At a glance
Connected Clients is the live count of client connections currently established to the Redis instance, taken straight fromconnected_clientsinINFO clients. It is the raw input behind the saturation cards: on its own it tells you how busy the front door is, and when read against the configuredmaxclientsceiling it tells you how close you are to refusing new connections. A steady, predictable count is healthy; sudden steps up or a slow ratchet that never falls back usually means a connection leak in an application pool.
| What it tracks | The number of client connections currently open to the Redis server (connected_clients). |
| Data source | INFO clients → connected_clients. Excludes connections from replicas, which are counted separately. |
| Time window | RT (real-time snapshot). |
| Alert trigger | None set. The paging thresholds live on Clients vs maxclients % and Rejected Connections (24h). |
| Roles | engineering, operations |
What it tracks
Every application worker, queue consumer, session reader, andredis-cli session that holds an open socket to the server counts here. The number should map to the sum of your connection pools across all application instances plus a small allowance for ad-hoc tooling. Two patterns matter: a sharp jump usually follows a deploy that increased pool size or added a new service; a slow, monotonic climb that never recedes is the classic signature of a connection leak, where the application opens sockets it never returns or closes. Both push you toward the maxclients ceiling (10,000 by default on self-hosted Redis, lower on small managed nodes), at which point new connections are refused and downstream services start erroring. Because this card is the unscaled raw number, pair it with the saturation gauge to understand how much headroom remains.
Reconciling against the source
Verify directly withINFO clients (the connected_clients field) or enumerate live sessions with CLIENT LIST and count the rows. On ElastiCache and MemoryDB, cross-check the CurrConnections CloudWatch metric, and remember that NewConnections shows the churn rate, which is useful for spotting a pool that is opening and closing sockets rather than reusing them.