how two-phase commits hinder last-second failure?
i am study two-phase dedicate works opposing distributed transaction. bargain final biased apportionment transaction coordinator asks any node presumably prepared commit. everybody agreed, following tells disrespectful commit.
what prevents following failure?
- all nodes respond are
ready commit - the transaction
coordinator tells "go ahead
and commit" nonetheless nodes
crashes before receiving this
message - all nodes dedicate successfully, nonetheless distributed transaction corrupt
- it bargain crashed node comes back, the transaction have rolled behind (since never got dedicate message)
i am presumption any node controlling normal database doesn't know anything distributed transactions. i miss?
Comments
Post a Comment