Comment 2 for bug 1821912

From what I was checking, in all those cases (or almost all) instance was able to reach metadata server so connectivity from instance to router was fine.
So, I think that issue is somewhere between "tempest runner"-- router, or maybe in router it self FIP is badly configured but that I'm not sure.

We should try to find some way to do additional debugging when issue happens. Something line dump configuration of interfaces e.g. in router when this issue happens.
Also maybe we could do some DNM patch with adding tcpdump on every qg-XXX and qr-XXX interface in router's namespace to check later where packets are dropped.

That are only some ideas how to maybe start debugging this issue but if someone would have something else than it's of course also great :)