CloudAMQP✓

Good morning, Haarlem! Let’s kick off @MQSummit with a big smile! 😄 Please swing by and say hi.

We’ve managed RabbitMQ clusters for over a decade and have learned a few things along the way. What do you wish you’d known before putting RabbitMQ in production?

What caused your worst queue incident? Reply with your story. (We’ve seen a few. 😉)

Things go wrong in production. The question is what your app does next. Part 2 of our Node.js series, we make a banking app on @LavinMQ handle failures. Article in comments! * Retry temporary errors with a delay queue. * Dead-letter the ones that won’t fix themselves. * Skip duplicate messages so no one gets charged twice.

Need a RabbitMQ message to arrive in 10 minutes, not now? The Delayed Message Exchange plugin holds it, then routes it. Lois Soto López covers how it works, its limits, and the alternatives. #RabbitMQ

Message queues are a natural fit for AI workloads. A worker takes a request from a queue, asks an AI agent a question about it, and publishes the answer to a response queue. The catch is that agents can take minutes to reason, which is often longer than the broker’s consumer timeout. Magnus Landerblom walks through three ways to handle long-running tasks without processing the same work twice.

Need a broker for Celery? Our new tutorial walks through setting up LavinMQ as the broker for your Celery workers, from spinning up an instance to running your first async task. Fast to set up, light on resources, and speaks AMQP natively, so switching is usually a one-line config change.

RT @KuhikarPranjal: @CloudAMQP Really happy to see my NeuroBank project featured by CloudAMQP! Building it was a great learning experience with RabbitMQ, CloudAMQP, and event-driven communication. Thanks for featuring the project! Project: GitHub:

Things go wrong in production. The question is what your app does next. Part 2 of our Node.js series, we make a banking app on @LavinMQ handle failures: * Retry temporary errors with a delay queue. * Dead-letter the ones that won’t fix themselves. * Skip duplicate messages so no one gets charged twice.

Pranjal was new to message queues and couldn't find beginner-friendly content. Hailay heard his call and wrote it. His four-part series builds a banking app with LavinMQ and Node.js, from the first message to retries, dead-letter queues, and reconnects. Part 1:

Running your own metrics stack? You can now push RabbitMQ and LavinMQ metrics straight to any Prometheus remote write endpoint, including self-hosted setups like Mimir, Thanos, or VictoriaMetrics. Native Prometheus format, four auth options, automatic cluster labeling.

Send your LavinMQ and RabbitMQ logs to any OpenTelemetry backend. CloudAMQP now ships broker logs as structured OTLP records to SigNoz, Honeycomb, your own Collector, or any other OTLP/HTTP endpoint. Severity is already mapped, so you can skip the parsing rules. Set it up in minutes:

How does RabbitMQ actually perform against Azure Service Bus under load? We ran the benchmarks. Our new blog covers throughput, message size impact, consumer concurrency, and latency, with RabbitMQ pulling ahead on raw throughput while trade-offs remain on both sides.

RabbitMQ or Azure Service Bus? It depends on what you're optimizing for. Our new blog compares message flow, delivery models, dead-lettering, and portability between the two, and breaks down when fine-grained routing control wins over fully managed simplicity. Read more:

OTLP log integration is now available for RabbitMQ and LavinMQ dedicated instances, for any OpenTelemetry-compatible backend, such as SigNoz or Honeycomb.

CloudAMQP is now available in the Amazon Web Services Asia Pacific (Taipei) region.

New region added: CloudAMQP is now available in Malaysia West.

If you run Uptrace for observability, your LavinMQ and RabbitMQ broker logs can now join it directly. Both were built around OpenTelemetry from the start, so logs arrive clean and queryable, with no extra tooling required.

The CloudAMQP Datadog log integration is rebuilt on OpenTelemetry OTLP. Logs now arrive as a typed number rather than embedded text, the source is auto-tagged as "cloudamqp," and multi-node clusters get one log stream per host. Read more: