# Prometheus alert rules for FinData OS (scrape http://<node-lan-ip>:8600/metrics from a private address; the
# endpoint also evaluates docs/OPERATIONS.md health checks on every scrape).
groups:
  - name: findata
    rules:
      - alert: FinDataLakeDiskLow
        expr: findata_lake_free_gb < 120
        for: 30m
        labels: {severity: warning}
        annotations: {summary: "Lake volume below 120 GB free"}
      - alert: FinDataLakeDiskCritical
        expr: findata_lake_free_gb < 60
        for: 5m
        labels: {severity: critical}
        annotations: {summary: "Lake volume below the 60 GB hard stop — ingestion refuses to write"}
      - alert: FinDataIngestDeadJobs
        expr: findata_jobs{status="dead"} > 0
        for: 10m
        labels: {severity: critical}
        annotations: {summary: "Jobs in the dead-letter queue"}
      - alert: FinDataStaleBars
        expr: findata_vendor_lag_days{dataset="market_bars.stock.1m.raw"} > 0
        for: 6h
        labels: {severity: warning}
        annotations: {summary: "1-minute bars are behind the vendor's last_update"}
      - alert: FinDataNoRecentSession
        expr: findata_dataset_freshness_days{dataset="market_bars.stock.1m.raw"} > 5
        for: 1h
        labels: {severity: warning}
      - alert: FinDataBackupStale
        expr: findata_backup_age_hours > 48
        for: 1h
        labels: {severity: warning}
      - alert: FinDataApiLatency
        expr: histogram_quantile(0.95, sum(rate(findata_http_request_seconds_bucket[5m])) by (le, route)) > 5
        for: 15m
        labels: {severity: warning}
      - alert: FinDataQualityErrors
        expr: findata_quality_anomalies_7d{severity="error"} > 100
        for: 1h
        labels: {severity: warning}