Phantom incidents are anomalies MyDBA spots on its own — unusual behaviour that isn't an outage yet, but could turn into one. Detection runs constantly in the background, with nothing to set up.
Threshold alerts only cover the failure modes you predicted. The incidents that hurt most are usually the ones nobody wrote a rule for.
Phantom incident detection runs continuously in the background against the telemetry MyDBA already collects. There are no rules to write and no thresholds to guess.
Threshold alerts only catch what you predicted. Phantom incidents catch unusual behaviour — lock waits, connection behaviour, error bursts, latency shifts — judged against how your database normally behaves.
Every observation is graded, so a curiosity is recorded quietly while a genuinely dangerous pattern is surfaced prominently. You see what matters without alert fatigue.
Each phantom incident is attributed to the cluster, instance, and database where it originated — so in a multi-instance estate you start the investigation in the right place.
When the same anomaly keeps coming back, MyDBA groups the occurrences into one recurring pattern instead of a pile of duplicates — nightly job contention looks like one story, not thirty incidents.
When an anomaly does turn into a real problem, the phantom incident record tells you when it started and what changed — the perfect starting point for a full incident analysis.
Detection never depends on AI. When you want extra depth, you can manually ask AI to explain a phantom incident in plain language, grounded in the evidence that triggered it.
75+ automated checks across 12 scoring domains
Index analysis across primaries and replicas
Visual plan analysis, guided walkthroughs, and regression detection
Blocker detection, recovery scripts, and runbooks
The most complete TimescaleDB monitoring anywhere
First-class pgvector index and query monitoring
Spatial data inventory, index health, and geometry checks
Manual incident reports with evidence, root cause, timeline, and remediation
Highest-impact queries, the best tuning candidates, and exactly what to do
Manual AI depth for incidents, query plans, tuning, and explanations
Shared managed, dedicated managed, or self-hosted collectors
Rate-of-change, compound conditions, and 11 prebuilt templates
Slot, subscription, apply lag, and PG17 conflict monitoring
Extension detection, auto-install, and provider-specific guides
Shared remediation plans with review workflow
Priority trend analyzers with severity cards and sparklines
Start free with one primary or standalone PostgreSQL instance and all its databases. Monitoring, alerts, query insights, index advice, team access and every non-AI feature included. No replicas. No credit card required. New Free connections get seven days of full collection, then reduced sampling. Free and Pro include 30 days of monitoring history.