Splitting Dual Port Cards on vPars

[Albert Smith_1]

Question - HP-UX 11.23 on an rx8620. I have a server that has multiple dual port Fibre and NIC's. Is it possible using IO padding to put 1 port of each on one vpar and the other ports on the vpar?

thanks...

[Sameer_Nirmal]

It is NOT possible.

Vpar allocates IO at LBA level. Saying that IO devices at and below a LBA could only belong one Vpar not matter what if they are dual ported.

Refer this link to know details about Vpar's allocation at LBA level.
http://docs.hp.com/en/T1335-90057/T1335-90057.pdf

In rx8620, there is one-to-one correspondance of a LBA and PCI slot.
Refer the diagram in this link
http://docs.hp.com/en/A7026-96031/ch01s02.html#d0e898

So the bottom line, each PCI slot is controlled by a LBA and as per Vpar pre-defined/fixed allocation, it can only belong to one Vpar and can't be padded or splitted across the Vpar.

[Albert Smith_1]

Intersting beacuse in HP Document's for Vpar's. It shows splitting dual port SCSI using padding. I would assume you could do the same for Network and Fibre.

T1335-90051
March 2006

Vpar's 8th Edition

[Torsten.]

Hi,

what chapter (page) you are refering to?

From the doc:

"When assigning IO, if you specify a path below the LBA level (for example,
cell/sba/lba/.../device, vPars automatically assign the LBA to the virtual partition.
...
The assignment rules of LBAs remain applicable: the LBA can only be owned by one virtual partition."
(page 158 - T1335-90057)

[Sameer_Nirmal]

As mentioned in the document the "padding" means to pad using .0 up to six elements after the first dot in the h/w path. It is a new hardware path format introduced since Vpar version A.02.02. This new format is useful for Combo Cards to identify/use the end device which are below cell/sba/lba/pci_bridge h/w structure in Vpar configuration. The word "padding" here doesn't mean hardware bifurcation/separation from Vpar perspective.

As per the Vpar rules, the h/w bifurcation is done at cell/sba/lba. All the devices under cell/sba/lba would be part of a single specific Vpar.

The statements which Torsten has mentioned appears in the document which you are referring ( page 162 T1335-90051 March 2006 )

[Albert Smith_1]

Took training and now know vpars better.