Open Stack

Hi all,

The new L3 HA and DVR features in the Juno release of Neutron are a big
leap in creating fault tolerant and highly available networking services.
We can run multiple active/passive routers and DHCP servers per tenant on
networking nodes and FWaaS rules are applied to each router. So far, so
good.

But how about:
1 )SNAT, on networking nodes when using DVR
2) VPNaaS
3) LBaaS, on either the networking node or compute node when using DVR

What happens to these services when the host node goes down? Not sure if
SNAT and VPN are covered by HA routers but LB for sure are not run HA yet.

So my question is: how does evacuation/recovery/fail-over work for these
services?

Does Neutron automatically start a new service on an other available node
when it detects that the old host went down? Are tenants simply out of luck
if that happens? Would there be any point to using the L3 agent
"allow_automatic_l3agent_failover" parameter on L3 agents on compute nodes
in a DVR setup?

Hoping somebody with a deeper understanding of Neutron could shed some
light on this. Any insights are very much appreciated!

Thanks,

Tom
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20141128/c1b70da3/attachment.html>