pvcreate fails on HP-UX 11iv3

[Ian Dennison_1]

Gents,

Get the following message when attempting to pvcreate on 11iv3,...

#pvcreate -f /dev/rdisk/disk10
pvcreate: clearing the bad block directory: Device busy

# Results of ioscan -dmf
/dev/rdisk/disk10 /dev/rdsk/c2t0d1
/dev/rdsk/c8t0d1
/dev/rdsk/c14t0d1
/dev/rdsk/c16t0d1

I cannot get this to pvcreate or vgcreate cleanly. Storage is an EVA5000 vcs 3.028. ANy ideas? Have tried previous postings, no luck there.

Share and Enjoy! Ian

[Jaime Bolanos Rojas.]

Ian,

It looks like LVM commands are not able to clean the BBdir in the disk, either you might want to replace the disk or open a support call , HP might be able to help clearing that up in the disk.

Regards,

Jaime.

[Johnson Punniyalingam]

Hi There,

Can please Check do you have this patch

PHKL_35970

looks like you need to patch with "LVM Cumulative patch"

Thanks,
Johnson

[Ian Dennison_1]

Gents

Thanks for input so far. Disk is an EVA vDisk so no chance of a repair by HP. OS is version 11.31 (Sep 2008) so there are no cumulative LVM patches available.

Is there anyway I can use SecurePath? Share and Enjoy! Ian

[Steven E. Protter]

Shalom,

To reinforce.

This is a known condition with HP-UX 11.31 supposedly corrected by PHKL_35970

SEP

[likid0]

I had a similiar case, I allready had the PHKL_35970(check it out) installed, but mine was a problem with scsi reservation, the disk had been used before by a true64, so I had to map it to a true server and clean the reservation, afterwards it worked fine

[Duncan Edmonstone · Accepted solution]

Ian,

Regardless of the errors you are getting, 11.31 doesn't support an EVA running VCS3.028 - it only supports the EVA when running in ALUA mode (i.e. an EVA 4/6/8x00 or an EVA 3/5000 running VCS 4.x)

HTH

Duncan

[Ian Dennison_1]

Johnson / SEP,

Thanks for the info on the PHKL patch, I have looked it up on the ITRC and it is definitely an 11.11 patch. IIRC these don't work on 11.31 kernels, or am I missing something? Would be great if you could elaborate.

In meantime, will go with single disk device visibility by removing SAN zoning and EVA masking for all but one path and controller. (Quick, easy and not potentially damaging).

Share and Enjoy! Ian

[Duncan Edmonstone]

Not sure where SEP & Johnson are going with PHKL_35970, but this patch is *only* relevant on an 11iv1 box when importing a VG that has previously been active on a 11iv3 box. It has no relevance at all to a 11iv3 system itself and a cursory read of the patch notes would make that obvious.

The issue you are seeing is (I suspect) simply that the load balancing alogrithms in the 11iv3 storage stack are sending some IOs to the controller on which the vdisk is not active. This isn't a problem with later releases of EVA which operate active/active and are able to proxy the IOs to the owning controller, but it is on VCS 3.x.

Your idea of masking all but one port should work (as long as you make it a port on the controller owning the vdisk), but it won't alter the fact that you'll be running in an unsupported configuration.

Unfortunately an upgrade to VCS 4.x is non-trivial, particularly if you have a bunch of other hosts attached to the EVA.

Good Luck

Duncan

[Armin Kunaschik]

This is a multipathing issue!
HP-UX 11.31 does not support EVA 5000 with active/passive setup.
You need to unpresent the disks visible over the passive host ports. Only the active ports need to be visible!
Maybe it also helps to deactivate the passive links with scsimgr.. I did not try this.
Attention: Both solutions are workarounds!!! EVA5000 still stays unsupported on HP-UX 11.31.

If you want to run production applications here, get a supported storage box (like an active/active EVA) :-)

My 2 cents,
Armin

[Ian Dennison_1]

Have had to revert back to 11iv2 on the install in order to use the EVA 5000 storage.

Thanks for assistance everyone. Share and Enjoy! Ian