseline): 25% increase from normal - worth investigating
• High (1.5x baseline): 50% increase from normal - significant issue
• Critical (2.0x baseline): 100% increase from normal - urgent attention required

Metric-Specific Thresholds

• CPU/Memory: Resource constraints that could impact performance
• Error Rates: More sensitive thresholds since small increases matter
• Latency: Response time increases that affect user experience

Supported Data Sources

AWS CloudWatch

• Native integration with AWS services
• Comprehensive metric coverage for EC2, RDS, ELB, Lambda, and more
• Custom metrics and dashboards
• Automated alerting and dashboard creation

Datadog

• Full-stack monitoring across infrastructure and applications
• Custom metrics and synthetic monitoring
• APM (Application Performance Monitoring) integration
• Log correlation and analysis

Getting Started

To use deployment monitoring:

1. Connect Data Sources: Ensure your CloudWatch or Datadog integration is configured
2. Deploy with Monitoring: Cased automatically monitors deployments when data sources are connected
3. Review Anomalies: Check the Mission Control dashboard for detected anomalies

Example Monitoring Scenarios

Post-Deployment Monitoring

After deploying version 2.1.3:
- CPU usage increased 45% above baseline (HIGH severity)
- Error rate increased 150% (CRITICAL severity)
- Memory usage within normal range
- Database latency increased 30% (MEDIUM severity)

Recommendation: Investigate error rate spike and CPU usage

Resource Anomaly Detection

Deployment monitoring detected:
- Memory usage consistently 80% above baseline
- Potential memory leak in new code
- Container restart rate increased 3x
- Database connection pool exhaustion

Action: Rollback recommended, investigate memory management

Performance Degradation

Gradual performance degradation detected:
- API response times increased 60% over 2 hours
- Database query latency doubled
- Cache hit rate decreased 25%
- No error rate increase

Analysis: Database performance issue, possibly query optimization needed

Best Practices

Monitoring Strategy

1. Establish Baselines: Allow sufficient time for baseline establishment before relying on anomaly detection
2. Gradual Rollouts: Use canary deployments to limit blast radius of issues
3. Multiple Metrics: Don’t rely on single metrics - correlate multiple data points
4. Regular Reviews: Periodically review anomaly detection accuracy and adjust thresholds

Integration with CI/CD

Cased’s deployment monitoring integrates seamlessly with your existing CI/CD pipeline:

• Automatic Activation: Monitoring starts automatically when deployments are detected
• Status Reporting: Deployment status updated based on monitoring results
• Rollback Triggers: Can trigger automated rollbacks based on anomaly severity
• Pipeline Integration: Works with GitHub Actions