Comment 26 for bug 1371723

Igor, version of document that you mentioned contains this text about the bug:

============
OpenStack services report multiple AMQP reconnects in the logs
nova-api and other services may call fork() soon after the process starts. This may lead to a situation when the
AMQP heartbeat sending thread is initialized and run in the context of the parent process, while the reply waiters
are added to the list in the child process. When this happens, the thread won't send any heartbeats and RabbitMQ
will close AMQP connections after timeout.
This problem is addressed in oslo.messaging packages from MOS 5.1.1 package repositories (CentOS, Ubuntu).
After installing these packages, restart the affected services. See LP1371723.
============

But current version of docs, updated for 5.1.1 (docs.mirantis.com/openstack/fuel/fuel-5.1/release-notes.html#openstack-bugs-are-fixed) has only this:
============
The Oslo messaging service had a bug that affected MOS HA environments. The bug could make OpenStack unstable; the problem got worse as the load increased. A large number of broken AMQP connections caused Oslo-related errors appear in different logs. See LP1371723.
============

So, instructions for those, who still use 5.1.0 is gone...