Installation guide
Introduction to Mirroring Data
4-4 Fault Tolerant System Administration (R1004H) HP-UX version 11.00.03
■ Mirrored logical volumes should use PVG-strict allocation to allocate physical
extents.
■ If you use single-initiated SCSI buses, make sure that you mirror disks
controlled by a single-initiated SCSI bus with disks controlled by a SCSI bus
attached to a controller port of a PCI card in the other card-cage.
This strategy will ensure that a logical volume can still be accessed in the event of
disk failure or SCSI bus failure.
Guidelines for Managing Mirrors
There are many ways you can set up data mirroring on your system. The Managing
Systems and Workgroups (B2355-90157) describes the guidelines to consider before
setting up or changing mirrored disk configuration.
The following options are presented when you use SAM to configure your mirrors:
■ Bad block relocation—If LVM is unable to store data on a particular block, it
stores the data at the end of the disk.
Always use with Continuum systems when hardware sparing is not available
for disks.
■ Contiguous allocation—Indicates that data is distributed in physical volumes
with no gaps.
Use for root logical volumes, /stand files, and swap space.
■ Number of mirrored copies (0, 1, or 2)—Creates the specified number of
mirrors.
Use 0 for data that rarely changes and is backed up or can be regenerated.
Use 2 when you need to back up the data without interrupting the mirror.
Use 1 for all other cases.
■ Mirror policy (separate physical volume groups, separate disks, or same
disk)—Specifies location of mirrors.
Use separate physical volume groups (also called I/O channel separation)
whenever possible. Physical volume groups should be set up such that
physical volumes are on different SCSI buses. Use separate disks when you
have only two physical volume groups and need two mirrored copies.
■ Scheduling (parallel, sequential, dynamic)—Specifies how mirror is to be
updated.
For higher performance, use parallel to update all copies at the same time.
For higher data integrity, use sequential to update the primary copy first.
For a high-integrity mixture (with better performance than sequential), use