LVM Errors when One Disk is Mapped from one VIOS to Multiple LPARs:
A disk shared to an LPAR from a VIOS can have three mapping configurations:
1. One-to-one: The disk is mapped directly from a VIOS to one LPAR
2. One-to-many by design: A disk is intentionally shared between multiple LPARs in a concurrent environment
3. One-to-many incorrectly: A disk is inadvertently shared from a VIOS to multiple LPARs
When a disk is inadvertently mapped to multiple LPARs, LVM inconsistencies can occur. Below is an example of incorrectly mapping one disk to multiple LPARs and the resulting errors.
1. Map the disk from VIOS --> vhost0
$ mkvdev -vdev hdisk15 -vadapter vhost0
$ lsmap -vadapter vhost0
SVSA Physloc Client Partition ID
--------------- -------------------------------------------- ------------------
vhost0 U9117.MMA.1010A00-V2-C12 0x0000000a
VTD vtscsi12
Status Available
LUN 0x8200000000000000
Backing device hdisk15
Physloc U789D.001.DQDYKHP-P1-C1-T1-W500507680B21683C-L0
Mirrored false
2. Map the disk from VIOS --> vhost1
$ mkvdev -f -vdev hdisk15 -vadapter vhost1
$ lsmap -vadapter vhost1
SVSA Physloc Client Partition ID
--------------- -------------------------------------------- ------------------
vhost1 U9117.MMA.1010A00-V2-C13 0x0000000b
VTD vtscsi13
Status Available
LUN 0x8200000000000000
Backing device hdisk15
Physloc U789D.001.DQDYKHP-P1-C1-T1-W500507680B21683C-L0
Mirrored false
Now the disk is shared between LPAR1 and LPAR2 and the PVIDs are 00c10a003757f29e.
LPAR1 (vhost0):
(lpar1-vhost0)# lspv
hdisk0 00c10a00e7970f11 rootvg active
hdisk1 00c10a003757f29e None
LPAR2 (vhost1):
(lpar2-vhost1)# lspv
hdisk0 00c10a000bf5ee59 rootvg active
hdisk1 00c10a003757f29e None
3. Create a VG on hdisk1 on LPAR1 (vhost0)
(lpar1-vhost0)# mkvg hdisk1
(lpar1-vhost0)# lspv
hdisk0 00c10a00e7970f11 rootvg active
hdisk1 00c10a003757f29e vg00 active
4. Change the PVID for hdisk1 on LPAR2 (vhost1)
(lpar2-vhost1)# chdev -l hdisk1 -a pv=clear
(lpar2-vhost1)# chdev -l hdisk1 -a pv=yes
(lpar2-vhost1)# lspv
hdisk0 00c10a000bf5ee59 rootvg active
hdisk1 00c10a0037b59fe5 None
5. Varyonvg error
At this point, LVM information on LPAR1 assumes the disk PVID is 00c10a003757f29e, but LPAR2 has changed the PVID to 00c10a0037b59fe5.
When we try to varyoff then varyon the volume group on LPAR1, it can not find 00c10a003757f29e.
(lpar1-vhost0)# varyoffvg vg00
(lpar1-vhost0)# varyonvg vg00
0516-013 varyonvg: The volume group cannot be varied on because
there are no good copies of the descriptor area.
6. Importvg error
From LPAR2, if we try to import the volume group we received the follow error.
(lpar2-vhost1)# lspv
hdisk0 00c10a000bf5ee59 rootvg active
hdisk1 00c10a0037b59fe5 None
(lpar2-vhost1)# importvg hdisk1
0516-1939 : PV identifier not found in VGDA.
0516-780 importvg: Unable to import volume group from hdisk1.
7. Root cause for varyonvg and importvg errors
Since the LPARs share the disk but have their own ODM entries, we can see the discrepancy.
PVID
ODM On Disk
lpar1-vhost0 00c10a003757f29e 00C10A00 37B59FE5
lpar2-vhost1 00c10a0037b59fe5 00C10A00 37B59FE5
The PVID in the ODM must match the PVID in the LVM metadata. If changes are made to a disk mapped to multiple LPARs, the PVIDs in the ODM may not match the PVIDs in the LVM metadata.
If you experience the errors cited, or similar errors referencing the VGDA, please contact IBM support to correct the LVM inconsistencies. Incorrect attempts at repairing LVM metadata may result in data loss or system instability.
No comments:
Post a Comment