HP-UX Logical Volume Manager and MirrorDisk/UX Release Notes HP-UX 11i v3 (October 2010)

Corrective Action: Do one of the following:
• Use the lvchange command to change the permission of the
space efficient snapshot to read-only to prevent the snapshot
from becoming over-committed and inoperative due to any
write on it.
• To prevent the snapshot from becoming over-committed and
inoperative due to any write to its successor, add physical
volume(s) to create necessary free extents in the volume group
and deactivate and activate the volume group to allocate the
pre-allocated extents.
The lvextend command fails to increase the number of pre-allocated extents
Defect ID: QXCR1001077247
Problem: When the number of pre-allocated extents of a space efficient
snapshot is zero; in other words, 'Current pre-allocated
LE' displayed by the lvdisplay command is zero, an attempt to
increase the number of pre-allocated extents using the lvextend
command fails even though there are enough free extents in the
volume group to satisfy the allocation policy of the snapshot.
Severity: Serious
Corrective Action: While some extents of a space-efficient snapshot still share data
with its successor, make sure that the lvmpud daemon is always
•
running and enable auto pre-allocation for the space-efficient
snapshot so that the number of pre-allocated extents does not
become zero.
• Deactivate and activate the volume group if the lvextend
command fails to increase the number of pre-allocated extents
in the case mentioned above.
Limitations
This section provides a list of limitations as known to HP at time of publication.
• Logical volumes that have snapshots associated with them might experience an increase in
latencies associated with writes. This is also applicable for writes on (writable) snapshots
themselves.
• With the HP-UX 11i v3 September 2010 Update release, when the number of extents
remaining in a space-efficient snapshot logical volume’s pre-allocated pool falls beyond a
certain internally computed threshold, by default, LVM tries to increase the pre-allocated
pool size by threshold value. This does not guarantee that snapshot will not become
over-commit. For example, if the I/O rate is faster than auto pre-allocation, the snapshot
becomes over-commit.
If auto pre-allocation is manually disabled, a message is logged in the syslog. This message
is displayed to inform you that further unsharing of data between the snapshot and its
successors might end up depleting the extents in the pre-allocated pool and lead to the
snapshot and its predecessors being marked as inoperative. As soon as you see this message
in the syslog, you must increase the pre-allocated extent pool size for the snapshot logical
volume.
• For more limitations on snapshots, see the Using LVM Logical Volume Snapshots white paper.
Installation requirements
This section describes the installation requirements for this release.
8