强制故障转移（FORCE_FAILOVER_ALLOW_DATA_LOSS）及故障转移后的恢复

命令

-- 1) On each candidate secondary: pick the most current one (highest end_of_log_lsn)
SELECT ag.name AS ag_name, DB_NAME(drs.database_id) AS db,
       drs.synchronization_state_desc, drs.end_of_log_lsn, drs.last_commit_time
FROM   sys.dm_hadr_database_replica_states drs
  JOIN sys.availability_groups ag ON ag.group_id = drs.group_id
WHERE  drs.is_local = 1;

-- 2) On the chosen replica (while the primary is unreachable):
ALTER AVAILABILITY GROUP [AG_Musteri] FORCE_FAILOVER_ALLOW_DATA_LOSS;

-- 3) On the other secondaries, resume the suspended data movement:
ALTER DATABASE [Musteri] SET HADR RESUME;

-- 4) When the old primary returns (now a SUSPENDED secondary):
--    take a snapshot first to inspect the lost transactions, then resume
CREATE DATABASE [Musteri_lost_snap]
  ON (NAME = N'Musteri', FILENAME = N'D:\Data\Musteri_lost_snap.ss')
  AS SNAPSHOT OF [Musteri];
ALTER DATABASE [Musteri] SET HADR RESUME;

作用

主副本完全丢失时，接受数据丢失将 DR 副本提升为主副本；随后恢复其他副本上挂起的数据移动，并在旧主副本恢复后创建快照，以检查未到达新主副本的事务。

适用场景

主数据中心宕机、只剩异步副本，或同步副本不是 SYNCHRONIZED 状态但仍必须故障转移时。

注意

数据丢失：最后一个已固化 LSN 之后的事务会丢失。先获得业务方对 RPO 的认可。强制故障转移后所有辅助数据库都处于 SUSPENDED 状态；在 HADR RESUME 之前日志不会截断并持续增长。若 WSFC 仲裁丢失，先以强制仲裁启动群集（PowerShell：Start-ClusterNode -FixQuorum），否则命令会失败。在旧主副本上执行 HADR RESUME 会永久丢弃丢失的事务；未创建快照前不要恢复。