LVM wedged after stopping some pvmove's

I was in the middle of doing pvmove for all my lv's which resided on two disks, that were being moved to a newly created raid5 pv.

Things got all crappy, and now I get this:

~# pvmove --abort -v
Finding all volume groups
Finding volume group "swissdisk_1"
VG swissdisk_1 metadata too large for circular buffer
ABORTING: Failed to write new data locations to disk. VG swissdisk_1 metadata too large for circular buffer
ABORTING: Failed to write new data locations to disk.

I'm unable to recover from this. Is there any way to forcefully remove the failed pv's? Perhaps clean this up? I can do nothing with this vg until this is resolved, and I cannot lose the data on the volumes.

Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/

1. Check if the metadata is genuinely too large for the metadata area. (eg vgcfgbackup into a file)

2. If so, see if there's anything you can change to make it smaller eg by changing LVs that are not involved in the pvmove.

We haven't written tools to resize the metadata area yet: you have to ask for more space when you ran the original pvcreates.

If you really can't shrink it, there are workarounds involving storing the metadata elsewhere: either on a new PV created with a larger metadata area that you add to the VG, or else in the filesystem.

Alternatively, depending what changes you've been making, you might be able to restore to a metadata backup - may lose changes if you've been updating the data in the areas moved though during the pvmove.

Alasdair

[email protected]