We have found that when using multiple pvscsi controllers the only reliable thing one can count on seems to be the LUN number. Controller host (first column) is completely random. We believe that the only way to reliably map these is to embed the controller and LUN id into the label or volume group name (assuming 1 vmdk per volume group)
If there are any linux method to sequentially map the host controllers this would make it more reliable.
Ideas anyone?
eg:
[0:0:0:0] disk VMware Virtual disk 1.0 /dev/sda
[0:0:1:0] disk VMware Virtual disk 1.0 /dev/sdb
[0:0:2:0] disk VMware Virtual disk 1.0 /dev/sdc
[1:0:0:0] disk VMware Virtual disk 1.0 /dev/sdd
[1:0:1:0] disk VMware Virtual disk 1.0 /dev/sde
[2:0:0:0] disk VMware Virtual disk 1.0 /dev/sdf
[2:0:1:0] disk VMware Virtual disk 1.0 /dev/sdg
[3:0:0:0] disk VMware Virtual disk 1.0 /dev/sdh
[3:0:1:0] disk VMware Virtual disk 1.0 /dev/sdi
^
Update: The following link describes the problem and we believe the problem is as stated that the pvSCSI driver just does now pass the info through such as the wwn for the controller.
vmware esx - How does Linux determine the SCSI address of a disk? - Server Fault