Open Stack

Heyho guys,

first of all, I know the Pacemaker failover with L3 RA but i think
ther's only working correctly if you're using no namespacing and
overlapping IP's.
In this case (pacemaker+RA) you will have a hot standby Network node but
not an active/active cluster.

ENV:
* Havana release
* Neutron + OVS Gre
* 2 Networknodes
* Seperated Api Nodes
* Storage nodes
* ...


I've tried to make a failover manually by detach all routers they mapped
on the broken l3 agent and attach again to the good one. Is there a way
to do this automatically ? Yes i can write new RA but this script needs
all keystone admin informations and this isn't a good way i think.

Secondly LBaaS-Agent, is there ANY way to make it HA ? So i see you
can't switch the lb pool/VIP to an other agent. The lb provider is
haproxy but i think the main problem is the agent.


Do anyone fine a solution ? I mean for a production environment are this
very critical parts and i don't found any really good solution.


Btw. DHCP has a good working failover.


Cheers and thanks
Heiko