At a glance
The live count of asynchronous replicas currently connected to and replicating from this MariaDB primary. It answers a simple availability question: how many standbys are actively following the primary right now, ready to absorb read traffic or be promoted in a failover? When this number drops below your expected topology, your redundancy has quietly degraded even though the primary itself looks healthy.
| What it tracks | Active Async Replicas: the number of replica hosts with a live replication connection to the primary, both I/O and SQL threads running. The detail line is Active Async Replicas for the selected period. |
| Data source | MariaDB SHOW REPLICA HOSTS (formerly SHOW SLAVE HOSTS) on the primary, counting registered replicas, cross-checked against Slave_running / Replica_running status on each replica via SHOW REPLICA STATUS. |
| Time window | RT: real-time, refreshed on each poll. |
| Alert trigger | None on this card. Pair it with Failover Readiness and Async Replication Lag (seconds) for the threshold alerts. |
| Roles | DBA, platform, SRE |
What it tracks
This card surfaces the count of active asynchronous replicas for the selected period, read live from the primary’s view of its replication topology. A replica counts as active only when it holds a live connection to the primary and both replication threads are running: the I/O thread pulling binary-log events and the SQL (apply) thread executing them. A replica that has fallen off the network, stopped with a duplicate-key error, or been taken down for maintenance drops out of the count immediately, which is why this number is your at-a-glance redundancy gauge. Read it against your intended topology: if you run a primary with two read replicas and the card shows one, a standby has gone silent and your failover and read-scaling capacity is halved, even though the primary is serving writes normally. The metric is asynchronous-replica specific; Galera (synchronous, multi-primary) cluster membership is tracked separately by the Galera cards.Reconciling against the source
To verify the count directly, runSHOW REPLICA HOSTS; on the primary to list registered replicas, then SHOW REPLICA STATUS\G on each replica and confirm both Slave_IO_Running and Slave_SQL_Running (or the Replica_* equivalents) read Yes. On managed services the active-replica count appears in the provider’s replication view: Amazon RDS / Aurora for MariaDB shows read replicas and their state in the console and via the ReplicaLag / replication-state metrics, while SkySQL and Azure Database for MariaDB surface replica membership in their own consoles. A brief discrepancy between the card and a manual check is usually a timing artefact between poll intervals.