EMC devices not available during Boot process

[Ya Ery Botelho]

Hi all,

We are running HP-UX 11.00 on 4 K-class boxes connected to 2 EMC 8430 frames (2 boxes connected to each frame).

Everything had been running pretty smoothly for 2+ years until we had a binfile change done 2 weeks ago.

Now we are having a very weird problem on both boxes connect to one of the frames.
The problem occurs during the boot process. When executing script /sbin/bcheckrc we can see the following messages for each device belonging to the database volume group (vgora):

***
vgchange: Couldn't attach to the volume group physical volume /dev/dsk/cntndn

The path of the physical volume refers to a device that does not exist, or is not configured into the kernel.
***

Funny thing is that the boot process goes on after that and after the box is finally up I just have to activate the vg with "vgchange -a y" and everything works out just fine.

Does anybody have any ideas on this???

Thanks for your help,
Yan.

[harry d brown jr]

With your binfile change, did you change the lun's or IO paths to previously existing disks in your VG's, or did you simply add more disks?

For giggles, does vgdisplay -v vg## produce warnings or errors about missing disks?

live free or die
harry

[James R. Ferguson]

Hi:

Do you have any custom volume group activation? This would be done in '/etc/lvmrc'. If so, its AUTO_VG_ACTIVATE would be set to <0> to enable the 'custom_vg_activation' function.

Regards!

...JRF...

[Rita C Workman]

Let's see..you have four servers connected to EMC disks. I'm guessing maybe you are using MC/SG.

Is it possible since there was a change to your bin in your EMC...that maybe some new disks got added, and thus a new bin had to be created.
Is it possible that you added some new disk or maybe volume group and did not properly halt the package...then stop cluster control of the vg with a vgchange -c n and so forth...Now when you start back up your package is trying to connect to disks/vg's inside and outside the package. At least that's what happened to me when I was first starting out with EMC & MC/SG.
If this scenario sounds familiar .. then let me know what you did and maybe I can post what to do to fix it...
If it's not the case...than maybe someone else will have a resolution for you and I don't have to do extra typing.

Regards,
Rita

..I'll stop by and check post in a while.

[Ya Ery Botelho]

Harry: You're right! I left out this very important piece of information.
The binfile change main purpose was to unify the hypervolumes size to 8.4GB instead of the 6GB/12GB mix we had before.

As a consequence of that all LUNs had to be reassigned and now all servers see approximately 50% more devices than before.
All previous data has been erased and the databases have been recreated from scratch!

After the volume gets activated vgdisplay gives no errors or warnings whatsoever.

JRF: No custom activation! No changes made to /etc/lvmrc file ever!

Rita: No MC/SG being used. We do failover manually with in-house customized scripts.

Thanks to all,
Yan.

[James R. Ferguson]

Hi:

I think you will find that your /etc/lvmtab isn't consistent. I'd recreate it:

# mv /etc/lvmtab /etc/lvmtab.old
# vgscan -av

Before you do this, read the 'vgscan' man pages. See the discussion of alternate links and boot information.

Regards!

...JRF...