In Zerto, when a VM undergoes a recovery operation (such as failover, failover test, or move) that requires a pre-commit decision point, all I/O writes generated by the recovered VM during that pre-commit period are redirected to a temporary scratch volume instead of being written directly to the final recovered virtual disks. This separation allows Zerto to easily roll back the operation—if a rollback is triggered, the scratch volume is simply discarded, and the recovered VM is removed/reverted as though the recovery never happened, avoiding any need to undo writes on the actual recovery disks. If the recovery is committed instead, the scratch data is merged into the recovered VM's disks.
Community Discussion