vgscan errors

[Robert-Jan Goossens_1 · Accepted solution]

This error was being caused because all of the lvm files in /usr/sbin had owner
and group permissions set to something other than root:sys. In this case the
dba's had changed all of the files in /usr/sbin to oracle:sys.

http://www4.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000072806637

Robert-Jan

[Stephen Keane]

OK, that explains the vgdisplay problem, the suid bit on vgdisplay had gone AWOL. Takes us back to the vgscan problem.

All the vg* lv* and pv* files in /usr/sbin are root/sys, except for lvmmigrate which is bin/bin and doesn't have suid bit set.

All the vg* lv* and pv* files in /sbin are root/sys, except for lvmrc which is bin/bin and doesn't have suid bit set.

[Stephen Keane]

Ooops, that was a mistake, I should have awarded that 10 pointer at the end, now people will think the original problem was solved :(

[Robert-Jan Goossens_1]

could you post your lvm patches.

# swlist -l product | grep -i lvm

[Stephen Keane]

# swlist -l product | grep -i lvm
LVM B.11.11 LVM
PHCO_23333 1.0 LVM Virtual Array support

[Eric Antunes]

Hi again,

Did you try to increase maxssiz?

[Stephen Keane]

Hello again Eric,
yes I increased it as you suggested and the 64-bit one (just in case), but sadly it didn't seem to have any effect.

[Stephen Keane]

# swlist -l product | grep -i lvm
LVM B.11.11 LVM
PHCO_29379 1.0 LVM commands cumulative patch
PHKL_26743 1.0 LVM Cumulative Patch


Just added those patches, despite cpm_collect script saying I was fully patched. Doesn't solve the problem though.

[Robert-Jan Goossens_1]

There are a few new lvm patches available. I can't find a definite fix for yours by the way.

http://www4.itrc.hp.com/service/patch/patchDetail.do?BC=patch.breadcrumb.main|patch.breadcrumb.search|&patchid=PHCO_29379&context=hpux:800:11:11

[Robert-Jan Goossens_1]

please ignore my last post :-)

[Stephen Keane]

OK, cracked it. Eric if you'd like to post again, I can award you some points, as your solution worked (when I increased maxssiz to 0x0A000000) and now vgscan seems happy.

It won't be until Monday though as I'm off home now.

[Eric Antunes]

Hi,

Do you have only 250Mb of physical memory?

Can you post your dbc_max_pct, dbc_max_pct and shmmax:

#kmtune -lq dbc_max_pct

#kmtune -lq dbc_min_pct

#kmtune -lq shmmax

[Eric Antunes]

All right!

Ignore my previous post then.

Best Regards,

Eric Antunes

[Stephen Keane]

What was trowing me was that vgscan had been working fine, then suddenly it stopped working. I hadn't installed any software, or patches so why would it suddenly break?

The answer turned out to be that the root VG disk used to be quite underused and one of the other admin guys had added a logical volume to the root VG that pretty much used up all of the remaning space on the disk. To compound the misery, the new logical volume did not have a filesystem on it and hence wasn't mounted or particularly visible. As I say, this isn't a production box so procedures on it aren't as tight.

So I guess the extra number of PE/LE in use was enough to cause vgscan to blow its stack?

Thanks for your help.