SQL Azure always keeps data in three replicas (a primary and two standby). The first standby copy is done synchronously (ie in the same transaction) and the second asynchronously. Uncommited data may be lost but your commited data is safe.
When a hardware error the first standby is brought online. New connections are made to the new online copy. According to this the failover is not transparent and may result in termination of your connection with a generic connection error (10053 or 40197).
Of course, if it may terminate, you have to assume it will terminate and your code will have to be prepared to retry failed transactions.
发布评论
评论（1）
SQL Azure 始终将数据保存在三个副本中（一个主副本和两个备用副本）。第一个备用副本是同步完成的（即在同一事务中），第二个备用副本是异步完成的。未提交的数据可能会丢失，但已提交的数据是安全的。
当出现硬件错误时，第一个备用数据库将联机。与新的在线副本建立新的连接。根据此 故障转移不透明，可能会导致您的连接因一般连接错误（10053 或 40197）而终止。
当然，如果它可能终止，您必须假设它将会终止，并且您的代码必须准备好重试失败的事务。
SQL Azure always keeps data in three replicas (a primary and two standby). The first standby copy is done synchronously (ie in the same transaction) and the second asynchronously. Uncommited data may be lost but your commited data is safe.
When a hardware error the first standby is brought online. New connections are made to the new online copy. According to this the failover is not transparent and may result in termination of your connection with a generic connection error (10053 or 40197).
Of course, if it may terminate, you have to assume it will terminate and your code will have to be prepared to retry failed transactions.