PRIMECLUSTER Installation and Administration Guide 4.1 (for Solaris(TM) Operating System)

6.6.1.3 Creating Gds Resources

Set up resources for the disk classes that are defined in Global Disk Services (GDS).

The disk classes must be created before this screen is used.

If a disk class is registered to a Gds resource when equivalent volume copy is being executed, the equivalent copy operation is canceled. Therefore, the data becomes invalid in some of the slices. However, recovery is not necessary because equivalent copy is executed automatically when the volume is started. If you want to recover the slice state before the volume is started, complete Gds resource setup and then start equivalent copy. For instructions on starting equivalent copy, see "5.3.4 Copying Operation" or "D.10 sdxcopy - Synchronization copying operation" in the "PRIMECLUSTER Global Disk Services Configuration and Administration Guide."

For instructions on creating disk classes, see "Setting Up Shared Disks."

[Flow of Gds resource creation]

For information on the above operations up to "Select SysNode," see "Creating Cmdline Resources." This section describes the operations starting from "Select disk class."

Selecting the disk class

Select the disk class.

[Select disk class]

Available Disk Class

Names of disk classes that can be set.

The disk classes (shared disks) that are defined in GDS are displayed.

Selected Disk Class

Names of disk classes to be set.

Select disk classes from Available disk class, and then click Add. To add all listed disk classes, click Add all. To delete a disk class to be set, select the disk class to be deleted from Selected disk class, and then click Remove. To delete all listed disk classes, click Remove all.

After completing the setup, click Next to go to the "Set Disk Class Attribute" screen.

Disk class attribute setup

Set the disk class attribute.

[Set up disk class attribute]

Disk Class

The disk classes (common disks) that were defined in GDS (described earlier) are displayed.

Flag

Outline

MONITORONLY

This flag determines whether disk class failures are to be reported to the userApplication. If "Yes" is set and a disk class failure occurs, the disk class is switched to faulted state but the Gds resources remain online, and userApplication failover does not occur. Carry out the preliminary design, including the higher applications" before determining the disk classes for which this flag is to be set to "Yes." The default value is "No."

The MONITORONLY attribute must be set to "No" for at least one disk class.

This will prevent userApplication failover in the event of all disk class failures and RMS' notifying the upper applications of the failures.

After completing the setup, click Next to go to the "Set up disk class use" screen.

Setting the disk class use

Configure the use of the disk class.

[Set up disk class use]

Exclusive use

Select "Yes" or "No" for Exclusive use. If you select "Yes", also select "Yes" or "No" for HotStandby operation.

Specify the options according to the use of the shared disk.
After completing the setup, click Next to go to the "Confirm Registration Information" screen.

Shared disk uses

[Shared disk uses and setting methods]

Use	Exclusive use	HotStandby operation
Switched disk	Yes	No
Shared disk for simultaneous access	Yes	Yes
Simultaneous shared disk	No	-

Shared disks have the following features for each use:

• Switched disk

  Only the OPERATING node can use the shared disk.
  The non-OPERATING nodes cannot access the volume.

• Shared disk for simultaneous access

  All nodes can use the shared disk simultaneously.
  Select this use if applications that require disk access from a node other than the OPERATING node are set inside the same userApplication.
  The exclusive control for maintaining data integrity in the shared disk must be handled by the applications, and not by PRIMECLUSTER.

• Simultaneous shared disk

  Select this use if multiple userApplications share the disk classes. An example is when Oracle real application clusters are used.
  The exclusive control for maintaining data integrity in the shared disk must be handled by the applications, and not by PRIMECLUSTER.

• During operation of a cluster application, try not to have applications or user processes that are not related to the cluster applications use the shared disks. If this occurs, the state transition might fail.
• Do not create a home directory for user accounts on a switchover disk. The reasons are as follows:
  • The file system on the switching disk is mounted only on the OPERATING node of the cluster application.
  • PRIMECLUSTER forcibly terminates process that uses the file system on the switching disk by executing the "kill(1M)" command to unmount the file system in the event of cluster application failover.
  • If the file system cannot be unmounted properly because of a failure of forced termination of process, PRIMECLUSTER might forcibly stop the node to prevent the file system data on the shared disk from being destroyed by double mounting.

Checking registration information for Gds resources

Check the registration information for the Gds resources. You can also set resource attributes by selecting the Attributes tab and switching the screen.

For information on the resource attributes, see "Attributes."

[Confirm registration]

SubApplication button

Use this button to associate a previously created Gds resource under the current Gds resource. This button can be selected only if there are resources of the same type that can be associated. For setting instructions, see "Resource association" in "Creating Cmdline Resources."

Check the registration information, and then click Registration.