pvmove issue

[harsr]

Hi,

We have issue with pvmove, while pvmove happening from one PV to other destination PV, the source PV got detached. After the source PV came back to online, we tried the same PV move operation on to different destination. Right now one PE is in stale state and we are getting I/O error at Oracle database level while accesing the particular table space. Can any one help on this and let us know how to recover the stale PE.

00286 /dev/dsk/c11t7d5 01046 stale /dev/dsk/c11t8d1 00882 stale PX_NOPV 00000 current

[John Payne_2]

Well, that certainly is not a good sign. It's kind of hard to recover a stale PE without having one that's a current PE. You might get lucky if you were to power the disk down and spin it back up, but I think you've got all the extents that follow this one available on the old disk, so I think the damage is already done.

Anyway you can recover that tablespace at this point if you created a new volume?

Were you moving from one disk to the other because you knew a failure was coming, or were you trying to go from one disk to another for some other reason?

John

[harsr]

Hi Jhon,

Unfortunately we don't have backup for this. The intention of pvmove is storage migration. We are moving the data from one EMC DMX3000 storage to other EMC DMX3000 storage. Please suggest us if there are any other solutions.

[Emil Velez]

pvmove will not work if the source disk is not available and stale. This is not a good situation since it seems that both disks with the extents are not available. You need to get one disk back online and them pvmove the lv. WIthout all of the extents a LV is not available.

You could try a vgchange -a y vgxx to see if it will recognize that the disk is back online. YOu need to get one of the disks online as far as LVM is concerned. IT is bad that it seems that both disks failed at the same time.

[SKSingh_1]

Hi,

could please share command executed for pvmove??

and output of affected LV.
lvdisplay
lvdisplay -v | grep -i stale