Agile Multipathing Problem

[Craig Johnson_1]

We have a situation where one server in a cluster isn't seeing the SAN disks properly.  We are seeing an agile address be assigned for alternate paths to the same device, as well as to a Clariion disk already in use (disk943).  We are using EMC Symmetrix and Clariion disk (migrating off the Clariion):

# ioscan -m dsf | grep t11d7
/dev/rdisk/disk1123      /dev/rdsk/c28t11d7
/dev/rdisk/disk1129      /dev/rdsk/c29t11d7
/dev/rdisk/disk943       /dev/rdsk/c12t11d7
                         /dev/rdsk/c11t11d7
                         /dev/rdsk/c10t11d7
                         /dev/rdsk/c9t11d7

agile getting apparently confused over the mixture of Clariion and Symmetrix disk:

# syminq | grep t11d7
/dev/rdsk/c9t11d7          DGC       CX4-960WDR5  HP03 Ch2 CONT     37748736
/dev/rdsk/c10t11d7         DGC       CX4-960WDR5  HP03 Ch2 CONT     37748736
/dev/rdsk/c11t11d7         DGC       CX4-960WDR5  HP03 Ch2 CONT     37748736
/dev/rdsk/c12t11d7         DGC       CX4-960WDR5  HP03 Ch2 CONT     37748736
/dev/rdsk/c28t11d7         EMC       SYMMETRIX    5773 2203AD8000    8923200
/dev/rdsk/c29t11d7         EMC       SYMMETRIX    5773 2203AD8000    8923200

Anyone seen anything like this before?

[Craig Johnson_1]

Note that on the first node, we didn't see the multiple agile addresses to the same disk, and vgextend worked.  The trouble was discovered when we tried to do the vgimport on the second cluster node.  It gets confused.

In any case, having the same tXXdXX for physically different disks seems awfully odd to me.

[Robert_Jewell · Accepted solution]

Could this be an issue with the Symm and its relation to HP-UX 11.31?

Check this link:

http://h30499.www3.hp.com/t5/System-Administration/11iv3-Adding-alternate-links-to-existing-volumegroup-using/m-p/5295793/highlight/true#M475088

And the link it refers to:

http://h30499.www3.hp.com/t5/LVM-and-VxVM/Can-t-add-alternate-path-to-VG-on-11-31/m-p/5192457/highlight/true#M47369

-Bob

[Craig Johnson_1]

Thanks Bob.  The disk team found the culprit, a missed port setting on masking setup for the node that didn't work.  Yes it was the SPC-3 bit as noted in your links!

I still find it odd that there could be other t11d7's, somehow the disk on the Clariion must be presenting itself differently enough from the Symmetrix that the server can tell them apart, despite the oddly identical td numbers.