> From 6+ years of experience it seems that we got fouled by the multi-az promise of being able to survive datacenter outage. You have quite a misunderstanding

> From 6+ years of experience it seems that we got fouled by the multi-az promise of being able to survive datacenter outage. You have quite a misunderstanding ... AWS' "multi-az promise" has always been that they will try to take only one AZ down at a time within a region. It was never "blend your AZ usage so we can't take one down." If you don't have a wiki page with some HA architecture diagrams for each of your systems, then you probably don't have HA. Hint: at every company that I've worked at, I drew the first diagrams. Something to think about.