[Azure Service Bus] JMS messages getting dead-lettered
![[Azure Service Bus] JMS messages getting dead-lettered [Azure Service Bus] JMS messages getting dead-lettered](https://cdn.techcommunity.microsoft.com/assets/Azure/BlogPreview_default-blue.png)
The article discusses a problem where numerous messages end up in the dead letter queue (DLQ) when the JMS service bus consumer connects to the Azure Service Bus using Qpid jars. The reason for the messages being dead-lettered is that they have reached the maximum delivery count.
The fundamental issue stems from Apache Qpid's message handling. Qpid utilizes a local buffer to prefetch messages from the Azure Service Bus, storing them prior to delivery to the consumer. The complication occurs when Qpid prefetches an excessive number of messages that the consumer is unable to process within the lock duration. Consequently, the consumer is unable to acknowledge or finalize the processing of these messages before the lock expires, leading to an accumulation of messages in the Dead Letter Queue (DLQ).
To address this problem, it is crucial to either turn off Qpid's local buffer or modify the prefetch count. Disabling prefetching is achievable by setting jms.prefetchPolicy.all=0 in the JMS URL. This configuration allows the JMS client to directly consume messages from the Azure Service Bus, circumventing Qpid's local buffer. Consequently, the consumer can process messages at a suitable pace, guaranteeing smooth processing and issue-free completion.
Why is Prefetch not the default option in Microsoft .NET/Java/Python libs?
Published on:
Learn moreRelated posts
Exploring azd extensions: Enhance your Azure developer experience
A deep dive into the introduction of the Azure Developer CLI (azd) extensions and the azd extension framework to build extensions. The post Ex...
Automate data management with Azure Storage Actions
Introducing the Azure Storage Connector for PyTorch
This post announces the Azure Storage Connector for PyTorch (azstoragetorch), integrating files from Azure Blob Storage into your PyTorch trai...