Comment 5 for bug 1402010

I originally thought to solve this by somehow serializing the keepalived notifier scripts, but after spending an enormous amount of time trying I determined that that's impossible. Instead, I intend to abandon the keepalived notifier scripts altogether and instead monitor the IP addresses in the router namespaces. Since keepalived configures IP addresses only on the master instance, that's a good way of knowing the state of an HA router. This approach is technically backportable (No DB upgrades, no RPC or API changes, etc) but it might just be too large / risky. We can determine that once the complete solution is up for review.