https://bugzilla.redhat.com/show_bug.cgi?id=1309684
Marian Krcmarik <mkrcmari(a)redhat.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |mkrcmari(a)redhat.com
--- Comment #2 from Marian Krcmarik <mkrcmari(a)redhat.com> ---
I have one example of justification for rabbitmq rebase to version 3.6:
If several failovers take place in Openstack cluster with three controller
nodes hosting rabbitmq-server, queue masters tend to be hosted on a single
controller node and with failover of the node they just move to a different
available node. Once a failover of such node happens recovery time of Openstack
cluster is increased significantly. The problem nicely described here -
https://www.erlang-solutions.com/blog/take-control-of-your-rabbitmq-queue...
Several queue master distribution strategies were introduced in upstream
version of 3.6.0 which could solve the problem -
https://github.com/rabbitmq/rabbitmq-server/pull/217 for example by using
min-masters strategy.
I would be happy to test a scratch build to see If it really helps.
--
You are receiving this mail because:
You are on the CC list for the bug.