I'm looking into using snapmirror migrate to move data between aggregates. All I can find is that it will shut down NFS and CIFS on the system during the cutover, which does not sound like "no downtime" to me. A coworker says that he has used snapmirror migrate for NFS volumes, and the shutdown happens so quickly it's not noticed.
I've read the man page, the official documentation, KB40272. From what I can gather, the shutdown is just for the final snapmirror update - so theoretically you could minimize this time by doing an update just before doing the migrate.
We have a bunch of volumes across a bunch of aggregates, only some of which are moving. CIFS and NFS are active on almost every volume.
I see lots of posts from people that use it, but everyone seems to gloss over the "shut down NFS and CIFS" bit. Can people that have used this to migrate data comment on impact? I'm especially worried about CIFS being shutdown - I understand that shares to the volumes being migrated will have to reconnect, but if EVERY share to the controller will have to reconnect just because I migrate one volume, it seems a bit harsh.
I've used snapmirror migrate a bunch of times (about a dozen migrates over two nights). It will shut down nfs (I don't know about cifs since I don't use it much) anywhere from less than a second to about 5 seconds (that was the longest one I personally experienced). In most cases, it didn't disrupt the servers (they are similar to nfs lag spikes), but the one that took about 5 seconds did disrupt the servers. We had server admins standing by to fix any disruptions.
We bit the bullet because that was the best way to move the nfs volumes without major disruption.