NDMPCOPY on NetApp 8.2.x Cluster Mode

So, I have a customer who I needs to move 3 “boot from iscsi san” luns from a Netapp OnTap 8.2 Cluster Mode vServer shared with regular iscsi luns, to their own dedicated vserver. How do you easily move between vservers? NDMPCOPY!

Why do they need to do this? Well their boot from iscsi is 1gb, and they have multiple 10gb, and Windows is being a stupid idiot and the iscsi initiator isn’t routing it right.

It’s a big ordeal with a managed service provider/datacenter .

I digress, simple fix, move the boot luns to their own vServer. k thx bye.

Not so fast. How will you move them? This customer cant have a complete outage of their 3 node SQL Cluster. If we could take an extended full outage, we could snapmirror and just bring down the old vol, break, and bring up the new volume, and remap the luns.

We could snapmirror to 3 destinations if SM is licensed. That leaves alot of cruft around though.

Here’s what I’m thinking.

Assume we have an old vserver (s0-fas3220win-iscsi1) with an old volume (boot_s0_pwdb_vol) with 3 luns.

What would be possible, is:

  • * Make new vserver s0-fas3220-win-boot-iscsi1
  • * Make 3 volumes for the luns. (ndmptarget1 – 3 in this example)
  • * Copy initiator group settings
  • * Move MS SQL cluster resources from Box 1 to other servers.
  • * POWER OFF Server
  • * Offline original lun boot-s0-pwdb03.lun
  • * NDMPCOPY original volume in original vserver to new vserver and new volume ndmptarget1

  • * Delete 2 unneeded luns from new volume ( boot-s0-pwdb04.lun & boot-s0-pwdb05.lun )
  • * Bring wanted lun online and map it.
  • * Resize volume down a bit
  • * Change server’s network port for 1gb vlan (This is getting a new vlan entirely for this vserver to fully isolate it)
  • * Change servers target ip to the new vserver lif IP
  • * Bring up server and test
  • * Repeat for the other luns to new volumes

    No flex cloning or lun cloning needed in this method.

    NDMP/ndmpcopy doesn’t work the same in 8.2 cDot mode. It actually has 2 different modes it can work:

  • * Cluster level/vserver level – sort of like the old way, but with /vservername/ as the base of the paths.
  • * “node-scope-mode” – Node-scope-mode is for compatibility with certain backup applications like TSM, and will cause all sorts of problem with easy ndmpcopy clusterip:/vserver/volume pathing.

    node-scope-mode makes the following bad mojo happen:

  • All NDMP operations are restricted to resources on the node
  • Vserver-aware NDMP commands are disabled
  • NDMP authentication falls back to DATA ONTAP 8.1 NDMP authentication scheme
  • So, I have lun /vol/boot_s0_pwdb_vol/boot-s0-db03.lun on s0-fas3220win-iscsi1, and I want it to be /vol/ndmptarget1/boot-s0-db03.lun on so-fas3220-win-boot-iscsi1

    Verify you have intercluster lifs. Why? You always should… this is how a node in the cluster talks and sends crap to another node in the cluster (SnapMirror, SnapVault, etc need this)

    There it is!

    Be Sociable, Share!
  • , , , , , ,

    Comments are closed.