vgexport run with -s or not??

[Paul Senior_1]

We have a 4-node cluster running Serviceguard 11.16 on 11iv2.
I need to add LV's to an existing volume group with space allocated from a disk already within the VG. These LV's have already been created on one node and we now need to propagate the changes to the other nodes.

I beleive I need to run a vgexport -p -m and vgimport -m but am unsure of whether to run the vgexport with a -s (i.e to include the VGID). I believe the different nodes have different VGID's so I currently believe I should run vgexport without the -s option. Is this correct?

[F Verschuren]

if you use the -s option you wil not have to name all disk whit the vgimport, so the best way to use the -s option.

[Geoff Wild]

vgexport -s -v -p -m /tmp/vgXX.map /dev/vgXX

That's the best way to create a map file in a SG environment...

Rgds...Geoff

[F Verschuren]

the vgID is the id that is on the disk, you can take it out of a disk by using:
echo '0X2010?2X' | adb /dev/dsk/c?d?t? | expand | tr -d '' | sed 's/2010:/VGID/'
please check the disk befor and you know for sure.

[A. Clay Stephenson]

The same VG will have the same VGID on all nodes. When you use -s, vgexport writes the VGID in text form to the mapfile along with the LVOL names. The -s vgimport command then reads the mapfile and searches all attached disks for that same VGID. If you know the PV paths for each disk on each node (which could vary greatly across the nodes --- and that's fine) then the -s option is "optional" but if you do not know the paths to each disk on each node then the -s option is essential.

One possible downside to using -s is that the i/o distribution will not be optimized because every LUN in an array will usually get the same primary path. If you want to optimize the i/o to use multiple paths then you you will need to run vgreduce and vgextend to rearrange the primary and alternate paths. This usually doesn't matter as much as it once did when copper SCSI was the norm.