Skip to contentSkip to conversation

lvremove or vgextend: metadata too large for circular buffer

linux-lvm Andrew Gideon 2 participants 4 messages Mar 11, 2011 – Mar 12, 2011

lore
Share & tools
lore
#0 lore raw

We're hitting the "metadata too large for circular buffer" problem. Last time this occurred, I was able to clear out a little space by removing some unused volumes, and then make more space by selective use of pvmove to defragment. This worked well.

However, that time I didn't get the error from lvremove. This time, that is occurring. So I don't seem to be able to make any space.

Once I execute lvremove on a volume, the "LV Status" becomes "NOT available". The command 'lvdisplay -m' still shows the segments, however.

How can I resolve this? Trashing the entire VG is not something I can afford.

I created new PVs with larger allocations of space for metadata, but I cannot even add these to the VG. The vgextend command also fails with "metadata too large for circular buffer".

- Andrew

#1 lore raw
Quoted 2 lines · Andrew Gideon

However, that time I didn't get the error from lvremove. This time, that is occurring. So I don't seem to be able to make any space.

I've tried vgcfgrestore, but this too yields:

metadata too large for circular buffer

I ran some commands with -vvv. The output which seems to always immediately precede the error is:

Doubling metadata output buffer to 131072

I also noticed that it appears to be looking at /dev/sdi and /dev/sdj. These are two PVs I'd tried to add, but where vgextend also issued the "too large..." error. These two PVs do not appear in 'vgdisplay -v'.

[root@backup2 backup]# /usr/sbin/lvremove -d -vvv -f /dev/jetarray004/i2-backup0 2>&1 | egrep 'sd[ij]'
Opened /dev/sdi RW O_DIRECT
/dev/sdi: block size is 4096 bytes
/dev/sdi: lvm2 label detected
lvmcache: /dev/sdi: now in VG #orphans_lvm2 (#orphans_lvm2)
Closed /dev/sdi
Opened /dev/sdj RW O_DIRECT
/dev/sdj: block size is 4096 bytes
/dev/sdj: lvm2 label detected
lvmcache: /dev/sdj: now in VG #orphans_lvm2 (#orphans_lvm2)
Closed /dev/sdj
[root@backup2 backup]#

Is there perhaps some "internal" VG (ie. #orphans_lvm2) that I'd be able to successfully remove?

Thanks...

Andrew

#2 lore raw
In reply to Andrew Gideon
Quoted 10 lines

However, that time I didn't get the error from lvremove. This time, that is occurring. So I don't seem to be able to make any space.

Once I execute lvremove on a volume, the "LV Status" becomes "NOT available". The command 'lvdisplay -m' still shows the segments, however.

How can I resolve this? Trashing the entire VG is not something I can afford.

You may be able to manually do your removes on the output of vgcfgbackup, then use vgcfgrestore. Be careful.

Quoted 3 lines

I created new PVs with larger allocations of space for metadata, but I cannot even add these to the VG. The vgextend command also fails with "metadata too large for circular buffer"

What I did was to allocate MUCH more space on the new PVs,
make a new VG from them, then copy LVs from one of the old PVs. Then recreate that PV with more metadata space, add it to the new VG, etc. until all of the PVs are in the new VG, with far more metadata space.

Signature · 16 lines
Ray Morris
[email protected]

Strongbox - The next generation in site security:
http://www.bettercgi.com/strongbox/

Throttlebox - Intelligent Bandwidth Control
http://www.bettercgi.com/throttlebox/

Strongbox / Throttlebox affiliate program:
http://www.bettercgi.com/affiliates/user/register.php




On Fri, 11 Mar 2011 20:05:12 +0000 (UTC)
Quoted 27 lines · Andrew Gideon

We're hitting the "metadata too large for circular buffer" problem. Last time this occurred, I was able to clear out a little space by removing some unused volumes, and then make more space by selective use of pvmove to defragment. This worked well.

However, that time I didn't get the error from lvremove. This time, that is occurring. So I don't seem to be able to make any space.

Once I execute lvremove on a volume, the "LV Status" becomes "NOT available". The command 'lvdisplay -m' still shows the segments, however.

How can I resolve this? Trashing the entire VG is not something I can afford.

I created new PVs with larger allocations of space for metadata, but I cannot even add these to the VG. The vgextend command also fails with "metadata too large for circular buffer".

- Andrew

_______________________________________________
linux-lvm mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

#3 lore raw
Quoted 4 lines · Ray Morris

What I did was to allocate MUCH more space on the new PVs,
make a new VG from them, then copy LVs from one of the old PVs. Then recreate that PV with more metadata space, add it to the new VG, etc. until all of the PVs are in the new VG, with far more metadata space.

Creating a new VG is my worst-case scenario, but that will be expensive in terms of time - time during which I cannot do expansion of volumes. It's a little frustrating.

How risky is the idea of editing a backup file and doing the restore? If I screw it up, I can always do a re-restore of the proper data, no?

Thanks...

Andrew