Forum Discussion

vxvmnewbie1's avatar
vxvmnewbie1
Level 2
14 years ago
Solved

Using VSFHA for Windows for array migration

Hi,

I'm relativley new to VSFHA for Windows (v5.1) and VxVM and have been asked to migrate a 2 node cluster that has a node in each of the companies data centres from a pair of old EVAs to a different pair of EMC arrays due to the EVAs going out of support.

The config has three volumes: regrep, data1 and data2.

  • regrep has 2 x 1GB LUNs from each of the EVAs
  • data1 is concat and has 2 x 500GB LUNs from each of the EVAs
  • data2 is concat and has 3 x 100GB LUNs from each of the EVAs

How can I use VxVM to migrate to the EMC arrays? I also want to resize the volumes as part of the migration so data1 volume is 1TB at the moment and I want it to be 1.2TB, and data2 is 300GB which I want 500GB.

I have configured the LUNs on the CX and presented them out to each node in each data centre and have written signatures to the new disks in preparation.

From what I've read we should be able to create and assocaite a new plex with the volumes and mirror them up then drop the old plexs from the EVAs?

Thanks in advance.

Best Practice
business continuity
Storage Foundation for Windows
Tip-How to
  • mikebounds's avatar
    mikebounds
    14 years ago

    Malc,

    Point 1: Yes, as long as the array is supported by SF - you should check the HCL at https://sort.symantec.com/documents/doc_details/sfha/5.1%20SP2/Windows/CompatibilityLists/

    Point 2: You are correct and you can use GUI or CLI.

    Point six.  DRLs are used for a quick resync when a server crashes - i.e a DRL keeps track of inflight writes so that when a server returns after a crash it can just sync the writes that were happening when the server crashes and not the whole volume.  DCOs are used to resync when an array crashes - i.e a DCO keeps tracks of writes when an array is down so that when array returns you only need to resync writes that occured while array was down.  You can add a DCO at any stage, but you need to allow space for it - they only occupy a few MB and you need one per plex (so if a plex is across 3 LUNS you only need a few MB on one of the 3 LUNs) and you need enterprise license to enable fastresync (DRL only requires standard license)

    So in Windows I believe you need a DCO AND a DRL (where as in SF for UNIX, you only need a DCO as the DCO also performs DRL functionality as from about version 5.0)

    Mike

  • mikebounds's avatar
    mikebounds
    Level 6
    14 years ago

    You are correct so procedure is:

    1. Add LUNS from new array of the size you require to your existing diskgroup
    2. Mirror volumes to new LUNS (this will be 2 mirrors so that you now have a total of 4 copies)
    3. Once the resync of the mirror is complete then remove plexes from old EVA's
    4. If there are any DRLs on old EVAs, then you can move these to new arays (I think you may be able to drag and drop in VEA), alernatively, remove DRLs and add new DRLs on new array.
    5. Remove EVA LUNs from diskgroup
    6. Extend volumes to your required size (I would recommend you do not extend to the full size of the LUN, but leave a couple of MB free so that you can add DCOs in the future.  DCOs will allow fast resync of mirrors if an array goes down and comes back and this feature requires Enterprise License)

    All of this can be done online while your application is still running.  Note you do not need to have the same number of LUNs on the new array, so for example for data1 you could use 3 x 400GB LUNs, so the initial mirror would probably put 400GB on 2 LUNS and 200GB on the 3rd LUN which would be extended in step 6.

    You should also consider using a striped mirror (RAID10) to improve I/O performance.

    Mike

  • vxvmnewbie1's avatar
    vxvmnewbie1
    Level 2
    14 years ago

    Thanks for this - I'm awaiting my manager to sign off on some test Windows kit so I can run through this procedure not on the production boxes and would like to ask a few more newbie questions as I've no test lab to learn this stuff on at the moment....

    • Point one: Will the new arrays come in as new enclosures? Mirror accross enclosure is the existing configuration on the EVA.
    • Point two: My understanding having done some reading is that this mirroring process in VSF is basically setting up two new plexes, associating them with the volume (as they will have a different sub-disk construct than the source plexes on the old arrays the aim is to mirror the plex and not the sub-disks) and sync the new plexes and then disassociate and remove the old plexes from the volumes. Is this correct and can this be done in the GUI or will we need to use CLI?
    • Point six: What are the main differences in DRL and DCO and should we add a DCO now when we're moving to larger capacities on the target arrays?

    Thanks, Malc

  • mikebounds's avatar
    mikebounds
    Level 6
    14 years ago

    Malc,

    Point 1: Yes, as long as the array is supported by SF - you should check the HCL at https://sort.symantec.com/documents/doc_details/sfha/5.1%20SP2/Windows/CompatibilityLists/

    Point 2: You are correct and you can use GUI or CLI.

    Point six.  DRLs are used for a quick resync when a server crashes - i.e a DRL keeps track of inflight writes so that when a server returns after a crash it can just sync the writes that were happening when the server crashes and not the whole volume.  DCOs are used to resync when an array crashes - i.e a DCO keeps tracks of writes when an array is down so that when array returns you only need to resync writes that occured while array was down.  You can add a DCO at any stage, but you need to allow space for it - they only occupy a few MB and you need one per plex (so if a plex is across 3 LUNS you only need a few MB on one of the 3 LUNs) and you need enterprise license to enable fastresync (DRL only requires standard license)

    So in Windows I believe you need a DCO AND a DRL (where as in SF for UNIX, you only need a DCO as the DCO also performs DRL functionality as from about version 5.0)

    Mike