Remove wait_apply
stage in disk snapshot restore
#50312
Labels
type/enhancement
The issue or PR belongs to an enhancement.
wait_apply
stage in disk snapshot restore
#50312
Enhancement
The
wait_apply
stage in disk snapshot restore was used for preparing "physical" flashback, which delete all entries committed after the flashback TS directly in each peers underlying storage (RocksDB), without any synchronization by raft.For keeping the consistency of the raft state machine, we must wait all peers' state machine go to the same state.
But for now, the flashback implementation should be the same as
FLASHBACK TABLE
SQL statement, which will "logically" delete the keys needed to be flashed back by committing new writes with raft. Hence it is no need to wait all peers(but only leaders) apply to last index.We can remove the useless step. Given it have brought other bugs, like make backup stuck if there are some stale peers.
The text was updated successfully, but these errors were encountered: