RabbitMQ Per Object

RabbitMQ Per Object

Per-queue and per-exchange debugging dashboard for RabbitMQ, powered by the /metrics/per-object endpoint of the rabbitmq_prometheus plugin.

Built for on-call engineers and service owners who need to answer questions like "is this queue backed up?", "is this consumer keeping up?", or "is the message I published actually reaching its queue?". It is not a broker-health overview — pair it with the stock RabbitMQ-Overview dashboard for that.

Source: github.com/SonyCore/rabbitmq-per-object-dashboard — file issues and feature requests there.

Prerequisites

• RabbitMQ with the rabbitmq_prometheus plugin enabled:

  rabbitmq-plugins enable rabbitmq_prometheus
  

  The per-object endpoint is exposed on port 15692 at /metrics/per-object. See the RabbitMQ Prometheus docs.

• Prometheus scraping that endpoint under job name rabbitmq-per-object with an env label:

  scrape_configs:
    - job_name: rabbitmq-per-object
      metrics_path: /metrics/per-object
      scrape_interval: 30s
      static_configs:
        - targets: ['staging.local:15692']
          labels: { env: staging }
        - targets: ['production.local:15692']
          labels: { env: production }
  

  Every query in this dashboard is pinned to job="rabbitmq-per-object". If you use a different job name, search-and-replace it in the JSON after import.

• Grafana 12.3+ with a Prometheus datasource.

How to use it

1. Pick the env (e.g. staging, production).
2. Narrow down the vhost (defaults to all).
3. Set queue and/or exchange to the object(s) you're investigating.
4. Walk top-to-bottom — rows are ordered to match a typical debug flow.

Template variables

Panels

QUEUE — state Queue depth (ready + unacked) · Messages ready · Messages unacknowledged · Consumers.

QUEUE — health Backlog per consumer (messages_ready / consumers) as a rough drain-timer signal.

QUEUE — rates Publish rate into the queue · Delivered & acknowledged rate · Redelivery rate.

EXCHANGE Publish rate · Confirm rate · Unroutable (dropped & returned) — any sustained non-zero value on the last one means message loss or a routing/binding bug.

EXCHANGE → QUEUE ROUTING Per-binding routing rate, grouped by (exchange, queue). Verifies that a message published to exchange X is actually landing in queue Y.

TOP QUEUES / TOP EXCHANGES Fleet-wide top 20 queues by depth and top 20 exchanges by publish rate. Ignores the queue/exchange filters so you always see the biggest offenders.

Metrics required

All from the RabbitMQ per-object endpoint:

• rabbitmq_queue_messages
• rabbitmq_queue_messages_ready
• rabbitmq_queue_messages_unacked
• rabbitmq_queue_consumers
• rabbitmq_queue_messages_published_total
• rabbitmq_queue_messages_delivered_total
• rabbitmq_queue_messages_acknowledged_total
• rabbitmq_queue_messages_redelivered_total
• rabbitmq_channel_messages_published_total
• rabbitmq_channel_messages_confirmed_total
• rabbitmq_channel_messages_unroutable_dropped_total
• rabbitmq_channel_messages_unroutable_returned_total

License

MIT.