Restrictions On Database Validator - Hitachi TAGMASTORE MK-90RD011-25 User Manual

2.7.2

Restrictions on Database Validator

Oracle Tablespace Location
File system-based Oracle files are not supported by DB Validator. All Oracle
–
tablespace files must be placed on raw volumes (including LVM raw volumes)
directly.
If host-based striping is used on the raw volumes, then the stripe size must be an
–
exact multiple of the Oracle block size.
Oracle redo log files (including archive logs) must be on separate volumes with
–
respect to the data files (including control files) to different LU. In other words,
Oracle redo log files and the data files must not be mixed on the same LU.
Restoring of Oracle Files
Before restoring Oracle data files from a backup, the data validation may need to be
–
temporarily turned off for those data files that were backed up prior to the Oracle
checksum being enabled.
Old blocks may exist on disk without checksum information in them if the database
was running without checksum enabled in the past.
Oracle on LVM (VxVM)
LVM block size must be a multiple of the Oracle block size. The Oracle block size
–
must be less than or equal to the minimum of the LVM stripe size and the largest
block size that LVM will not fracture (known as "Logical Track Group" in LVM), which
is 256 KB in LVM.
When adding new physical volumes (PVs) to a logical volume (LV) to be used as an
–
Oracle datafile, controlfile, or online log, in order to have HARD checking take
effect on those new PVs, the data validation should be re-enabled.
Similarly, in order to have HARD checking no longer performed on PVs that have
been removed from an LV that had previously been used by Oracle, HARD checking
should be explicitly disabled on the device corresponding to the PV.
If host-based mirroring is used such as LVM mirroring, all component PV mirrors must
–
be HARD-enabled, otherwise the entire logical volume (LV) is exposed. That is, if a
user takes an unmirrored HARD-enabled LV, then makes it mirrored on the fly
without HARD-enabling all sides of the mirror, that entire LV is exposed to data
corruption.
LVM bad block relocation will not be allowed on PVs that are HARD-enabled.
–
Oracle and LVM (VxVM) on HA Cluster Server
If HA Cluster software will be writing to LVM metadata at regular intervals in order
–
to confirm whether its disks are available or not, then its LVM's area must be out of
checking for Database Validator by using "-vs <bsize> SLBA ELBA" option.
Hitachi Command Control Interface (CCI) User and Reference Guide 41