Valkey cluster monitoring

A cluster fails differently than a single instance: hot slots, lagging replicas, stuck migrations, split-brain. Valkey 8.0 made much of this directly observable for the first time - if your tooling reads it.

What to monitor in a Valkey cluster

Topology & roles

Which nodes are primaries, which replicas follow whom, and whether every node agrees on the cluster state.

Slot coverage & distribution

All 16,384 slots assigned, and keys spread evenly - or one shard quietly holding half the keyspace.

Per-slot load

Reads, writes, key counts, and CPU time per slot (CLUSTER SLOT-STATS, Valkey 8.0+). The only way to see a hot slot rather than infer it.

Migrations & resharding

Slot migrations in progress, their pace, and their effect on latency.

Replication health

Replica lag, broken links, and replicas that stopped following.

Failovers & safety

Node failovers, split-brain conditions, and quorum risks - state changes, not statistical spikes.

How BetterDB monitors clusters

Topology graphs

Live cluster visualization: nodes, roles, slot ranges, and health at a glance, with per-node performance stats.

SLOT-STATS heatmaps

Per-slot key counts, reads, writes, and CPU rendered as heatmaps (Valkey 8.0+), plus Prometheus series for the top slots - so hotspot detection and resharding decisions are data-driven.

Cluster-wide slowlog

Slowlogs aggregated across every node into one queryable view, instead of node-by-node SLOWLOG GET.

Safety detectors

State-change anomaly detectors for failovers, split-brain, and stuck replicas (Pro, free in early access) - conditions a Z-score on a single metric can never see.

FAQ

What is CLUSTER SLOT-STATS?

CLUSTER SLOT-STATS is a Valkey 8.0+ command that reports per-slot statistics in cluster mode: key count, expiring keys, total reads, total writes, and CPU time per slot. It makes hot slots directly observable instead of inferred, and supports data-driven resharding. Redis does not have this command; BetterDB renders it as heatmaps and exports Prometheus metrics for the top slots.

How do I find a hot slot or hot shard in Valkey?

On Valkey 8.0+, read CLUSTER SLOT-STATS ordered by reads, writes, or CPU - BetterDB visualizes this as a heatmap and keeps history, so you can see when the hotspot appeared. Combined with hot-key tracking, you can usually go from "this shard is hot" to "this key pattern is responsible" in one investigation.

Does BetterDB monitor Redis Cluster too?

Yes. Topology visualization, per-node stats, cluster-wide slowlog aggregation, and failover detection work on Redis 6+ clusters. The per-slot SLOT-STATS views are Valkey 8.0+ exclusive, and the UI degrades gracefully when the capability is absent.

Does BetterDB support Sentinel deployments?

Yes, Sentinel-managed topologies are supported alongside cluster mode. Note that when connecting through an SSH tunnel, cluster and Sentinel monitoring fan out to other nodes using their advertised addresses directly - so per-node views require network reachability to those nodes.

What cluster metrics should I alert on?

The high-signal ones: slot coverage incomplete, failover events, replication link down or lag growing, split-brain indicators, and sustained per-slot CPU concentration. BetterDB raises these as diagnosed anomalies and exposes the underlying data as Prometheus metrics for your existing Alertmanager.

Part of Valkey monitoring and observability →

Point it at any cluster node - it discovers the rest

docker run -d betterdb/monitor:latest

Then open http://localhost:3001 - see the Docker guide.

Ready to get started?

Start monitoring in minutes - no infrastructure to maintain. Team collaboration, agent-based monitoring for private databases, and more. Or self-host - open source core, zero lock-in.

Valkey Cluster Monitoring: Topology, Slots & SLOT-STATS | BetterDB