UNSOLVED

prashant_shah

updated

14 years ago

P

prashant_shah

1 Rookie

•

37 Posts

0

2848

December 13th, 2012 10:00

Celerra migration - NFS and VDM question

I'll be attempting a migration soon for a customer with NS960 as source and VG2 as target.  Customer has some multiprotocol filesystems that are used by CIFS and NFS simultaneously.  They have NIS configured.  Everything on source is on physical data mover but on the target, I'm thinking to create a VDM so that future migrations are easier.  The plan is to configure a VDM on the target and then replicate everything under that VDM since with the new code, NFS is supported for VDM.

Are there any issues I should watch out for?  I looked at some Primus cases and it looks like all NFS exports will have to be done at CLI with the VDM config.  GUI is not supported.  I'm looking for some info on any other gotchas I should look for.  I could just keep everything as is and not worry about the VDM complexity, but I'd like to make easy in the future.

  • Rainer_EMC

    6 Operator

    •

    8645 Posts

    1006

    0

    Posted December 17th, 2012 11:00

    Move the CIFS servers to VDM on the source before doing the migration and save yourself some work

  • umichklewis

    4 Apprentice

    •

    1190 Posts

    1006

    0

    Posted December 18th, 2012 13:00

    Agreed.  To elaborate, Rainer recommends moving the CIFS server on the physical DM to a VDM on the NS960, before migrating it.  By moving the CIFS servers to a VDM first, all you need to do is replicate that VDM to get the CIFS information.

    On the NFS side, your exports point to the specific filesystems they're mounted on.  When you mount a filesystem on a VDM, the root becomes "/root_vdm_<#>/", so if VDM #1 has fs02 is mounted on it, the NFS export path is now starting at "/root_vdm_1/fs02".  This is probably the biggest gotcha you'll see.

  • prashant_shah

    1 Rookie

    •

    37 Posts

    1006

    0

    Posted December 18th, 2012 14:00

    Thanks Rainer and Karl.  Yeah I've done that in the past where I've moved the CIFS server to a VDM and then the associated file systems.  However, in this case where there is multiprotocol involved, I was thinking that by moving a files system that has NFS exports on it under a VDM will break it since that's only recently supported under DART 7.  Maybe I'm getting mixed up and it's not that way.

  • Rainer_EMC

    6 Operator

    •

    8645 Posts

    1006

    0

    Posted December 19th, 2012 03:00

    No – even before the latest code you could NFS export a VDM mounted file system.

    You just couldn’t do it via the GUI and had to use a path from the physical DM like /root_vdm_x/fs

    With the NFS VDM support the path changes to being relative to the VDM root like /fs

    Also if its exported on the VDM the NFS share definition is kept in the VDM and replicated with it like for CIFS

  • prashant_shah

    1 Rookie

    •

    37 Posts

    1007

    0

    Posted December 19th, 2012 11:00

    Cool.  Thanks.  I will try to go that route.

  • Clustor

    57 Posts

    1007

    0

    Posted March 31st, 2013 07:00

    How did the migration go eventually? I'm in the process of doing the exact same thing moving from NS20 to VNX5300.

    So you didn't have any issues when moving to a VDM first at the source although NFS was used ?

    Would these be the steps?

    1.) move everything to a VDM on source

    2.) replicate filesystems and VDM to target

    3.) switchover filesystems

    4.) switchover VDM

    I'm still a bit unsure about what happens to my NFS exports when performing step 1

  • Rainer_EMC

    6 Operator

    •

    8645 Posts

    1007

    0

    Posted March 31st, 2013 12:00

    Moving a fs from the DM to a VDM will of course change the mountpoint and therefore the path in the export

    I would would suggest to test with a test fs or simulator till you are comfortable before changing production file systems

  • prashant_shah

    1 Rookie

    •

    37 Posts

    1007

    0

    Posted March 31st, 2013 20:00

    As Rainer explained, the mount points will be different so you'll have to redo the exports. 

    In my case, it worked perfectly and we we about to call it a success until I realized that there was one global share (a CIFS share presented through 3 servers) and so we had to back out.  I ended up migrating the hard way by doing sharedup/lgdup with a swing server.

  • Rainer_EMC

    6 Operator

    •

    8645 Posts

    1007

    0

    Posted April 1st, 2013 04:00

    Well, there are ways to make the export path the same on the dst but that requires more work and understanding

    Unless the client change is too painful I would just change the client mount