CockroachDBGrafana

Critical node detection indicating potential data unavailability

critical
reliabilityUpdated Feb 21, 2026

Nodes become critical when their failure would cause replica unavailability. Detecting critical nodes before termination prevents data loss and service disruption. This is indicated by the /_status/critical_nodes endpoint showing non-empty criticalNodes array.

How to detect:

Query the /_status/critical_nodes endpoint showing criticalNodes array is non-empty, or report.unavailable or report.underReplicated contain ranges. This indicates nodes whose failure would cause data unavailability.

Recommended action:

When critical nodes detected: (1) DO NOT terminate nodes in criticalNodes list, (2) Investigate report.violatingConstraints to identify zone config issues, (3) Check report.underReplicated ranges and allow time for rebalancing, (4) If decommissioning, verify report.unavailableNodeIds is empty before proceeding. Add capacity if cluster is chronically under-provisioned.