Description of Problem:
It sometimes happens that a fail-over interface is not removed when pulse is
stopped ("ifconfig ethN:X down" not run). This can create BIG problems when
running with both a master and a backup director (resulting in having the VIP
on both directors).
Version-Release number of selected component (if applicable):
In my setup it was quite often, but it could be just 1 time out of 100 :(
Steps to Reproduce:
1. pulse -v -c /etc/sysconfig/ha/lvs.cf
2. wait a few seconds
3. service pulse stop
4. ifconfig might show the VIP whereas it should not
VIP configured on director whereas service not running, or worse, running on
VIP removed when pulse stoppped.
The problem *seems* to be in pulse.c, in function deactivateFos. Swapping the
HA_process termination and the VIP cleanup *seems* to be a fix. But not a real
Piranha's FOS functionality (though not piranha itself) is slated for removal.
There are much better two node failover solutions out there, notably Red Hat