openstack team mailing list archive

["Hathaway.Jon" <Jon.Hathaway@xxxxxxx>]

All

I was wondering whether my current understanding of HA in Quantum is currently correct or not as there does seem to be a little confusion.

>From what I understand, the Quantum Service, DHCP and L3 agents are all supported through Pacemaker RA's or using the native messaging in Openstack Grizzly. Using the Grizzly Quantum messaging heartbeats, the Quantum service will keep deploying configurations until a point in time the agents are not responding. When this occurs it will look to deploy configurations on other available Network nodes running the agents.

However, the bit I am really fuzzy about it as follows. What happens if the network node completely fails? How would I recover the current networking configuration from the failed network node and re-deploy it on a working network node automatically?

I also see from reading multiple sources, that it would be possible to deploy the same configuration to multiple network nodes at the same time using the Quantum API, but it isn't supported currently through the normal GUI API calls so it would be an enhancement or a separate orchestration step.

If possible, would someone be able to point me in the right direction of how I can provide full HA Quantum based services for not only new configurations, but how I would re-deploy and recover configurations from a failed network node. I'm looking at this from an SLA perspective as well - is it possible to achieve failover between two active network nodes?

Thanks in advance to the great community out there.

Jon