openstack team mailing list archive

Thread
Date

[Quantum] Grizzly: how to setup multi l3-agents for HA

To: "openstack@xxxxxxxxxxxxxxxxxxx (openstack@xxxxxxxxxxxxxxxxxxx)" <openstack@xxxxxxxxxxxxxxxxxxx>
From: Robert van Leeuwen <Robert.vanLeeuwen@xxxxxxxxxxxxx>
Date: Wed, 8 May 2013 11:13:42 +0000
Accept-language: en-US, nl-NL
Thread-index: Ac5Lz5ndLP7msk9OSdSfBFc72G8rFw==
Thread-topic: [Openstack] [Quantum] Grizzly: how to setup multi l3-agents for HA

Hi,

>From the Grizzly release notes I understood it should now be possible to setup multiple l3-agents for HA purposes.

I tried this by just running the l3-agent on two machines but this does not work:
The first one works fine, however when I stop the first l3-agent:
* the ip's are kept on that machine.cat 
* the second node never brings up the router IP's when I start the l3-agent.

I've setup both l3-agents with the id of the the router (using namespaces=False)
Do I need to create multiple routers?
If that is so, I guess I got two gateways in each segment so how is the failure handled from the virtual machines perspective?

Maybe I missing something or is this not even possible?
I'd rather not use Pacemaker for HA because it add's complexity and another thing that can fail.

Thanks,
Robert van Leeuwen