LVM and VxVM

Re: Help on swaping cables between servers and disk arrays.

Go to solution
Jack C. Mahaffey
Super Advisor

Help on swaping cables between servers and disk arrays.

Here's my scenario.

I currently have three HP9000 servers, server1, server2, server3.

Each server is attached to a SureStore Autoraid.

server1 diskarray1
server2 diskarray2
server3 diskarray3

server1 was a failover server. Served as failover for server1 and server2.

server2 and server3 were previously production servers.

server1 is currently connected to diskarray2 and diskarray3 in addition to diskarray1.

Diskarray2 and diskarray3 would only be activated on server1 if server1 served as production.

I'm going to disconnect diskarray2 from server2 at the IO Expansion port so that server2 can be a failover server for server3 and be connected to diskarray3.

server1 will remain connected to diskarray2.

server3 will remain connected to diskarray3.

diskarray3 connection with server1 will be removed and moved to server2.

The cable that was previously connected server2 to diskarray2 will be connected to the IO Expansion port on server3.

Final setup:

server1 connected to diskarray1
server1 connected to diskarray2
server2 connected to diskarray3 (primary)
server3 connected to diskarray3 (failover)

diskarray1 and diskarray2 will be connected by two cables each. Cables will remain connected to server1.

diskarray3 will be connected with two cables to server2 and two cables to server3.

Now the questions:
1 - When I connect the diskarray3 cable to server2 that was previously connected to server1, will the normal mkdir, mknod and vgimport with the mapfile -s sequence be sufficient for accessing diskarray3 from server2?

2 - Will it be necessary to use vgfrestore?

3 - Am I missing anything? Hardware path to diskarray3 to server2 will be different. It pointed to diskarray2 previously.

Hope I wasn't too confusing...

A. Clay Stephenson
Acclaimed Contributor

Re: Help on swaping cables between servers and disk arrays.

1) Your normal vgimport stuff will suffice. If you power the hosts down nothing more will be needed; if you do this "hot" then you will need to run insf manually to create the new /dev/rdsk and /dev/dsk device nodes.

2) No but after the VG is activated it would be a good idea to run vgcfgbackup.

3) The one downside to using vgimport -s is that all I/O will use the primary controller. A better way is to initially do a vgimport -s; strings etc/lvbtab; vgexport; and using the data from the strings /etc/lvmtab listings do a conventional vgimport listing the primary/alternate paths to each LUN so that I/O is split between both of your external SCSI paths to each array.
If it ain't broke, I can fix that.
Michael Steele_2
Honored Contributor

Re: Help on swaping cables between servers and disk arrays.

I didn't see anything about PVlinks being used. If you have a primary and alternate path into each disk array then just by removing one PVlink to the same HBA will free up the slot you need for the new disk array. With this arrangement you now have new disk connections as well as your old. So you can create new vg's, lvols and mount points and just copy the data over from old to new disk arrays. (* Comprende? *)

1) Yes
2) No
3) See above comments.
Support Fatherhood - Stop Family Law