Forum Discussion

Karim_Raslan's avatar
11 years ago

Migrating to new Array

Hi,

We will be doing a Storage Migration from EMC Clariion to EMC VNX. Our vendor is suggesting to use MirrorView to block replicate the LUN from CX to VNX.

Once they have completed the LUN replication, they will remove the CX LUN and present the new VNX LUN to the Cluster Server.

Below are the planned steps:

 

1. Sync the LUN

2. Stop Cluster services

3. Bring down all the nodes

4. Break the replication

5. Remove old LUN and present the new LUN to the hosts

6. Bring up only 1 node

7. Start cluster services

8. Bring Up the rest of the nodes.

 

Is there anything missed? do we need to deport and import the DG?

  • I assume mirrorview works something like

    Hardisk1 is on Cx

    Hardisk 11 is on Vnx

    You mirror LUN Cx to Vnx so that Harddisk11 is then a copy of Harddisk1, and then you remove Harddisk1 and present Harddisk11 to the server.  This being the case you need to replace step 5 with:

    • Deport diskgroup
    • Remove old LUNS
    • Scan disks in VEA so that VEA sees old LUNS are gone
    • Present new LUNs
    • Scan disks in VEA so that VEA sees new LUNS
    • Import diskgroup

    You can migrate to new arrary completely online if you use SFW mirroring, i.e, you can leave your cluster running while you break mirror - the process being:
     

    1. Add the new LUNs from new array to the diskgroup
    2. Add a mirror in VEA (or whatever tool you normally use to manage SFW) to the volume(s) on the old LUN and select new LUN on new array - repeat for all volumes on all LUNs
    3. When mirror has finishing syncing, remove mirror of volume on original LUN - this is fine to do online
    4. When all old LUNs are empty, you can remove LUNs from the diskgroup.

    I believe you can do this in VOM also which automates the steps for you

    The process is the same concept as for SF on UNIX, - see posts:

    https://www-secure.symantec.com/connect/forums/add-disk-vcs-shared-disk-group

    https://www-secure.symantec.com/connect/forums/array-migration-using-vxvm-solaris-vcs-cluster

    https://www-secure.symantec.com/connect/videos/storage-provisioning-and-array-migration-vom-sf

     

    Mike

  • Hi Karim,

    If you have the SFW-HA product, then this entire operation can be done online with no downtime to the application using the mirror feature of SFW.  We have customers do this all the time.

     

    There is also no need to deport the diskgroup when removing the old luns.  You will need to break the mirror and then remove the old luns from the disk group.  VCS and your application will continue working with the new mirror plex on the new array without skipping a beat.

     

    If you are not completely comfortable with the process, please open a case with Symantec Technical Support.  We will be happy to answer any questions that you have ahead of time so that your array migation occures without application interuptions.

     

    thank you,

    Wally

8 Replies

  • I assume mirrorview works something like

    Hardisk1 is on Cx

    Hardisk 11 is on Vnx

    You mirror LUN Cx to Vnx so that Harddisk11 is then a copy of Harddisk1, and then you remove Harddisk1 and present Harddisk11 to the server.  This being the case you need to replace step 5 with:

    • Deport diskgroup
    • Remove old LUNS
    • Scan disks in VEA so that VEA sees old LUNS are gone
    • Present new LUNs
    • Scan disks in VEA so that VEA sees new LUNS
    • Import diskgroup

    You can migrate to new arrary completely online if you use SFW mirroring, i.e, you can leave your cluster running while you break mirror - the process being:
     

    1. Add the new LUNs from new array to the diskgroup
    2. Add a mirror in VEA (or whatever tool you normally use to manage SFW) to the volume(s) on the old LUN and select new LUN on new array - repeat for all volumes on all LUNs
    3. When mirror has finishing syncing, remove mirror of volume on original LUN - this is fine to do online
    4. When all old LUNs are empty, you can remove LUNs from the diskgroup.

    I believe you can do this in VOM also which automates the steps for you

    The process is the same concept as for SF on UNIX, - see posts:

    https://www-secure.symantec.com/connect/forums/add-disk-vcs-shared-disk-group

    https://www-secure.symantec.com/connect/forums/array-migration-using-vxvm-solaris-vcs-cluster

    https://www-secure.symantec.com/connect/videos/storage-provisioning-and-array-migration-vom-sf

     

    Mike

  • Hi Mike,

    Thanks for the reply.

    With this we dont need to shutdown/reboot the nodes before presenting the new LUNs?

     

    regards,

    Karim

  • No need for reboot if the O/S can see LUNS after presenting them.

    Slight correction on last post in that stopping Cluster services should deport diskgroups and starting Cluster services should import diskgroups, so just need to check that all diskgroups are deported before removing LUNs.

    Mike

  • To add to the comments above. Starting from SF 6.0.1 , you can add/delete luns dynamically using Dynamic reconfiguration tool. Please find the details here.

    https://www-secure.symantec.com/connect/articles/dynamic-reconfiguration-tool
  • Hi Karim,

    If you have the SFW-HA product, then this entire operation can be done online with no downtime to the application using the mirror feature of SFW.  We have customers do this all the time.

     

    There is also no need to deport the diskgroup when removing the old luns.  You will need to break the mirror and then remove the old luns from the disk group.  VCS and your application will continue working with the new mirror plex on the new array without skipping a beat.

     

    If you are not completely comfortable with the process, please open a case with Symantec Technical Support.  We will be happy to answer any questions that you have ahead of time so that your array migation occures without application interuptions.

     

    thank you,

    Wally

  • I agree, strongly, with Wally.  If you have VxFS on the hosts, then, this entire operation can be done online without impacting the cluster or the application.  No downtime at all.  No need for MirrorView. 

    Using VOM, the entire process can also be highly automated.   It is also possible to cleanup storage, as by default, only used blocks will be moved, making it possible to take advantage of any thin-provisioning supported by the new array.

     

    Cheers