SAN switch migration with NetApp Controllers

[pankajgadhari]

Hi All,

We are doing a SAN migration project wherein we will be replacing the SAN switches from broacade to Cisco Nexus 5596UP. The Storage array is NetApp v3170 running ONTAP version 8.0.2 P1 7-mode. Now on each netapp controller we have 4 fiber channel ports out of which 2 are used and 2 are unused. All these ports are configured as targets. The name of the controllers are netapp1a and netapp1b (each controller having 4 fiber ports) and configured in single image mode.

Now, I want to connect the unused 2 ports from each controller to new switches without disrupting the current connections, currently the initiators or hosts are connected to 0a and 0d while 0b and 0c are not in use. Once this is connected, is there any thing else I need to worry from NetApp perspective for enabling that unused fiber ports. I will do the zoning on the new switches.

After this, I have to work on moving the SAN connected hosts (SQL servers running on windows 2008 SE) from old switch to new switch. Can anybody share the steps on what needs to be done to migrate the servers online without taking the downtime. As these servers are multipathed, I would like to know how I can move one path at a time so that there is less downtime of the server involved in the migration. If any of the experts has done this before, please share the procedures or steps of doing that on cisco + NetApp configuration?

Thanks

[billshaffer]

You can connect the two unused ports to the new switch with no disruption.  You said the unused ports are already targets; you'll also need to verify/configure them with the fcp command.

The host multipathing software should eliminate downtime.  If the software allows you to disable one path, do it, then remove the zoning for that path from the old switch, and add it to the new switch.  Verify that the host WWN logs in and associates with the correct igroup on NetApp.  I can't recall if the WWN that shows up on NetApp is associated with the port or not - if it is, you may need to create a new igroup and map the luns to it as well as the existing igroup.

At this point you should be able to flip the active path in the multipathing software to the one coming off the new switch, remove the zoning for the second path from the old switch and add it to the new switch, and again verify WWN and igroup connectivity.

If you can't manually disable a path with your multipathing software, at least verify that it sees both paths before changing the second zone!

Bill

[pankajgadhari]

Hi Bill,

I verified the adapters using fcp config command and the unused adapters are showing as UP, but link not connected. I think that should be good, right?

Regarding migration of hosts. Whether it will create any problem if one path from the host is connected to Brocade switch and the other path is connected to Cisco Nexus switch, as there are different vendor fabrics?

As of now, I am thinking of following plan:

1. Connect unused ports 0c and 0b to Cisco Switch and verify they are online.

2. Disconnect one path from the Server and connect to Cisco switch, verify they are online and zone the WWN to 0c and 0b.

3. On NetApp, no need to change the igroups as the hosts WWWN will be same as I am using the same hba.

But my question is - once the zoning igroups is configured properly whether I need to take the reboot of the Server to make the disks online, or it will just appear as soon as the zoning is in place and I should see the disks online, is that right?

Let me know if I am missing anything else on this.

Thanks

[billshaffer]

If fcp config shows the same for all the ports, then you should be good.

I have not done it, but I don't see why there would be a problem with two paths to the same device being from different vendor SANs - assuming your multipath driver supports them both.  As long as, after switching the first path, the luns come up with two paths, you should be okay.

Once you connect/zone the first path to the new switch, you'll need to follow whatever your normal procedure is for scanning in new luns - typically a scan on the HBA side, then an import of some sort to the OS, then an import of some sort to the multipath driver.  (You should probably do a proper delete of the lun prior to pulling the zoning, too.)  At this point the multipath driver should identify the luns from the new switch with the luns from the old switch - that is, instead of showing a bunch of new luns, it should show everything with dual paths again.  I mentioned the possibility of creating a new igroup because I can't remember if the WWN presented to NetApp references the port on which it comes in; if your "igroup show" only shows the standard WWN, then you probably don't have to worry about it.  One of the things you want to do once the server is attached via the new switch is make sure it shows online in the igroup.

No reboot should be necessary as long as you can typically add luns online, since that is essentially what you are doing.

If you can, stand up a test box and do a test run, just to be sure you have the right steps.

Bill