Open Stack

On Sun, Mar 15, 2015 at 22:37:59, Sławek Kapłoński wrote:
> Maybe good idea for the beginning could be to implement some periodic
> task
> called from agent to check db config and compare it with real state on
> host?
> What do You think? Or maybe I'm competly wrong with such idea and it
> should be
> done in different way?

This is almost exactly what we do in our Calico ML2 driver. Each of our agents will periodically request its complete state from a neutron-server node and will ensure that its local state matches that expected state. This interval is configurable, to allow administrators to make a trade-off between DB/network load and convergence time.

With reliable transport this is in principle almost never needed (messages only really get lost on agent crash, and the agent will resynchronize when it starts back up anyway), but it provides assurances that the fabric is capable of bringing itself into consistency without administrator intervention.

Having similar function in other neutron agents would be valuable for the exact same reasons, but do bear in mind the potentially increased load this kind of resynchronization can place on databases and servers.

Cory