HPUX 11.x and EMC Symmetrix LUN address limitation

[Aaron Newcomb]

Has anyone successfully been able to see a LUN from an EMC Symmetrix that is assigned at 08 through 0F? There seems to be a serious limitation on the part of HP to be able to see all 256 possible LUN addresses from a single fibre channel initiator. I can only see devices that are addressed in the 0 - 7 range (i.e. cXt0d7 but not cXt0d8). I thought that patch PHKL_25118 was supposed to fix this, but so far I cannot get it to work. Also, what is the stated upper limit of the number of devices that HPUX will recognize on a single HBA (A5158A in this case) given that it could connect to multiple subsystem ports through a switch?

[harry d brown jr]

You might want to take a look at this patch:

http://us-support3.external.hp.com/wpsl/bin/doc.pl/screen=wpslDisplayPatch/sid=166ec6510691702921?PATCH_PATH=/hp-ux_patches/s700_800/11.X/PHKL_24268&HW=s800&OS=11.00

live free or die
harry

[Deshpande Prashant]

Hi
On my system I see luns through cXtYd0 through cXtYd7.
I remember, EMC Eng. mentioning me on HP fiber adaptors the performance degrades after 128 LUNs and so we converted them to Meta-volumes.

Thanks.
Prashant.

[Aaron Newcomb]

Harry,

Thanks for the info. The patch you reference was re-released to resolve issues with PHKL_25118, but does not include the LUN address fix. We are not having problems with the PHKL_25118 patch so I don't think going backwards would resolve the situation.

Prashant,

I would re-visit that issue with EMC if I were you. The number of LUNs alone should have no bearing on performance unless your data suggests that throughput would be an issue. I could have thousands of LUNs (theoretically) on an HBA and as long as only one LUN was being accessed at a time I would never run into performance problems. It all depends on your particular setup.

[Aaron Tatone]

Aaron,

You are correct. All HP-UX hosts can only have 7 Disks (luns) per Target. That is because the Controller is an 8-bit (1 byte)field (00-FF), the Target is a one-byte field (00-FF) and the Disk is a 3-bit field (00-08). I have called HP and talked to their back line about this. The only way to get sixteen LUNS per target is to be running a superdome with HP-UX 11i and all the appropriate patches.

Hope this helps

Aaron

[Aaron Tatone]

As a side note, this same issue (only 8 LUNS) occurs for all storage devices attached to HP servers, including the HP storage (VA7x00, XP, etc).

Aaron

[Vincent Fleming]

I beg to differ, Aaron.

What actually happens is that the actual LUNs are mapped to "ghost" targets in FC. If there are 32 LUNs, 4 ghost targets are created... You certainly can "see" all of the LUNs on the SAN.

[Aaron Tatone]

you're right. You see all LUNs, but after the eighth disk, the t# increments.

i.e.
c0t0d1 - LUN 1
c0t0d2 - LUN 2
c0t0d3 - LUN 3
c0t0d3 - LUN 4
c0t0d4 - LUN 5
c0t0d5 - LUN 6
c0t0d6 - LUN 7
c0t0d7 - LUN 8
c0t1d0 - LUN 9
c0t1d1 - LUN 10

[liam Fogarty_1]

I agree with Aaron,
With hp you can only have a total of 8 luns per target. HP-UX simply does not scan the bud above those luns...


Liam

[Stuart Abramson_2]

1. There is a very good short white paper on the hpux web sited called

Maximum LUN configuration considerations for HP-UX.
www.docs.hp.com

2. The limits are:

cCtTdD:
cC:
tT: 0-15
dD: 0-7

15 x 7 = 105 PVs on a single "vbus"

3. But you can have multiple "vbus"-es.

vbus might be an EMC term.

When the vbus changes, the C number increments.

4. The limit is 256 PVs/HBA (I think!).

It turns out that is both an HP and an EMC limit (I think).