Managing Shared Disks

Creating or defining Virtual Shared Disks or Hashed Shared Disks

Remember, your procedure is based on whether or not you already have logical volumes. The create actions and commands take care of logical volumes and global volume groups for you. If you already have them, you must do the define steps instead.

If you are using the create actions or commands, it's a good idea to check for old rollback files.

Checking for old rollback files

If you issue a virtual shared disk action or command that operates on multiple nodes, such as createvsd or createhsd, and the command fails, a rollback file will be created so that a second invocation of that command can start at the last successful operation. (A command that is issued against multiple nodes fails if any of the nodes cannot execute it.) If you later change your virtual shared disk configuration or run a different command, createvsd attempts to complete processing using the rollback file and will fail. Be sure there are no rollback files from failed invocations of commands on your system. If there are any, they can be found in /usr/lpp/csd/sysctl/new_rollback and /usr/lpp/csd/vsdfiles/vsd_rollback. You can delete them with the rm command. Old rollback files can interfere with the processing of new invocations of these commands.

Any time a virtual shared disk or hashed shared disk command that operates on multiple nodes fails, check for old rollback files.

Creating Virtual Shared Disks

You can create several virtual shared disks with a single graphical user interface action or a line command (on both primary and secondary nodes if you have the IBM Recoverable Virtual Shared Disk component running). You must first have used the IBM Virtual Shared Disk Perspective or the vsdnode command to set up information in the SDR about each node involved in this virtual shared disk configuration.

Do not perform this task if you have already used the Logical Volume Manager of AIX to establish logical volumes. The create process generates them for you. Instead, see Defining Virtual Shared Disks and Hashed Shared Disks. Also, if you want data striping see Creating Hashed Shared Disks instead.

Creating Virtual Shared Disks with the IBM Virtual Shared Disk Perspective

To create virtual shared disks using the IBM Virtual Shared Disk Perspective graphical user interface, do the following:

Click on the IBM VSDs pane
Click on Actions>Create...

That opens the Create VSDs dialog where you can :

Enter all the pertinent information
- Number of VSDs per node
- VSD name prefix
- Logical volume name prefix
- Volume group Name
- VSD size (MB)
- VSD type (concurrent or nonconcurrent)
- Mirroring count
- Physical partition size (MB)
Select a cache option (cache or no cache)
Select from a list of nodes that have been designated as virtual shared disk nodes, the primary node and the backup node
Select from a list, the physical disks that the virtual shared disk is to span
Use the arrow button to move the node and disk combinations to the Create on Nodes list

The equivalent command is createvsd. See the book PSSP: Command and Technical Reference for syntax.

Information about virtual shared disk definitions is stored in the SDR VSD_Global_Volume_Group object and the VSD_Table object. You can view the information using the IBM Virtual Shared Disk Perspective graphical user interface.

Examples of creating Virtual Shared Disks with createvsd

The following examples range from simple to fairly complex.

To create identical virtual shared disk definitions on each of three nodes in a system partition, type:
```
createvsd -n 1,3,5 -s 4 -g SYS2VG -v SYS2VSD
```
This creates the following virtual shared disk definitions:
- SYS2VSD1n1 on node 1. The local volume group name on node 1 is SYS2VG. The global volume group name is SYS2VGn1. The logical volume is lvSYS2VSD1n1.
- SYS2VSD2n3 on node 3. The local volume group name on node 3 is SYS2VG. The global volume group name is SYS2VGn3. The logical volume is lvSYS2VSD2n3.
- SYS2VSD3n5 on node 5. The local volume group name on node 5 is SYS2VG. The global volume group name is SYS2VGn5. The logical volume is lvSYS2VSD3n5.
No secondary nodes are defined. The space allocated to a virtual shared disk is spread across all the physical disks (hdisks) within its local volume group on each node (1,3, and 5).
To assign each disk in the previous example a secondary node (with the IBM Recoverable Virtual Shared Disk component running), type:
```
createvsd -n 1/2/,3/4/,5/6/ -s 4 -g SYS2VG -v SYS2VSD
```
This creates the following virtual shared disk definitions:
- SYS2VSD1n1 on node 1 with a twin-tailed connection to node 2. The local volume group name on node 1 is SYS2VG. The global volume group name is SYS2VGn1. The logical volume is lvSYS2VSD1n1.
- SYS2VSD2n3 on node 3 with a twin-tailed connection to node 4. The local volume group name on node 3 is SYS2VG. The global volume group name is SYS2VGn3. The logical volume is lvSYS2VSD2n3.
- SYS2VSD3n5 on node 5 with a twin-tailed connection to node 6. The local volume group name on node 5 is SYS2VG. The global volume group name is SYS2VGn5. The logical volume is lvSYS2VSD3n5.
The volume groups in this example are imported to the secondary node.
To create three virtual shared disk definitions on primary and secondary nodes (with the IBM Recoverable Virtual Shared Disk running), where the logical volume created on nodes 3 and 4 spans two disks and the volume group spans three disks, type:
```
createvsd -n 3/4:hdisk1,hdisk2+hdisk3/,5/6/,7/8/ -s 12 -g datavg\
-v USER -x
```
This command creates the following virtual shared disk definitions:
- USER1n3, with logical volume lvUSER1n3 defined on a volume group with the global volume group name datavgn3 on node 3, imported to node 4. The volume group datavgn3 spans hdisk1, hdisk2, and hdisk3. The logical volume lvUSER2n3 spans hdisk1 and hdisk2.
- USER2n5, with logical volume lvUSER2n5 defined on a volume group with the global volume group name datavgn5 on node 5, imported to node 6.
- USER3n7, with logical volume lvUSER3n7 defined on a volume group with the global volume group name datavgn7 on node 7, also imported to node 8.
  The volume groups datavgn5 and datavgn7 are created with one 4MB partition from a single physical disk. The volume group datavgn3 is created with one 4MB partition from the three physical disks hdisk1, hdisk2, and hdisk3.
The local volume group name on each node is datavg. The volume groups are not imported to the secondary node because the -x flag was used in the command.

Creating Hashed Shared Disks

You can create an entire configuration of hashed shared disks and underlying virtual shared disks with a single graphical user interface action or line command (on both primary and secondary nodes if you have the IBM Recoverable Virtual Shared Disk component running).

Creating Hashed Shared Disks with IBM Virtual Shared Disk Perspective

To create hashed shared disks using the IBM Virtual Shared Disk Perspective graphical user interface, do the following:

Click on the IBM HSDs pane
Click on Actions>Create...

That opens the Create HSDs dialog where you can :

Enter all the pertinent information
- Number of VSDs per node
- HSD name prefix
- Logical volume name prefix
- Volume group Name
- VSD type (concurrent or nonconcurrent)
- HSD size (MB)
- Mirroring count
- Physical partition size (MB)
- Stripe size (KB)
Select a cache option (cache or no cache)
Select a protect LVCB option (protect, do not protect)
Select from a list of nodes that have been designated as virtual shared disk nodes, the primary node and the backup node
Select from a list, the physical disks that the virtual shared disk is to span
Use the arrow button to move the node and disk combinations to the Create on Nodes list

The equivalent command is createhsd.

Examples of creating Hashed Shared Disks with createhsd

For the syntax of the createhsd command see the book PSSP: Command and Technical Reference. The following are examples which range from simple to fairly complex.

To create a hashed shared disk that stripes data across three identical virtual shared disk definitions on each of three disks in a system partition, type:
```
createhsd -n 1,3,5 -s 12 -g SYS2VG -t 4 -d SYS2HSD
```
This creates the hashed shared disk definition SYS2HSD and its underlying virtual shared disk definitions:
- SYS2HSD1n1 on node 1. The local volume group name on node 1 is SYS2VG. The global volume group name is SYS2VGn1. The logical volume is lvSYS2HSD1n1.
- SYS2HSD2n3 on node 3. The local volume group name on node 3 is SYS2VG. The global volume group name is SYS2VGn3. The logical volume is lvSYS2HSD2n3.
- SYS2HSD3n5 on node 5. The local volume group name on node 5 is SYS2VG. The global volume group name is SYS2VGn5. The logical volume is lvSYS2HSD3n5.
  The usable hashed shared disk size is 12MB. The stripe size is 4KB. The first stripe on each disk is skipped to allow space for the LVCB.
No secondary node is defined. The space allocated to the hashed shared disk is spread across all the physical disks (hdisks) connected to each node (1, 3, and 5).
With the IBM Recoverable Virtual Shared Disk component running, to create a hashed shared diskdefinition whose underlying virtual shared disks are twin-tailed and have secondary direct client nodes assigned, type:
```
createhsd -n 1/2/,3/4/,5/6/ -s 48 -g SYS2VG -t 8 -d SYS2HSD
```
This adds the backup node number to the naming convention and creates hashed shared disk SYS2HSD with the following virtual shared disk definitions:
- SYS2HSD1n1 on node 1 with a twin-tailed connection to node 2. The local volume group name on node 1 is SYS2VG. The global volume group name is SYS2VGn1. The logical volume is lvSYS2HSD1n1.
- SYS2HSD2n3 on node 3 with a twin-tailed connection to a second direct client on node 4. The local volume group name on node 3 is SYS2VG. The global volume group name is SYS2VGn3. The logical volume is lvSYS2HSD2n3.
- SYS2HSD3n5 on node 5 with a twin-tailed connection to a second direct client on node 6. The local volume group name on node 5 is SYS2VG. The global volume group name is SYS2VGn5. The logical volume is lvSYS2HSD3n5.
  The usable hashed shared disk size is 48MB. The stripe size is 8KB. The first stripe on each disk is skipped to allow space for the LVCB.
With the IBM Recoverable Virtual Shared Disk subsystem running, to create a hashed shared disk with three underlying virtual shared disk definitions on primary and secondary nodes, where the logical volumes and volume groups span two physical disks, type:
```
createhsd -n 3/4:hdisk1,hdisk2/,5/6:hdisk1,hdisk2/,7/8:hdisk1,hdisk2/
          -s 48 -t 12 -S -g datavg -d USER
```
This command creates the hashed shared disk USER, with the following underlying virtual shared disk definitions:
- USER1n3, with logical volume lvUSER1n3 defined on a volume group with the global volume group name datavgn3 on node 3, imported to node 4. The volume group datavgn3 and the logical volume lvUSER2n3 span hdisk1 and hdisk2.
- USER2n5, with logical volume lvUSER2n5 defined on a volume group with the global volume group name datavgn5 on node 5, imported to node 6. The volume group datavgn5 and the logical volume lvUSER2n5 span hdisk1 and hdisk2.
- USER3n7, with logical volume lvUSER3n7 defined on a volume group with the global volume group name datavgn7 on node 7, imported to node 8. The volume group datavgn7 and the logical volume lvUSER2n7 span hdisk1 and hdisk2.
  The usable size of the hashed shared disk is 48MB. The stripe size is 12KB. The first stripe on each disk is not skipped to allow space for the LVCB.

Creating Hashed Shared Disks in a node-pair configuration

If your system partition is configured in a node-pair arrangement like that shown in Figure 16, where sets of physical disks are twin-tailed on nodes 1 and 2, nodes 3 and 4, nodes 5 and 6, and so forth, you will need to use the following process to create a single hashed shared disk for the system partition. In this configuration, odd-numbered nodes are the backups for virtual shared disks defined on even-numbered nodes and even-numbered nodes are the backups for virtual shared disks defined on odd-numbered nodes.

Figure 16. A Node-Pair Configuration

View figure.

To set up the hashed shared disk for this configuration, use createvsd to define one or more virtual shared disks on the odd-numbered nodes, and then to define one or more virtual shared disks on the even-numbered nodes. Virtual shared disks on odd-numbered nodes must be defined separately from those on even-numbered nodes in this configuration if they are to be part of a single Hashed Shared Disk. You then use defhsd to define the Hashed Shared Disk.

For example, to define a Hashed Shared Disk for the configuration in Figure 16, you could use the following sequence of commands:

createvsd 1/2/,3/4/,5/6/,7/8/ -s 8192 -g s21ovg -v ovsd

createvsd 2/1/,4/3/,6/5/,8/7/ -s 8192 -g s21evg -v evsd

defhsd protect_lvcb s21hsd 8192 ovsdn1 ovsdn3 ovsdn5 ovsdn7
evsdn2 evsdn4 evsdn6 evsdn8

Defining Virtual Shared Disks and Hashed Shared Disks

When you already have global volume groups and logical volumes you cannot use the create actions or commands, you must perform a define process instead. First you must define all the virtual shared disks then, if you want data striping, define the hashed shared disks.

To define all the virtual shared disks:

Start the IBM Virtual Shared Disk Perspective by issuing spvsd or by double-clicking on the IBM VSD Perspective icon on the SP Perspectives Launch Pad
Click on the IBM VSDs pane
Click on Actions>Define VSDs...

That opens a dialog window where you can enter the pertinent information:

Logical volume name
Global volume group name
IBM VSD name
Cache option (cache or nocache)

Alternatively, you can use the defvsd command.

To define all the hashed shared disks:

Start the IBM Virtual Shared Disk Perspective by issuing spvsd or by double-clicking on the IBM VSD Perspective icon on the SP Perspectives Launch Pad
Click on the IBM HSDs pane
Click on Actions>Define HSDs...

That opens a dialog window where you can enter the pertinent information:

IBM VSD name
Protect LVCB option (protect or do not protect)
Stripe size (KB)
Underlying IBM VSD names

Alternatively, you can use the defhsd command.

[ Top of Page | Previous Page | Next Page | Table of Contents | Index ]