Managing Shared Disks

Planning to use the shared disk components

This section describes things to consider and information that you should know about the IBM Virtual Shared Disk, Hashed Shared Disk, and Recoverable Virtual Shared Disk components before you use them. They are each optional components of PSSP with the following relationships:

• The base components of PSSP (contained in the install image ssp, file sets ssp.basic and ssp.sysctl in particular) are required for any of the subsystems to work.
• The IBM Virtual Shared Disk component can function independently.
• The IBM Virtual Shared Disk component is a prerequisite for the Hashed Shared Disk component.
• The IBM Virtual Shared Disk component and the IBM Recoverable Virtual Shared Disk component are corequisites.
• The RS/6000 Cluster Technology (RSCT) Group Services and Topology Services components of PSSP (contained in file sets rsct.basic and rsct.clients) are prerequisites for the IBM Recoverable Virtual Shared Disk component.
• The ssp.vsdgui image is a prerequisite to use the IBM Virtual Shared Disk Perspective component.

Planning for IBM Virtual Shared Disk

Before you can use the IBM Virtual Shared Disk component of PSSP, you must:

1. Refer to RS/6000 SP: Planning, Volume 2, Control Workstation and Software Environment for related planning considerations.
2. Check IBM Virtual Shared Disk restrictions.
3. Determine which applications are to have access to the virtual shared disks you plan to create. See Tuning Virtual Shared Disk performance and Application programming considerations for Virtual Shared Disks for considerations in selecting applications to exploit your virtual shared disks and planning your virtual shared disk configuration.
4. Consider the size of the applications you will run.
5. Consider how you want to spread data across the virtual shared disk nodes.
6. Plan your volume groups and logical volumes, using AIX System Management Guide: Operating Systems and Devices.
7. Install the IBM Virtual Shared Disk component of PSSP on the control workstation and all nodes that will have or use virtual shared disks (servers and clients).

Planning for Hashed Shared Disk

After planning for IBM Virtual Shared Disk, if you decide to use the Hashed Shared Disk component of PSSP as well, to stripe data across multiple virtual shared disk nodes, you must:

1. Check Hashed Shared Disk restrictions.
2. Read Tuning Hashed Shared Disk performance to understand tuning considerations.
3. Determine where the I/O bottlenecks are on the virtual shared disks.
4. Determine which nodes and disks will be used to eliminate these bottlenecks.
5. Determine the stripe size.
6. Install the Hashed Shared Disk component of PSSP with the IBM Virtual Shared Disk component on all nodes that will have or use virtual shared disks (servers and clients).

Planning for IBM Recoverable Virtual Shared Disk

The IBM Recoverable Virtual Shared Disk component provides recoverability for your virtual shared disks and operates in the same manner whether or not you use the Hashed Shared Disk component. If you want recoverability, in addition to planning for the IBM Virtual Shared Disk component, and the Hashed Shared Disk component, if you choose it, you'll need to consider both primary and secondary nodes for your volume groups and plan your hardware configuration. You must:

1. Check IBM Virtual Shared Disk restrictions and IBM Recoverable Virtual Shared Disk restrictions.
2. Plan your twin-tailed volume groups and decide which twin-tailed disks will be in each group. The IBM Recoverable Virtual Shared Disk generally supports adapters and disk subsystems that have the characteristics in the ODM CuDv class shown below. It is assumed that device reserves are reset in the standard way for the various class-subclass-types and that they only affect the device for which the reserve is being broken.

    parent = *scsi* or                # native scsi, vscsi, fscsi
             ssar*  or                # ssa
             dar*   or                # 7135 disk array
             dpo                      # Data Path Optimizer
    
    
    PdDvLn =
             disk/dpo*   or           # Data Path Optimizer also known \
                                      # as Subsystem Device Driver (SDD)
             disk/fcp*|disk/fdar*  or #fibre channel or fibre disk array
             disk/ssar*  or           # ssa
             disk/dar*)  or           # 7135 disk array
             disk/scsi*  or           # scsi
   

   The following example shows how to determine this information.

    odmget -q"name=hdisk0" CuDv
    
    CuDv:
         name = "hdisk0"
         status = 1
         chgstatus = 2
         ddins = "scdisk"
         location = "04-C0-00-4,0"
         parent = "scsi0"
         connwhere = "4,0"
         PdDvLn = "disk/scsi/scsd"
   

3. Ensure that the supported disks are properly installed and connected to both nodes if they are twin-tailed or concurrently accessed.
4. Each of the currently supported levels of IBM Recoverable Virtual Shared Disk (2.1.1, 3.1, 3.2, and 3.4) can interoperate with each of the supported levels of PSSP (2.4, 3.1, 3.2 and 3.4). Interoperate means that nodes in the same system partition but at different PSSP levels can access each other's virtual shared disks at the lowest functional level.

   However, to get the newest level of recovery function, the control workstation and each node in the system partition that will have or use virtual shared disks must have AIX 4.3.3 or AIX 5L 5.1 and PSSP 3.4, with the IBM Virtual Shared Disk component and the IBM Recoverable Virtual Shared Disk component installed.

   The IBM Recoverable Virtual Shared Disk component uses the RSCT Group Services and Topology Services utilities, so those components of PSSP must also be installed.

Note:

If you plan to migrate disk servers to AIX 5L 5.1 or higher, all the servers that share volume groups should be migrated at the same time.