You are going to be using the "storage adapter" command to disable and re-enable all the adapters.

If you want to do this live, you have to be careful of the four paths.

For instance, if your multipath paths are:

Node 1 0a/1c

Node 2 0a/1c

I would do something like this:

Node 1> storage adapter disable 0a Node 2> storage adapter disable 0a (provided it the same path as node1) --> basically, we want to shut down the entire path from both heads --> i.e. shelf 1 top ESH4 module and shelf 2 top ESH module

--> re-cable the link attached to 0a including the new shelves, daisy-chaining as needed.

Node 1> storage adapter enable 0a Node 2> storage adapter enable 0a

The system should pick up the new disks and will complain about not being multipath-HA

Repeat the process for the other path. Node 1> storage adapter disable 1c Node 2> storage adapter disable 1c (provided it the same path as node1) --> i.e. shelf 1 bottom ESH4 module and shelf 2 bottom ESH module

--> re-cable the link attached to 1c including the new shelves, daisy-chaining as needed.

The system should pick up the new disk paths and should verify multipath-ha

not sure this is supported, but in my head it should work ;)

--tmac

*Tim McCarthy* *Principal Consultant*

Clustered ONTAP Clustered ONTAP NCDA ID: XK7R3GEKC1QQ2LVD RHCE5 805007643429572 NCSIE ID: C14QPHE21FR4YWD4 Expires: 08 November 2014 Expires w/release of RHEL7 Expires: 08 November 2014

On Tue, May 21, 2013 at 7:34 PM, Fletcher Cocquyt fcocquyt@stanford.eduwrote:

This is something I've never had reason to attempt until now.

We have a 3270 cluster (8.1.2) Head 2 aggr1 consists of these two multipathed DS14MK4 trays (IDs 1,2):

http://i.imgur.com/821iN35.jpg

We want to hot add another DS14MK4 trays racked right above (IDs 3,4):

http://imgur.com/FlMxpyz

I've read the 2011 DS14MK4 guide and verified the loop speeds are all set consistently and IDs are all unique. (note: trays IDs are 1-4 from bottom to top)

Can I simply:

1 - remove the top two pairs of fiber and transceivers (these go to head 1 (partner)) 2 - daisy chain up from those vacated ports to new trays (3,4) 3 - plug the fiber and tranceivers back in on the top tray?

We would then grow (effectively double) this aggr1's capacity

What are the risks ? (2011 guide mentions panic's are possible with loop misconfig)

(we have 60 VMs running on head 2 on a separate loop and aggregate)

Any other tips appreciated

thanks

---

Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters

I've read the 2011 DS14MK4 guide and verified the loop speeds are all set consistently and IDs are all unique. (note: trays IDs are 1-4 from bottom to top)

Can I simply:

1 - remove the top two pairs of fiber and transceivers (these go to head 1 (partner))

Not sure what you are saying here. Both pairs of fiber from the same shelf shouldn't be going to the same head in an MPHA config.

Properly done, MPHA has head 1 connected to shelf 1 ESH A and head 2 connected to shelf 1 ESH B. Then the last shelf in the loop has ESH A go to head 2 and ESH B go to head 1. The way to think of it is that the primary path is 'forward' through the one loop and the secondary path is 'backward' through the other loop. Traffic on each loop is in both directions, but thinking forward and backward through the loop helps me visualize it better.

If MHPA is properly configured, you can disconnect the fiber at the end of the loop, connect the new shelves, and then reconnect the end of the loop to the new last shelf in the loop.

I'd recommend running ConfigAdvisor (in the toolchest on support.netapp.com) to verify your configuration both before and after adding the shelves. It verifies MPHA configurations, among other things.

John