RabbitMQ running out of memory

user_10035746 0
Viewed 66

I have a cluster of three RabbitMQ nodes and I would like to understand why my cluster is running out of memory after few days. When I start the cluster, the used memory is about 400MB but every day the memory is increasing even if the traffic and the number of queues are pretty much the same.

This is a screenshot 2 days after restart: enter image description here

Here are details: enter image description here

Can you explain me what are the quorum queue tables and why they are growing its size every day ??

After several days the rabbitMQ hit the high watermark so it stops to accept connections. I tried already to increase the watermark to 1.3GB but the only difference was that the problem occurred after more days.

This is memory details when the node hits the watermark: enter image description here

Thank you for help

rabbitmq
1 Answers

When you hit your water mark:

  • The quorum queues did not grow significantly over time.
  • The quorum queue tables are a lot smaller

The quorum queues not your problem.

However when you hit the watermark you have a lot of Preallocated Memory.

From Reasoning About Memory Use - Preallocated Memory

When a node reports a large amount of allocated but unused memory on a long running node, it may be an indicator of runtime memory fragmentation.

You need to change you settings to something that matches your situation:

A different set of allocator settings can reduce fragmentation and increase the percentage of efficiently used memory. The right set of settings depends on the workload and message payload size distribution.

user_349890
Related