Why does RabbitMQ stop and what can be done about it
summary
RabbitMQ often stops because an Erlang process hogs memory or CPU (mailbox never empties), clustering/schema-store issues, or hardware/OS failures; observed failures include convert-to-lazy infinite recursion (queues stuck converting), shovel buffers causing upstream mailbox buildup (fixed in 3.11 via flow control), explosions of Erlang processes (26,000+ pids) and disks creeping to ~120GB. Debugging uses Prometheus, crash dumps, perf/strace, observers and logs and the USE (utilization, saturation, errors) method. Remedies: set all limits, monitor/alert, prefer quorum queues, apply updates (not bleeding edge) and restart or replace affected nodes.
AI-written summary · about
chapters (10)
- 0:0001:56 - Intro & How does CoudAMQP work with Support?
- 1:5703:41 - Most Common Issues
- 3:4204:35 - Chart of Debugging Tools
- 4:3608:04 - Problem
- 8:0509:42 - Why Does RabbitMQ Stops?
- 9:4310:37 - USE Method
- 10:3812:15 - Finding Servers with Issues and System Components
- 12:1619:12 - Examples of RabbitMQ Stopping
- 19:1320:40 - What can be done about it?
- 20:4121:20 - Conclusion and QnA
