Direct network connection to instance or external load balancing?

Recently during OVH outage, what I observed was that the instances itself were running, but the public-internal routing was causing the issue due to the reasons you mentioned in your post mortem post.

What measures we can take as failover, to reach our apps directly during such time? Can we set up an A record, which will directly point to the instance, or can we set up a load balancer outside fly network?

You can’t right now. We’re buying a bunch of IPs so people can directly access VMs, we’ll hopefully have a way to do this in several months.