On 4/17/07, Learmonth, Peter <Peter.Learmonth@netapp.com> wrote:

Hi Mike
The path is never determined by the storage device.  The server/host (or
it's multipath implementation) determines that.  Remember "initiator"
and "target" - the initiator decides by specifying a particular target
WWPN.

I haven't played with DSM3.0 so I can't answer that part, but the NetApp
ESX Host Utilities analyzes the target WWPNs and figures out preferred
paths to avoid using proxy paths.

If you look at the link Vic gave, you can see in the picture that there
are 8 paths (2 server HBAs each going to a switch with 4 paths to
controllers).  "lun_1" should be accessed through ports on "Filer X".
The way you tell which ports are the ones is the "fcp config" or "fcp
show adapters" on the controller on which you created the LUN.  Look at
the first digit of the 5th octet of the portname, which is usually
either 8 for one controller or 9 for the other.

Here's the decoder on this:
50:0a:09:83:82:00:96:d5
5 = IEEE Registered Name format
00a098 = OUI for NetApp - just like for Ethernet
3 = Port number on this controller
8 = Which head (usually 8 or 9)
0096d5 = Unique vendor-assigned part.  Usually determined by one of the
hardware WWNs on one of the ports, then the same one is soft-assigned to
all the other targets.  Also written to disk in case you swap heads
(upgrade, etc.) so it keeps it.

The asup error (FCP Partner Path Misconfigured) that Jack mentioned
indicates that the server is using a non-preferred path.  There could be
a performance impact, but other than the annoyance of lots of asups and
corresponding case notifications, that's it.  You can figure out which
server it is by figuring out which LUN, which you do with "lun stats -o
[lun_path]"

homer*> lun stats  -o
    /vol/peterwintest/w2k3ent  (23 hours, 20 minutes, 50 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops
QFulls  Partner Ops Partner KBytes
        5               0               3         0          148
0       0           0
    /vol/vmfstest2/vmfstest2  (23 hours, 20 minutes, 50 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops
QFulls  Partner Ops Partner KBytes
        686582          96097           21313     23149      396
0       0           0

If you see "Partner Ops" and "Partner Kbytes" then a server using that
LUN is using a non-preferred path, and you need to check the multipath
settings on that server.  You can reset the counters with "lun stats -z
[lun_path]".

Share and enjoy!

Peter

-----Original Message-----
From: Mike Ball [mailto:MBall@datalink.com]
Sent: Tuesday, April 17, 2007 4:56 AM
To: toasters@mathworks.com
Subject: RE: Information on how Netapp single_image mode works

Thanks for the responses but I am still confused about how the paths are
determined. For example, lets say I have a FAS6030C (Node1 and Node2)
running Data Ontap 7.2.1 with each head having 2 target ports each in a
switched fibre fabric. In this fabric, I have a Windows host (running
Snapdrive 4.2.1 and DSM 3.0) with a dual port fibre card. I create lun0
on Node1 from the Windows host via Snapdrive. How is the path chosen by
the FAS6030C? Is there a round robin algorithm which gives out the
target port on Node1 to the Windows host?

-----Original Message-----
From: Paul Rosham [mailto:paulrosham@yahoo.com.au ]
Sent: Monday, April 16, 2007 11:26 PM
To: 'Jack Lyons'; 'Nils Vogels'
Cc: Mike Ball; toasters@mathworks.com
Subject: RE: Information on how Netapp single_image mode works

In addition, anything with a reasonable NVRAM/Cluster interconnect setup
(i.e. 3000, 6000 series) the additional latency of the request passing
to the correct head via the cluster interconect is almost negligible.

-----Original Message-----
From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com]
On Behalf Of Jack Lyons
Sent: Tuesday, 17 April 2007 11:10 AM
To: Nils Vogels
Cc: Mike Ball; toasters@mathworks.com
Subject: Re: Information on how Netapp single_image mode works

one caveat, based on my experience

the host based software will see all paths to the filer as equal, so if
you hit the WWPN on the "other" head of the cluster, you will get an
error message and an ASUP.  It is annoying, and as far as I know, not a
problem.

Jack

Nils Vogels wrote:
> Hey Mike,
>
> On 4/16/07, Mike Ball <MBall@datalink.com> wrote:
>>
>> I am trying to find additional information on how Netapp single_image

>> mode works in a clustered environment as it pertains to pathing.
>> Single_image mode pulls all the WWPN's from the Netapp cluster and
>> presents it to the SAN as one WWNN. What I would like to know is how
>> paths are determined from hosts to cluster nodes if no target portals

>> are defined?
>>
>
> There will be multiple paths available (at least 2, one for ClusterA,
> one for ClusterB) and you generally will need host-based path
> selection software to select the active one.
>
> Some software selects on 'Most Recently Used', other selects on fixed
> path, this may vary per multipathing-software.
>