Red Hat Bugzilla – Bug 970846
ucarp - failure of 1 node in a 3 node cluster results in 2 masters
Last modified: 2013-06-05 02:54:11 EDT
Description of problem:
* ucarp isn't honouring "--advskew"
* failure of 1 node in a 3 node cluster results in 2 masters
Version-Release number of selected component (if applicable):
ucarp 1.5.2 - Oct 25 2012
Steps to Reproduce:
Run a 3 node cluster with following parameters:
host 1: ucarp -k 15 -s 192.168.122.191 -v 1 -p badpass -a 192.168.122.200
host 2: ucarp -k 10 -s 192.168.122.13 -v 1 -p badpass -a 192.168.122.200
host 3: ucarp -k 5 -s 192.168.122.38 -v 1 -p badpass -a 192.168.122.200
* Bring up host 2 first, then host 3 and then host 1.
* Kill ucarp on host 2
* host 1 and host 3 both switch to masters
= When run as a service =
within vip-common.conf set the OPTIONS parameter as:
host1: OPTIONS="--shutdown --advskew 15"
host2: OPTIONS="--shutdown --advskew 5"
host3: OPTIONS="--shutdown --advskew 10"
* start ucarp service on host2, host3, host1
* host2 execute: service ucarp stop
Both host1 and host3 become masters
* Bring host2 back online and host1 and host3 return to backup
Online one master to take over VIP after a node failure
Just to expand on this:
The hosts in question are KVM guests. The KVM server is also Fedora 18 (3.9.4-200.fc18.x86_64) It seems that could be a kernel bug as I have the same issues when using keepalived.
My keepalived symptoms are the same as noted here: http://permalink.gmane.org/gmane.linux.keepalived.devel/3815
Given the circumstances I think this bug is a duplicate of: https://bugzilla.redhat.com/show_bug.cgi?id=880035
*** This bug has been marked as a duplicate of bug 880035 ***