The recovery takes place when the database software is started, not at system startup. This reduces the overall impact of recovery when the system reboots. Because the recovery is controlled by the database, the recovery time for the volume is the resilvering time for the database (that is, the time required to replay the redo logs).
Because the database keeps its own logs, it is not necessary for VxVM to do logging. Data volumes should be configured as mirrored volumes without dirty region logs. In addition to improving recovery time, this avoids any run-time I/O overhead due to DRL, and improves normal database write access.