What you're describing is a single writer distributed system, which is one way to achieve consistency. Since you can't write directly to server B, the system i

What you're describing is a single writer distributed system, which is one way to achieve consistency. Since you can't write directly to server B, the system is unavailable for writes when a writer happens to live on the side of the network partition where B is located. Often when discussing "availability", it's very important to qualify for whom and for which operations.