There is no guarantee on the time to move (for that you need to use Data Motion for vFilers with Protection Manager which uses snapmirror and not between nodes in the same cluster).  SnapMover isn’t often used but since it does a disk reassign in between the vfiler stop and vfiler start on the target, it should be fast without a data migration.  It should be much faster than a vfiler migrate with snapmirror since there is no snapmirror update.

I would run a test if you have a few spare drives with a test aggregate…just to make sure things like networking and any routes in vfiler0 /etc/rc get moved or put them in manually if not moved.  I haven’t run it in a long time so a test would be a good thing to confirm the vfiler comes up on the partner node with all networking available.

From: Kawakubo, Ken [mailto:kkawakub@fhcrc.org]
Sent: Monday, February 06, 2012 1:49 PM
To: Gelb, Scott; toasters@teaparty.net
Subject: RE: anybody has experience using snapmover?

Hi Scott,

Thank you for the pointer. I checked volumes in the vfiler and they are all in the same aggregate. Can you tell me how long snapmover no-copy vfiler migration takes if the vfiler contains 18 volumes totaling about 5TB.

Regards,

Ken Kawakubo

From: Gelb, Scott [mailto:sgelb@insightinvestments.com]
Sent: Monday, February 06, 2012 12:11 PM
To: Kawakubo, Ken; toasters@teaparty.net
Subject: RE: anybody has experience using snapmover?

The key requirement is that the vFiler owns ALL volumes in the aggregate that will be reassigned. If there is a volume from any other vFiler in that aggregate, it will fail.

From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net] On Behalf Of Kawakubo, Ken
Sent: Monday, February 06, 2012 11:25 AM
To: toasters@teaparty.net
Subject: anybody has experience using snapmover?

Dear list,

We have an active-active cluster with vfilers. One vfiler in node-1 is having performance issues. Since node-1’s CPU usage is much higher than node-2. I would like to migrate the vfiler from node-1 to node-2.

I found a feature called snapmover in vfiler manual. I found that the command "vfiler migrate -m nocopy" migrates a vfiler from node-1 to node-2 without actually copying data. This command merely changes the software disk ownership of disks that make up the vfiler and is non-disruptive. Furthermore, snapmover is now a non-license feature. Sounds too good to be true?

I want to know if there is any gotcha using snapmover. Does anybody have experiences using "vfiler migrate -m nocopy" command?

Our cluster is NetApp V3170.

Regards,

Ken Kawakubo