Using Veritas Volume Manager will make your data accessible, fast, and reliable. But as we said before there is nothing that cannot be improved. Consider the case where a mirrored volume is open while a system crash occurs. Because the plexes may not have been updated at identical times (this is optimistic - it would be a great coincidence if they actually had been updated at the same time) they probably contain different data in at least some volume regions. This would break the definition of volume management behavior: that a volume shall behave exactly the same as a partition. While a partition will always deliver the same data for the same blocks unless you change the blocks, a mirror with non-synchronous contents may not. This could cause the file system or database to crash, resulting in system panic or database corruption.
As another example of optimization potential consider a plex in a mirrored volume that is disabled because of a trivial, transient error, such as a disk that was switched off or otherwise unreachable when the volume was started. Most of the contents of the disk are probably identical to the contents of the other plex(es), but after the disk is back online VxVM will nevertheless have to resynchronize the whole volume because it simply did not keep track of which regions is changed in the remaining active plexes.
KeywordsUser Data File System Access Mode Parity Data Transient Error
Unable to display preview. Download preview PDF.