WG: LVM / DBCA error

UPDATE!
I finally managed to create the volume as desired. As it didn't work at first, I thought the problem maybe was the LVM and thus tried to add the vpathes one by one, no volume, just as single devices, formatted with OCFS. This didn't work at all, if I touched a file on one node it didn't appear on the other, obviously some mapping problem because debugocfs showed different vol_ids on both machines for the same device. Therefore I opened a TAR on Metalink, but the guys there couldn't help me :/ So I just gave LVM another try, and it finally worked! For those among you who are interested read on (I created one volume out of eight new vpathes):

1. Check that the vpathes are visible to the machine: datapath query adatper, datapath query device, cfgvpath devices
2. Create physical volumes: pvcreate /dev/vpath[e-l]. Take care that you take the whole vpath, do not make a partition out of every vpath (e.g. /dev/vpath[e-l]1).
3. Create volume group: vgcreate vg_oradata2 /dev/vpath[e-l]
4. Find out total physical extents: vgdisplay vg_oradata2 |grep "Total PE"
5. Create logical volume: lvcreate -l <insert Total PE> vg_oradata2 -n lv_oradata2
6. Create OCFS: mkfs.ocfs -b 128 -L oradata2 -m /oradata2 -u 700 -g 701 -p 0755 /dev/vg_oradata2/lv_oradata2 -F (I _guess_ the -F was the reason why my first try with LVM didn't work, I think maybe I forgot this option.)

Then I just needed to create the new mountpoint and mount the new volume. That's how it worked for me.

Cheers,

Nora

> -----Ursprüngliche Nachricht-----
> Von: BERNHARD Nora
> Gesendet: Mittwoch, 31. August 2005 11:49
> An: oracle-l_at_freelists.org
> Betreff: LVM / DBCA error
>
> Hello all,
>
> For a new database on my 10g Linux RAC (on SLES 8) I got assigned several
> small devices on our IBM Shark storage. To use them properly, I installed
> LVM in one of the recent versions that support vpath. I easily managed to
> tie all the devices together in one volume and mounted it on both nodes.
> Everything ok so far, if I touch a file on one machine it naturally
> appears on the other one as well, no errors. But as I now try to create a
> new database on both nodes I receive the following error in DBCA: "The
> directory "[/oradata2/sid]" is not on cluster file system partition shared
> across nodes "node1, node2". Hmmm... Google couldn't help me yet - anybody
> got an idea? And I wonder whether I could get some support on Metalink on
> this topic?
>
> Thanks a lot!
>
> Nora Bernhard

--
http://www.freelists.org/webpage/oracle-l