MC/ServiceGuard - Virtual Array 7400

[matt_22]

I am trying to configure two nodes service guard on HP virtual array to share the same disks space.

At the moment, ioscan shows the same disks with different hardware paths on the individual machines.

Any clues?

[Christopher McCray_1]

It is possible that your array is connected to cards on each server with two different addresses. Or, it is possible you are looking at the primary/alternate paths from each server. I'm just guessing, but does either of these fit your case? Can you post your exerpt of ioscan? Good luck.

Chris

[G. Vrijhoeven]

Hi Matt,

This is something that you would like to adjust. You can do that by adjusting the hardware pad in you kernel.
for references see man ioinit.

[melvyn burnard]

Well it depends on how you have configured the VA7100.
The HBA's in each node could have a difierent Instance or "c" number in the /dev/dsk/c_t_d_
fields, but as long as they have the same t and d(target and lun) numbering then there is nothing wrong with that.
Havig said that , I have seen instances where for some weird reason, arrays have been configured such that on one node (primary) they have the luns at one address, say c1t2d5, and on the other node (backup) they have set the array to show the lun at c?t15d7.
I guess you need to ensure you are presenting the Luns on the array at the same target an dlun number, and then lets see the ioscan outputs.

[matt_22]

Thanks for your comments:

see attached ioscan output from both machines:

Note: I have already created a dummy FS into the primary node PV. It works fine.

Qs: Even though I cannot pvcreate the same device on secondary node, should I take it that serviceguard software will take care of the data access share concern?

[A. Clay Stephenson]

Hi Matt:

The fact that the paths differ on the nodes matters not at all - once the vg has been imported by all the nodes. After you do the vgchange -c y to make the vg cluster aware data is written on the disks themselves to not allow mounting by more than one node (assuming you want 'exclusive' mode).

It is very common to not be able to have identical device path names for all the volume groups to all the nodes in the cluster. Life is a bit easier (or more convenient) that way but it is certainly not necessary or even very difficult if the device paths differ.

Consider the case of a cluster with a D-box and and N-box sharing a package. It would be most unusual if the device pathes were the same and yet MC/SG works just fine.

Clay