At a glance
Active Replicas is the count of replicas currently online and participating in replication for the selected period. On a ReplicatedMergeTree setup, every replica should be active and talking to ClickHouse Keeper (or ZooKeeper). A drop below the expected count means a node has gone read-only or lost its Keeper session, which puts redundancy and write availability at risk.
What it tracks
The card reads the active-replica picture fromsystem.replicas, where each row represents one replicated table on this node and the is_session_expired, is_readonly, and active_replicas columns describe replication health. A healthy cluster shows active_replicas equal to total_replicas for every table; any shortfall means a peer replica is offline, read-only, or has an expired Keeper session and is no longer keeping its copy current. Because this is a real-time (RT) state read, the value reflects the cluster as it stands at the moment of refresh, not an average over time. It carries no alert threshold of its own, but it is the context number you check first when Replication Lag (absolute_delay) or Replication Queue Size (Top Tables) starts climbing: a missing replica is the usual root cause.
Reconciling against the source
Confirm the value directly inclickhouse-client with SELECT database, table, active_replicas, total_replicas, is_readonly, is_session_expired FROM system.replicas. On ClickHouse Cloud, the same query runs in the SQL console and the managed service surfaces node status in its monitoring view.