At a glance
Active Nodes (status=live) counts the CockroachDB nodes currently reporting a live liveness status to the cluster. It is the count of nodes that are heartbeating their node-liveness range and are eligible to hold leases and serve traffic. A node that has crashed, been network-partitioned, or is mid-restart drops out of this count within a few seconds, so the card is a fast, real-time pulse on how much of your cluster is actually carrying load.
| What it tracks | Active Nodes (status=live) for the selected period, read from node-liveness state. |
| Data source | crdb_internal.gossip_liveness (and the liveness.livenodes time-series metric), which exposes each node’s is_live flag. The DB Console Cluster Overview shows the same live/suspect/dead breakdown. |
| Time window | RT (real-time, refreshed on each poll). |
| Alert trigger | None on this card directly. Watch Cluster Node Count for the “node lost” alert and Unavailable or Under-Replicated Ranges for quorum risk. |
| Roles | DBA, platform, SRE |
What it tracks
This card reports the number of nodes whose liveness status islive at poll time. CockroachDB tracks node liveness through a dedicated heartbeat: every node periodically writes to the node-liveness range, and a node is live only while its lease on that record is valid (the default liveness duration is roughly 9 seconds). When a node stops heartbeating, it transitions to suspect and then, after the time-until-store-dead window (5 minutes by default), to dead. Active Nodes is the headcount of the live set: the nodes that can currently take on range leases, accept SQL connections, and participate in Raft quorum. A healthy steady-state value equals your provisioned node count. A dip means a node has stopped heartbeating, which precedes lease transfers, possible under-replication, and rebalancing.
Reconciling against the source
To confirm the figure natively, runSELECT node_id, is_live FROM crdb_internal.gossip_liveness ORDER BY node_id; or open the DB Console Cluster Overview, where nodes are grouped as Live, Suspect, and Dead. On CockroachDB Cloud, the same live-node count appears on the cluster’s Overview page. Brief differences are usually a node that is suspect (heartbeat just missed) rather than truly dead, so allow a few seconds for liveness to settle before treating a dip as a real node loss.