Operating System - HP-UX
1834089 Members
2227 Online
110063 Solutions
New Discussion

Re: unable to determine unique identifier for PV

 
vijay alur alur
Frequent Advisor

unable to determine unique identifier for PV

Hi All,

 

We have a 2 node cluster and we has storage migration from EMC to Hitachi VSP. After the migration was completed when we tried to cmchheckconf, we got the error

 

"unable to determine unique identifier for PV--- pv create to give disk identifier"

Anyone has got any idea about this error?? Please help....

 

 

 

Lead Engineer, IMS.
iGATE
1 REPLY 1
Matti_Kurkela
Honored Contributor

Re: unable to determine unique identifier for PV

This might be harmless: it means you have a disk device on your system that is not controlled by LVM. Some versions of Serviceguard might produce this message on CD-ROM drives. Alternatively, if you have presented a disk to the system and haven't (yet?) run pvcreate on it,  this will cause cmcheckconf to display this message too. Some storage systems may even have pseudo-disk devices that are used to send management commands to the storage system: in EMC storage systems, these are called "gatekeeper devices": the OS will identify them as very small disk devices, but you cannot actually use them to store data, so you cannot even use pvcreate on them. If Hitachi uses the same technique, they might have a different name for them.

 

Cmcheckconf is looking for shared disks: the standard situation in a Serviceguard cluster is to have all the shared disks using LVM, because Serviceguard uses LVM to control access to shared disks. Shared disks may also be used to contain the cluster lock, which is a very important part of a 2-node cluster.  When a disk has no LVM PV identifier (i.e. "pvcreate" has not been run on the disk, or the disk is in a non-LVM format), cmcheckconf cannot reliably determine if that disk is usable as a shared disk or not.

 

Technically you can have non-LVM disks in your Serviceguard cluster nodes if you have a good reason to do so, but they will be entirely your responsibility as a sysadmin: Serviceguard cannot enforce their proper use so you must understand the requirements of the disk format used, and make sure they aren't violated. Usually this means "do not start the application that uses the non-LVM disk unless you are absolutely sure it isn't already running on some other node, and make sure nobody else can do it by mistake". If the non-LVM disk is not shared, Serviceguard places no special restrictions on it.

 

The message is essentially a notification. Serviceguard is telling you "I hope you know what this disk is used for, because my normal way of finding out is not applicable to it." It should mention a hardware path, a disk device name or some other method to identify the disk it's talking about. The suggestion to use pvcreate is because the creator of cmcheckconf assumes the most common reason for having non-LVM disks in Serviceguard cluster nodes is because you've run the cmcheckconf command while you're in the middle of adding new LVM disks to the system.

 

Newer versions of Serviceguard (at least A.11.19 and newer, not sure about A.11.18) allow excluding some disks from probing, by adding the device names or type strings to /etc/cmcluster/cmnotdisk.conf or /etc/cmcluster/cmignoretypes.conf, respectively. If you have a CD-ROM drive that is not recognized as such by cmcheckconf or a disk device that is supposed to be a non-LVM disk, you might want to use these to disable the notification if you know it is not necessary.

 

MK

MK