Recovering from storage transaction failures using checkpoints

Abstract

The disclosed technology facilitates recovery from storage-related failures by checkpointing copy-on-write operation sequences. An operation sequence incorporating such checkpoints into a copy-on-write can include the following: receive a write request that identifies payload data to be written to a first data store, read original data associated with the first data store, copy the original data to a second data store, record transactional information associated with the write request, generate a first checkpoint to confirm the successful recordation of the transactional information and the successful copying of the original data to the second data store, write the payload data to the first data store, acknowledge a successful completion of the copy-on-write operation sequence, and generate a second checkpoint that confirms the successful completion of such operation sequence. The first and second checkpoints are used to form a pre-failure representation of one or more storage units (or parts thereof). The checkpoints can be stored with other transactional information, to facilitate recovery in the event of a failure, and can be used to facilitate the use of optimizations to process I/O operations.