Monitoring Configuration

SLOzy exposes Prometheus-compatible metrics and ships with pre-configured alert rules and Grafana provisioning.

Prometheus Targets

The main prometheus/prometheus.yml defines the following scrape jobs:

Job	Target	Scrape Interval	Purpose
slozy-web	slozy-web-service:80	30s	Application metrics (/metrics)
kubernetes-pods	Auto-discovery via pod annotations	15s	All annotated pods
kubernetes-api-server	Kubernetes API	15s	Cluster health
postgres-exporter	postgres-exporter:9187	15s	Database metrics
redis-exporter	redis-exporter:9121	15s	Redis metrics

For non-Kubernetes deployments, use the simplified deploy/prometheus.yml which targets localhost:8080:

scrape_configs:
  - job_name: 'slozy'
    static_configs:
      - targets: ['localhost:8080']
    metrics_path: /metrics
    scrape_interval: 15s

Metrics Endpoint

The application exposes a /metrics endpoint (enabled via METRICS_ENABLED=true, default port 9090). Key metric categories:

• http_requests_total — request count by method, path, status code
• http_request_duration_seconds — histogram of response times
• slo_status — current burn-rate status per SLO
• cache_hits / cache_misses — cache efficiency
• websocket_connections_total — real-time connection count
• process_resident_memory_bytes — memory usage

Alert Rules

Alerting rules are defined in prometheus/alerts/alerts.yml (Kubernetes) and deploy/slozy-alerts.yml (standalone). Pre-configured alerts:

Alert Name	Severity	Condition
ApplicationDown	critical	up{job="slozy-web"} == 0 for 2m
HighErrorRate	warning	5xx rate > 5% over 5m
HighLatency	warning	P95 latency > 1s over 5m
VeryHighLatency	critical	P99 latency > 5s over 2m
HighMemoryUsage	warning	Container memory > 80% for 5m
DatabaseDown	critical	Postgres pod unreachable for 2m
HighDatabaseConnections	warning	Active connections > 80
CriticalSLOViolation	critical	Any SLO in critical status for 5m
MultipleSLOViolations	warning	3+ SLOs in warning/critical for 10m

Alertmanager is configured to deliver notifications via Slack, email, or PagerDuty. Set alertmanager:9093 as the target in your Prometheus config.

Grafana Dashboards

The docker-compose.yml includes a Grafana container (port 3001) with provisioning from grafana/provisioning/ and dashboard definitions from grafana/dashboards/.

Conceptual dashboards to create:

1. SLO Overview — burn-rate status for all SLOs, error budget remaining, time to exhaustion
2. Application Performance — request rate, latency (P50/P95/P99), error rate by endpoint
3. Database — connection pool usage, query duration, cache hit ratio
4. Infrastructure — CPU/memory per pod, disk usage, network I/O

Default Grafana credentials: admin / admin. Change these in production.

Running the Stack

# Full stack with monitoring
docker-compose up -d

# Or bring up monitoring standalone
docker-compose up -d prometheus grafana

# Access Grafana
open http://localhost:3001

Retention

Component	Setting	Default
Prometheus TSDB	retention.time	30d
Prometheus TSDB	retention.size	10GB
Grafana	Snapshots / DB	Configurable via GF_* env vars

Adjust retention in prometheus/prometheus.yml under the storage.tsdb section.