Any EVA admin's out there?

[Manuel Contreras]

I was conveniently excluded from some EVA8000 training, and now find myself with an EVA zoned, but no physical devices visable...

My ioscan (already ran insf -e) appears as follows:

ba 1 0/1 lba CLAIMED BUS_NEXUS Local PCI Bus Adapter (782)
fc 0 0/1/0/0 td CLAIMED INTERFACE HP Tachyon TL/TS Fibre Channel Mass Storage Adapter
/dev/td0
fcp 2 0/1/0/0.1 fcp CLAIMED INTERFACE FCP Domain
ext_bus 11 0/1/0/0.1.12.0.0 fcparray CLAIMED INTERFACE FCP Array Interface
target 16 0/1/0/0.1.12.0.0.0 tgt CLAIMED DEVICE
ctl 9 0/1/0/0.1.12.0.0.0.0 sctl CLAIMED DEVICE HP HSV210
/dev/rscsi/c11t0d0
target 17 0/1/0/0.1.12.0.0.1 tgt CLAIMED DEVICE
ctl 10 0/1/0/0.1.12.0.0.1.0 sctl CLAIMED DEVICE HP HSV210
/dev/rscsi/c11t1d0
target 18 0/1/0/0.1.12.0.0.2 tgt CLAIMED DEVICE
ctl 11 0/1/0/0.1.12.0.0.2.0 sctl CLAIMED DEVICE HP HSV210
/dev/rscsi/c11t2d0
target 19 0/1/0/0.1.12.0.0.3 tgt CLAIMED DEVICE
ctl 12 0/1/0/0.1.12.0.0.3.0 sctl CLAIMED DEVICE HP HSV210
/dev/rscsi/c11t3d0
target 20 0/1/0/0.1.12.0.0.4 tgt CLAIMED DEVICE
ctl 13 0/1/0/0.1.12.0.0.4.0 sctl CLAIMED DEVICE HP HSV210
/dev/rscsi/c11t4d0
target 21 0/1/0/0.1.12.0.0.5 tgt CLAIMED DEVICE
ctl 14 0/1/0/0.1.12.0.0.5.0 sctl CLAIMED DEVICE HP HSV210
/dev/rscsi/c11t5d0
target 22 0/1/0/0.1.12.0.0.6 tgt CLAIMED DEVICE
ctl 15 0/1/0/0.1.12.0.0.6.0 sctl CLAIMED DEVICE HP HSV210
/dev/rscsi/c11t6d0


We have two controllers zoned to the same luns, but It is my understanding that we may need a product called secure path.

Is there a way to configure an EVA w/pvlinks (no secure path)?
There appear to be a few failover modes on the EVA...currently may be to "no preference"?

Do I need to unzone one path for this to work?

server OS = 11.0

any assistance would be appreciated,
manny

[Manuel Contreras]

bouncing the server resoved the problem...I think?
The following devices are now visable on CRAB:

------------------------------------------------------------------------
DEVICE :VEND :PROD :REV :SER NUM :CAP(kb)
------------------------------------------------------------------------
/dev/rdsk/c1t0d0 :SEAGATE :ST173404LC :HP03 :3CE0HMM2 :71687368
/dev/rdsk/c1t2d0 :SEAGATE :ST173404LC :HP03 :3CE0JAD2 :71687368
/dev/rdsk/c2t0d0 :SEAGATE :ST173404LC :HP03 :3CE0JPS8 :71687368
/dev/rdsk/c2t2d0 :SEAGATE :ST173404LC :HP03 :3CE0JQZ9 :71687368
/dev/rdsk/c3t2d0 :HP :DVD-ROM 305 :1.01 : 2000/09 : 658948
/dev/rdsk/c11t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c11t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c13t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c13t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c15t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c15t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c17t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c17t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c19t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c19t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c21t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c21t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c23t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c23t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c25t0d1 :HP :HSV210 :5031 :8398ASF0 :104857600
/dev/rdsk/c25t0d2 :HP :HSV210 :5031 :8398ASF0 :104857600
#

A question I now have, is I see 16 100GB devices...I believe the EVA guy only setup TWO 100GB luns.

Also, they all appear tobe the SAME device...

the c10 and above are visable from one HBA, while the c19 and above are visable from the other HBA.

[Nigel Poulton]

Hi Manuel,

Im not an expert on the OS that you're using but to me it looks like you can see at least two disks -

dev/rdsk/c11t0d1
dev/rdsk/c11t0d2

I believe that this output should be read as -
Controller 11 Target 0 Disk 1
Controller 11 Target 0 Disk 2

So you have two disks seen down the same controller and target. The same applies for the rest of the lines, they all appear to see 2 disks.

Obvioulsy the fact that you seem to be seeing the same two LUNs down many controlelrs/paths appears to be a mulipathing problem.

Hope this helps

[Luk Vandenbussche]

Hi Manny,

You have only two luns defined, which are visable though your hp9000
cxtyd1 and cxtyd2. You see each Lun 8 times, this depends on the setup of your san (fe. number of cards in your server, zoning setup)

You can bundle this luns to one lun be installing securepath for hpux, this software is not free.

Our you can use pvlinks, this is for free.

In both configurations you can balance the load over all the controllers

[Uwe Zessin · Accepted solution]

> You can bundle this luns to one lun be installing securepath for hpux,

Sorry, but that is not possible with the EVA8000, which is an Active/Active storage array. In that case you need to install the 'Autopath' driver, which does not map the paths to one new pseudo device.

[Luk Vandenbussche]

Securepath V3.0F, is mentioned in the quickspec as Failover software for HPUX 11.11 of 11.23

This means also that your version 11.00 is not officialy supported on an EVA8000

[Manuel Contreras]

You guys are right, d1 and d2 are the two luns that where created...

I am very interested in using pvlinks for fail-over, as secure-path may not be an option...
Is it due to the 8 back-end EVA ports, I see 8 hwpaths for each lun?

Is it due to the active-active 8000 features, secure-path is the ONLY answer (pvlinks is NOT an option)?
Having 2 (x8) is one thing, but having 355 (x8) is not something I'm NOT looking forward too...


also,
I opened a case w/HP last night, and they talked about a tool called sssu.
This tool would allow me to query the EVA from cli on hpux, and retrieve lun info, like wwn.

Sounds like something I want installed on my box...

thanks,
manny

[Uwe Zessin]

> I am very interested in using pvlinks for fail-over,

OK.

> as secure-path may not be an option...

It is if you want dynamic load balancing between the paths. PVlinks does not do that. Whether you really need the additional bandwith, of course, I don't know.

>Is it due to the 8 back-end EVA ports,

Please allow me a minor correction: the back-end is where the disk drives live. The servers are connected to the front-end.

> I see 8 hwpaths for each lun?

Yes, although the exact number of LUNs depends on the connected host ports, the number of Fibre Channel adapters, the zoning in the switches and so on - it can look quite messy.

[Thomas J. Harrold]

What you are seeing is one path to each LUN from each of the EVA's front-end ports, zoned to each of your HBA's.

You don't have to do anything to use PVLinks, other than add multiple (up to 8) different paths into your LVM configuration when you build Volume Groups.

If you really plan to have several hundred physical LUN's, you may want to be carefull about how you add them to the system.

I'd suggest creating them on the EVA, but NOT presenting them to the system until you need them. Then present them one at a time. (using the gui "command view eva" on your storage management server)

It will be easier to determine the new devices, and you can immediately add the new devices to a volume group.

Just curious, why so many devices? Is it an option to create fewer, larger LUNs instead? With EVA, there is no real performance benefit to creating multiple LUNs.

-tjh

[Manuel Contreras]

Creating them one-at-a-time makes sense for simplicity/identification.
I'm not used to 7 alternates when creating volumes, but I'll accept it only if the configuration is valid...which it sounds like it is in a non secure-path env.

The reason the smaller lun sizes are due to lun sizes on production systems, and leveraging a tool to perform back-end copies from lun-to-lun (must be same size or larger) for periodic re-fresh/cloning.

thanks everyone for helping me become more familiar w/this box...
manny