HP-UX Logical Volume Manager and MirrorDisk/UX Release Notes (5900-2442, August 2012)

Fixed issues in this version
The following table lists the known LVM and MirrorDisk/UX defects fixed in the September 2012
release of HP-UX 11i v3.
Table 1 LVM Fixes in HP-UX 11i v3 September 2012 Release
DescriptionDefect ID
mkfs command will fail with following errors when performed on LVM 2.x VG with LV minor
2047 (7ff hexadecimal).
QXCR1001197843
# mkfs -F hfs /dev/vg01/lvol2047
mkfs (hfs): read of lvm record failed
mkfs (hfs): No such device
# mkfs -F vxfs /dev/vg01/lvol2047
vxfs mkfs: V-3-21203: read of lvm record failed
# ll /dev/vg01/lvol2047
brw-r... 128 0x0007ff /dev/vg01/lvol2047
The last three digits of the device minor is LV minor as highlighted above.
Every time the cluster is stopped (cmhaltcl) and restarted (cmruncl), LVM sub system makes
memory leak of about 0.5625 MB on all the nodes. This memory leak happens irrespective
of whether any LVM Volume group is activated on any of the node.
QXCR1001178600
lvchange -s g on a 1.0 LV fails with following error messages if lvmpvg contains any 2.X VGs:QXCR1001125477
# lvchange -s g /dev/vg01/lvol1
Error: Information in "/etc/lvmtab" or "/etc/lvmtab_p" differs from that in
"/etc/lvmpvg".
VG "/dev/anothervg" is not defined in file "/etc/lvmtab".
VG "/dev/anothervg" is not defined in file "/etc/lvmtab_p".
lvchange: Error detected when reading from file "/etc/lvmpvg".
When pvchange –T is used to handle media corrupted PVs by specifying threshold time, the
physical disk may get detached prematurely even before the threshold is reached. This will
QXCR1001197577
occur only when the first occurrence of media error occurs before the system clock wraps
(497 days since UP time) and the next occurrence is after the wrap.
When this happens, 'Media Error Elapsed Time' field of pvdisplay shows unusually huge
value as below:
# pvdisplay -v /dev/disk/disk1
...
Media error threshold time 2000
Media error elapsed time 18446744073666602006
Running multiple vgimport in parallel causes usage of same minor number for more than one
VG.
QXCR1001204527
In a cluster with 3 or more nodes, when VG recovery is in progress for at least 5 VGs on
each node following a previous server node crash, one or more nodes may get TOC'ed by
SG. On the TOC'ed node, most of the lvmkd daemons will have the following stack trace:
QXCR1001196582
sleep+0x3f0 ()
slvm_send_vgsuspend_resume_req_common+0x330 ()
slvm_send_vgsuspend_cstate_request+0x40 ()
slvm_vgsuspend_recovery+0xb0 ()
slvm_do_remaining_recovery_asynchronously+0xb0 ()
lvmkd_daemon+0xe0 ( )
lvmkd_fork1+0x170 ()
lvmkd_fork+0x50 ()
lvmkd_init+0x50 ()
Performance Collection Component services like midaemon that are part of HP Operations
Agent experience hang when volume groups are in deactivated state.
QXCR1001212048
Export the environment variable LVML_SKIP_INACTVG=1 in the midaemon start up scripts to
avoid this condition with the 1209 version of the liblvml library.
Fixed issues in this version 5