PostgreSQL rollback ratio % alert


The Rollback Ratio % alert notifies you when its configured condition is met on PostgreSQL instances so you can investigate and respond.

Screenshot pending: Mini DBA PostgreSQL Rollback Ratio % alert screenshot placeholder

Alert summary

  • Platform: PostgreSQL
  • Alert category: Query
  • Default enabled: true
  • Default evaluation frequency: Minute
  • Threshold label: % rollbacks
  • Unit: %

What Mini DBA checks

Mini DBA describes this alert as: A high rollback ratio can indicate application errors, retry loops, failed transactions, or aborted work. Mini DBA evaluates this alert once a minute so changes are detected quickly. Because duration is used, Mini DBA can avoid treating a single short spike as a full incident when the condition clears quickly.

How this alert helps

This alert highlights a high share of rolled-back transactions, which can indicate application errors, retry loops, failed transactions, deadlocks, or work abandoned by clients.

When to enable it

Enable it on any monitored PostgreSQL instance where this condition would require an administrator to investigate or communicate with the application team. Tune noisy development systems after you have a short baseline.

Threshold guidance

Threshold meaning: % rollbacks. Major threshold: 30 %. Minor threshold: 10 %. Comparison direction: over. Duration is used, so prefer requiring the condition to persist before paging people for transient spikes. For an over-threshold alert, decreasing a threshold makes that severity fire sooner; increasing it tolerates more load or pressure. Set the minor threshold as an early-warning level and the major threshold at the highest acceptable value for the service.

Remediation for an active alert

Identify the applications, users, statements, and errors associated with rollbacks. Correct transaction-handling defects, retry loops, constraint failures, timeouts, or deadlocks; index and statistics changes are relevant only when evidence shows they caused the failed work.

Investigation workflow

  1. Confirm the rollback ratio, transaction volume, affected database, and time window.
  2. Correlate rollbacks with application logs, database errors, deadlocks, timeouts, and retry behaviour.
  3. Identify whether one client, deployment, or statement caused the change.
  4. Confirm that both the rollback ratio and the underlying errors return to baseline.

Avoiding alert noise

If the alert creates repeated noise, review whether it is enabled on the correct instances, whether maintenance windows are configured, and whether the minor and major routes match the real business impact. Keep enough history to prove the new setting before changing it broadly.

Related pages