Fixing minor numbers for shared Volume Groups

[Shawn Jones_4]

I have a 5 node cluster that I am cleaning up. The VG minor numbers are not identical across all 5 nodes. Can I rm the group file and then do a new mknod, or do I have to vgexport the VG, then mkdir, mknod, and reimport the VG across all the out of sync nodes?

[A. Clay Stephenson]

You really don't have to do anything but life is probably a bit simpler if the minore device numbers are consistant. You need to do a vgexport and build your new vgxx directory and group file and then vgimport.
Don't forget to do a vgcfgbackup on each node so that you can easily replace a failed disk regardless of which node is active.

[Shawn Jones_4]

So I can't just rm the group file on the inactive vg's and the just do a mknod to create a new group file with an identical minor number? I just want to be clear on that.

[G. Vrijhoeven · Accepted solution]

Hi,

What you can do is:

1. Let the node on with the package is active be the leading one.
2. On all nodes on which the volume group is not active you can do a vgexport -v -f vgname.devfile -m vgname.mapfile vgname of the inactive vg. mknod -c /dev/vgname/group 64 0xnumber ( the one configured on the active node) vgimport all the disks using the vgname.mapfile and vgname.devfile. vgchange -a r vgname && vgchgbackup vgname && vgchange -a n vgname.

Now all your minor numbers are the same with no down time.


HTH,

Gideon

[A. Clay Stephenson]

By the way, make absolutely certain that the minor device number you choose is unique across all 5 nodes. If two active VG's happen to use the same minor device numbers, absolute chaos is assured.

[Shawn Jones_4]

Thanks for all your help, I ended up recreating the vg's since they weren't active.