EMQX MQTT Broker Dashboard for Connections and Messages

Use this dashboard to monitor your EMQX brokers. It covers cluster and connection health, message throughput, message loss, subscriptions and routing, access control, and Erlang VM runtime.

EMQX exports these metrics from inside the broker over OTLP. The names are dotted and carry no emqx prefix, which separates them from the emqx_* names on EMQX's Prometheus scrape endpoint.

Dashboard Coverage

Use these panels to:

• Confirm cluster health: See cluster size, stopped nodes, and current connections and sessions per node.
• Track connection churn: Watch connect attempts against successful connects and disconnects, and follow session creation, resumption, takeover, and termination.
• Measure throughput: Track message and packet rates in and out, bytes on the wire, and the QoS mix on both the receive and send paths.
• Find message loss: Separate messages the broker never routed from deliveries it failed to complete, and read the drop reason.
• Inspect subscriptions and routing: Track topics, subscriptions, routes, and retained messages, plus subscribe and unsubscribe rates.
• Review access control: Watch authentication success and failure, authorization allow and deny, cache hit rates, and banned or flapping clients.
• Watch broker runtime: Track CPU use, the Erlang run queue, and memory used against total per node.

Metrics Included

Cluster & Connections

• Cluster Size and Nodes Stopped: Nodes in the cluster as each node reports it (node.running, node.stopped). A standalone broker reports 1.
• Connections and Sessions: Current client connections and sessions (connections.count, sessions.count).
• Connections by Node: Total and live connections per node (connections.count, live_connections.count).
• Sessions and Channels by Node: Sessions, cluster sessions, and channels per node.
• Connection Churn: Connect attempts, successful connects, and disconnects (client.connect, client.connected, client.disconnected).
• Session Lifecycle: Sessions created, resumed, taken over, discarded, and terminated.
• Connections: Current vs Peak: Current values against the high-water marks (connections.max, sessions.max).

Message Throughput

• Message Rate In / Out: Messages received and sent (messages.received, messages.sent).
• Delivery and Acknowledgement Rate: Messages delivered to subscribers and acknowledged.
• Bytes In / Out and MQTT Packets In / Out: Wire volume and packet counts.
• Received / Sent Messages by QoS: QoS 0, 1, and 2 split on both paths.
• Publish, Forward and Store: Published, retained, forwarded, persisted, and delayed messages.
• Rule Engine: Rules matched and actions executed.

Message Loss

• Messages Dropped and Deliveries Dropped: Messages the broker never routed, and deliveries it failed to complete.
• Message Drops by Reason: No subscribers, expired, quota exceeded, and receive maximum.
• Delivery Drops by Reason: Queue full, too large, expired, no local, filtered, and QoS 0 message.
• Packet Errors: CONNACK and PUBLISH errors, including authorization errors.
• Missed QoS Acknowledgements: Missed PUBACK, PUBREC, PUBREL, and PUBCOMP packets.

Subscriptions & Routing

• Topics, Subscriptions, Routes, and Retained Messages: Current table sizes.
• Subscriptions and Subscribers: Subscriptions, subscribers, subscription options, shared and durable subscriptions.
• Subscribe / Unsubscribe Rate: Client subscribe and unsubscribe rates.
• Subscribe Packets and Errors: SUBSCRIBE, SUBACK, UNSUBSCRIBE, and UNSUBACK packets, plus subscribe errors.

Authentication & Authorization

• Authentication Outcomes: Success, failure, and anonymous success (authentication.success, authentication.failure).
• Authorization Outcomes: Allow, deny, superuser, and no match.
• Authorization Cache: Cache hits and misses.
• Banned and Flapping Clients: Banned clients and flapping detected by client ID, peer host, and username.
• CONNACK Outcomes: Authenticate and authorize counts against CONNACK errors.

Node Runtime

• CPU Use by Node: Broker CPU use (cpu.use).
• Erlang Run Queue by Node: Scheduler run queue length (run.queue).
• Memory Used vs Total by Node and Memory Utilization by Node: Memory consumption and percentage.
• Process Mailbox Sizes by Node: Broker pool and Mnesia transaction manager mailbox sizes.

Dashboard Variables

Use these filter variables:

• service.name: Filter by broker service name. EMQX reports emqx unless you set OTEL_SERVICE_NAME.
• service.instance.id: Filter by EMQX node name, such as emqx@10.0.1.7.

Next Steps

• Send the data this dashboard reads with the Monitor EMQX with OpenTelemetry guide.
• Explore individual broker metrics in the Metrics Explorer.
• Create metric-based alerts for dropped messages, connection drops, or node memory.
• Create log-based alerts for authentication failures and broker errors.

Related Dashboards

• RabbitMQ
• Kafka
• Hostmetrics