LVM messaging

[Joseph Primm]

I would like a clarification on determining what the numbers (as indicated for each) mean.

Example #1
vg[23] - does this refer to the 23rd volume group listed in /etc/fstab?
pvnum=2 - what does the number 2 refer to?
dev_t - this one I do understand

Dec 18 00:07:24 hostname vmunix: LVM: vg[23]: pvnum=2 (dev_t=0x1f240400) is POWERFAILED


Example #2
PV 14 - is this the same as pvnum and/or what does it refer to?
vg[35] - same as example #1 I would guess that it is the 35th entry in the /etc/fstab?

Dec 18 00:26:31 hostname vmunix: LVM: PV 14 has been returned to vg[35].

Lastly - if you had multiple controllers going to multiple arrays. Each with failover capability and pvlinks was set up correctly. What is the expected behavior of LVM in respect to a single path failure. Meaning, how does it keep checking to see if the path is recovered/repaired. Does it simply probe just the one path or does it scan all paths looking for a response?

[Volker Borowski]

Hello Joseph,

vgXXX is the logical name of the volume group, and has usaly nothing to do with /etc/fstab. It is quite common convention, to number your volumegroups vg00 vg01 vg03, as SAM is doing it this way, but you could also use vgDATA, vgSWAP or vgORACLE. In fact you do not need to use 'vg' as a starter, but this would be difficult for any outsider. This name is set up, when the VG is created.

A VG consists of PV (physical volumes), which are in fact disks, logical units from a disk-sub-system or something like that.
A PV is attached to to one ore more controllers, and the access to this PV is given by as many device-files as there are controllers to this PV. When putting a PV into a VG, you have to specify a PRIMARY and ALTERNATE links, which are in fact the diffrent devicefiles to the same disk.
LVM uses usally the primary link to access the disk, but if it fails (controller is dead), it tries the alternate link.

Inside a VG you can create LV (logical volumes) which can be part of one PV in a VG or even spread above several PV in a VG. Logical volumes are used for filesystems and swap, and they appear in /etc/fstab as the device to be linked to a mountpoint (or swap).

So your messages refer to the loss of a PV, which was notified by LVM, and the other one to the detection, that the disk got powered up again.

Hope this brings some clarification
Volker

[Patrick Wallek · Accepted solution]

The vg[xx] and pvnum=x do not represent the order in the fstab, but rather the order in the /etc/lvmtab. Here are a couple of links to documents in the TKB that should help you.

http://us-support2.external.hp.com/cki/bin/doc.pl/sid=127c60871b8733aaec/screen=ckiDisplayDocument?docId=200000048469426

http://us-support2.external.hp.com/cki/bin/doc.pl/sid=28149da80b3c4d32d7/screen=ckiDisplayDocument?docId=200000024645037