HP-UX Multipathing & NetApp SAN

[Richard Low]

Hi,

I have a problem making HP-UX and our NetApp SAN worked together during the SAN fail-over.

During SAN fail-over, my HP-UX servers would lost all the paths to the SAN LUNs. The hardware path to the LUNs would changed & I had to manually do an ioscan, followed by vgexport/vgimport in order to get my servers to see the LUNs again.

The ioscan output in normal situation is:
===========================================
scsi:wsio:T:T:F:31:188:1179648:disk:sdisk:0/2/1/0.1.1.0.0.0.0:0 0 4 50 0 0 0 0 17 113 200 159 132 181 203 219 :5:root.sba.lba.td.fcp.fcparray.tgt.sdisk:sdisk:CLAIMED:DEVICE:NETAPP LUN:18
/dev/dsk/c18t0d0 /dev/rdsk/c18t0d0

scsi:wsio:T:T:F:31:188:1310720:disk:sdisk:0/3/1/0.1.1.0.0.0.0:0 0 4 50 0 0 0 0 17 113 200 159 132 181 203 219 :6:root.sba.lba.td.fcp.fcparray.tgt.sdisk:sdisk:CLAIMED:DEVICE:NETAPP LUN:20
/dev/dsk/c20t0d0 /dev/rdsk/c20t0d0
============================================

The path for the VG is:
============================================
--- Physical volumes ---
PV Name /dev/dsk/c18t0d0
PV Name /dev/dsk/c20t0d0 Alternate Link
PV Status available
Total PE 19964
Free PE 1989
Autoswitch On
============================================

However, after the SAN failed over, the ioscan output becomes:
============================================
scsi:wsio:T:T:F:31:188:1179648:disk:sdisk:0/2/1/0.1.1.0.0.0.0:0 0 4 50 0 0 0 0 17 113 200 159 132 181 203 219 :5:root.sba.lba.td.fcp.fcparray.tgt.sdisk:sdisk:NO_HW:DEVICE:NETAPP LUN:18
/dev/dsk/c18t0d0 /dev/rdsk/c18t0d0
scsi:wsio:T:T:F:31:188:786432:disk:sdisk:0/2/1/0.1.2.0.0.0.0:0 0 4 50 0 0 0 0 17 113 200 159 132 181 203 219 :4:root.sba.lba.td.fcp.fcparray.tgt.sdisk:sdisk:CLAIMED:DEVICE:NETAPP LUN:12
/dev/dsk/c12t0d0 /dev/rdsk/c12t0d0
scsi:wsio:T:T:F:31:188:1310720:disk:sdisk:0/3/1/0.1.1.0.0.0.0:0 0 4 50 0 0 0 0 17 113 200 159 132 181 203 219 :6:root.sba.lba.td.fcp.fcparray.tgt.sdisk:sdisk:NO_HW:DEVICE:NETAPP LUN:20
/dev/dsk/c20t0d0 /dev/rdsk/c20t0d0
scsi:wsio:T:T:F:31:188:458752:disk:sdisk:0/3/1/0.1.2.0.0.0.0:0 0 4 50 0 0 0 0 17 113 200 159 132 181 203 219 :3:root.sba.lba.td.fcp.fcparray.tgt.sdisk:sdisk:CLAIMED:DEVICE:NETAPP LUN:7
/dev/dsk/c7t0d0 /dev/rdsk/c7t0d0
============================================

The hardware path has changed from c18t0c0 & c20t0d0 to c12t0d0 & c7t0d0 respectively.

Hence I have to manually do a vgimport/vgexport to get the server to see the LUN again & my data back:
============================================
--- Physical volumes ---
PV Name /dev/dsk/c12t0d0
PV Name /dev/dsk/c7t0d0 Alternate Link
PV Status available
Total PE 19964
Free PE 1989
Autoswitch On
============================================

We used NetApp FAS940c filers in partner mode.

Q1) Is there any way that I can get HP-UX to see all 4 paths to the VG, or get it to automatically switch to the new set of paths?

Q2) When I do a vgexport of a VG, the directories "/dev/VG" & "/dev/VG/group" would disappear. Is this normal?

[Ludovic Derlyn]

hi,

for Q1, for me no possibility to switch automatically to the new set of paths

In 11iv3, a new is paths will be stay the same

for Q2

normal if you execute vgexport /dev/vgxx
(export all informations of vg)

Regards
L-DERLYN

[Sandy Chen]

Hi Richard,

For Q1, can you tell us how you test the failover? It should be automatically switch and there should be no new path for a failover.

Q2: It is normal when you do vgexport those directories gone.

Regards,
Sandy

[Richard Low]

Hi,

I did not test it initially. We only found out when one of our SAN's controller failed a few days ago. :-p

This is the first of the few HP-UX servers that we are planning to connect to the SAN. All our existing SAN servers are Windows & have no such issues.

Our SAN vendor are still trying to come up with a solution. I thought that I may be able to get some advice, with the large number of SA in this forum & likely some will use the NetApp with HP-UX servers.

PS: Is there any multipathing product (besides PV Links) that is available for HP-UX & NetApps (something like Powerpath for EMC)? Our HP-UX is 11iv2 & 11iv1.

Thanks for all your responses.

[Torsten.]

Don't know about your netapp, but I guess it is very similar to the continuous access solutions by hp.

First, it is pretty normal to get different (not really changed) device files.

Look at this:

0/2/1/0.1.1.0.0.0.0 - /dev/dsk/c18t0d0

0/2/1/0.1.2.0.0.0.0 - /dev/dsk/c12t0d0

Can you see the difference in the hardware path? It is here:

0/2/1/0.1.2.0.0.0.0
=========+++=======

This is a different port on the switch (for the other array.

You need probably additional software/scripts to adjust the "new" device to the VG. I think you will have some add-on software on your Windows hosts too (again, no clue about this device).

Q2) If you don't want the VG is gone, vgexport with "-p" option. The files will stay there.

[Sandy Chen]

Hi,

I've used netapp but didn't get the chance to try the failover myself actually :p my cust using it for a backup disk only so there is no multipath there.

Most SAN will be showing 4 disk on configuration like you, 2 link from SAN to Switch and 2 link from Switch to Server. Not sure with NetApp if there any configuration should be changed.

NetApp itself support for Veritas DMP on HPUX, it is on their website:
http://www.netapp.com/products/storage-networking/fibre-channel-san/host.html

regards,
Sandy

[whiz_kid]

I dont think there is a way to address this issue as the luns will be published by the second header when the first fails. you will get a new set of device path for the second controller.

since netapp filer clusters are pure active/passive, only paths from one controller will be seen to the hosts at a time. when u do a 'cf takeover' all services from the first head is brought down, the disk ownerhsip will be passed over to the second head and it brought up the services. you will definitely loose disk connectivity.

i guess secuerpath can do the job as it masks all the paths underneath. when the connectivity to the controller fails, it will show as the paths to that controller as failed and rest will be active/available/standby. and when the paths are restored it will show them to be active and available/standby.

[Richard Low]

Hi Torsten,

Thanks for pointing out the port difference in the hardware path, I didn't noticed that.

Hi Sandy,
To test the NetApp failover, you can issue the command "cf takeover" & "cf giveback" in one of the filer.

Hi Whiz_kid,
I'll look into securepath.

By the way, I've just receive word from my SAN vendor that the problem may be due to the cfmode of my filers. My current filers' cfmode is "standby", which they said have some issues with HP-UX & AIX. They will confirm whether we need to switch the cfmode to "partner" mode & its impact on our existing Windows, VMware ESX, Linux & HP-UX servers as well.

Thanks for all your responses.