6 Replies Latest reply: Jan 25, 2014 10:45 PM by rohang RSS

aix lvm volume group migrations

john121

we are migrating from IBM array to EMC VMAX array. The host OS is AIX  and planning to use AIX LVM. We want to do migrations based on volume groups

i.e if a volume group has 3 devices 20gb 30gb 40gb then we will assign a 90gb vmax lun and move the volume group to the new VMAX lun.

I want to know the AIX LVM commands involved in this process of migrating vg's.

  • 1. Re: aix lvm volume group migrations
    dynamox

    John,

     

    are your system admins not familiar with AIX LVM mirroring ? It's very simple.

     

    Regarding you device sizes, have you establish a "building block" on you array. For example we have a building block of 17G, meaning all of your symdev are 17G in size. Whenever an application needs let's say 68G we can create either a 4 x 17G meta or 2 x 34G metas. For Unix/Linux systems try to come up with some kind of standard, each device presented to the system has an OS I/O path, so if you were to present one huge 90G meta vs 3 x 30G metas ..it matters because in the latter case your system has 3 I/O paths in the OS so it's more efficient. Your system admin add these devices in the volume group and gets the capacity they need either way ..but from performance perspective it's better to have more smaller devices. For example for our Oracle RAC boxes we have established a guideline where we increase their storage in 34G increments.  So if we get a request for 100G , we provision 3 x 34G Metas (each meta is comprised of 2 x 17G symdevs). Yes it's not exactly 100G but you establish a standard instead of building custom meta/devices (like you do on Clariion). In my opinion it's just so much easier to manage, configure for local and remote replication.  My 2 cents

  • 2. Re: aix lvm volume group migrations
    john121

    Dynamox,

    you are right, we are implementing Virtual Provisioning in our environment and we do have standard thin devices curved in the thinpool. ex:- hyper of 16,32,64,128&240gb standard sizes. We decided not to use metas for AIX volume groups. To match the volume group size we will select the appropriate hyper luns in multiples. AIX admin does know the procedure but I want know the procedure and would want to know if he is doing the right way. We are migrating VIO Lpars and also some physical to logical migrations.

  • 3. Re: aix lvm volume group migrations
    dynamox

    Using Smit

     

    1) Go to Volume Groups and select "Set Characteristics of a Volume Group
    2) select "Add a Physical Volume to a Volume Group"
    3) select VOLUME GROUP name and then PHYSICAL VOLUME name (these should be your new hdiskpower devices
    4) At this point new hdiskpower devices are part of the volume group
    5) In Smit select "Mirror a Volume Group"
    6) Select volume group you are mirroring and hdiskpower devices you added earlier
    7) When mirroring is done you will need to "un-mirror" old disks
    8) In smit select "Unmirror a VOlume Group"
    9) select volume and then pick old disks
    10)next step is to remove old disks from the volume group
    11) In Smit select "Set Characteristics of a Volume Group
    12) In smit select "Remove a Physical Volume from a Volume Group"
    13) select volume group and old disks
    14) run lspv and old disks should have "None" in the column.
    15) final step is to remove devices from AIX perspective (rmdev -dl hdisk2)

     

    by the way, you might want to still consider using striped metas as they may provide better performance in certain workloads. There have been a couple of threads discussing that.

  • 4. Re: aix lvm volume group migrations
    john121

    Will it be the same for VIO clients too? I mean the VIO client only has hdisks right? PP is installed only on VIO servers.

  • 5. Re: aix lvm volume group migrations
    dynamox

    i don't have experience with VIO ..sorry.

  • 6. Re: aix lvm volume group migrations
    rohang

    pretty old thread that I've stumbled upon ..

    anyways.

    shouldn't be any different for the VIO clients, john.

    allocate your EMC LUNs to the VIO servers, make them visible to your VIO clients and do a mirrorvg to mirror your VGs to the EMC LUNs .. verify and once done, unmirror your VGs, etc. [you could use SMIT like dynamox mentioned]

    yes, the VIO clients do not have powerpath installed because your EMC LUNs are masked to the physical HBA's on the VIO server (run an lspath instead to see the LUNs coming from the VIO servers) .. on the VIO client you'd still see the VG information across your PV's in a lspv, right? So, you do not have to touch powerpath anywhere, when you do LVM migrations. Just make sure you're mirroring to the right set of disks, whether or not, they are under powerpath control.

     

    thanks .