sharing a logical volume

[zenari]

Hi,

I've got two 3440 servers running HP-UX 11.11 and connected to a DS2405 disk enclosure.

I wonder how to share data.
Do I need to declare logical volumes on each server ?

Thanks,
Olivier

[RAC_1]

A shared volume group. But can be activated on one node at any given point in time.

NFS. Export it other node and mount. If set up properly could serve your purpose.

MCSG(Service Guard)

Anil

[Dave Olker]

Hi Olivier,

Are you talking about sharing a file system on both systems for simultaneous access? If so, this is really only supported for read-only access. If you're going to be doing write access, there would be nothing to stop both systems from corrupting each other's data.

A supported way of sharing volumes between systems is to use ServiceGuard to manage the filesystem and move it between systems if one of them becomes unavailable. You could then use NFS to access the filesystem on the client where the filesystem is not currently locally mounted.

Again, you *could* use vgexport/vgimport to create the same logical volume on both systems and potentially have both servers accessing the same disks simultaneously, but you'll be taking your chances with disk corruption.

Regards,

Dave

[Muthukumar_5]

To share the datas on networks use Network File System to get effective performace.

Make the available one system and export them to all so that they can mount and use the data's out there.

It is good to keep the secured / allowable users to access the data. See the exportfs man page so that, you can allocate permission for the data to the machines.

NFS is the best way to share and use data's.

HTH.
Muthu

[Sundar_7 · Accepted solution]

You need to create the volume group in one node and export it to the other node.

But as mentioned above, you should not activate the volume group for write access on both the machines at the same time.

This is mainly becoz the HP-UX kernel doesnt have any kind of mechanism to communicate/sync up the filesystem metadata changes to a kernel running in an another system except ofcourse, with the exception of using NFS (Network File system).

You can enable the volume group in rw mode in one node and only r-mode in another mode. But again there is a school of thought that even this can lead to the data corruption.

[Brian Markus]

With the disk enclosure you can have cables going to multiple systems, and partition the disks off. Iâ m not sure how smart that particular enclosure is, most of the newer mini SANâ s and enclosures have a "system partitioning" or "secure LUN manager" built into them. You should be able to go into the disk manager tool and click disks 1-5 to system A and 6-10 to system B. As far as actually sharing the same disks between systems, I agree with the rest of the group. I don't advocate making a volume group shared unless need to do it for Oracle Parallel server or something like that using MC/Serviceguard OPS/RAC edition. However, if you do want to do it, here is how:

If your disks have multiple paths and are visible on both nodes in a cluster, you can put them in shared cluster mode.

Make vg05 shared:
vgchange -c y -S y vg05

Make vg05 active:
vgchange -a y vg05

Deactivate vg05
vgchnage -a n vg05

Turn sharing off for vg05
vgchange -c n -S n vg05

You will most likely corrupt your data if you donâ t do this right. Youâ ll need to have a lock disk and a lot of other things setup prior to doing this. Make sure you have good backups and you read the docs.

If you just want to have an exclusive disk for each machine then share them off to the other machine, you could use samba, NFS, or any other of the tools out there. I actually use rsyncd for syncing directories on some machines.

I hope this was helpful. Good lu

[Brian Markus]

Wow these posts are delayed. Here is how to export your vg then import it on the second node.

On node1:
vgexport -pvs -m /tmp/node1_vg05.map /dev/vg05

**The -p option will preview, so your vg will not be lost on node1**

Copy this map to node2 (FTP, RCP, etc... you choose)

On node2:

mkdir /dev/vg05
mknod group c 64 0x050000

**Do not duplicate numbers on the same node.**

vgimport -v -s -m /tmp/node1_vg05.map /dev/vg05

Then do all the vgchanges I mentioned before.


Good luck!

-Brian.