Solaris Volume Manager Administration Guide

Chapter 4 Solaris Volume Manager for Sun Cluster (Overview)

This chapter provides an overview of Solaris Volume Manager for Sun Cluster.

This chapter includes the following information:

Introduction to Solaris Volume Manager for Sun Cluster

Starting with the Solaris 9 9/04 release, Solaris Volume Manager can manage storage in a Sun Cluster environment using multi-owner disk sets. Multi-owner disk sets allow multiple nodes to share ownership of a disk set and to simultaneously write to the shared disks. Previously, shared disk sets were visible from all participating hosts in the disk set, but only one host could access it at a time. Multi-owner disk sets work with Sun Cluster and with applications such as Oracle Real Application Clusters.

Multi-owner disk sets and Solaris Volume Manager shared disk sets can coexist on the same node. However, moving disk sets between the two configurations is not supported.


Note –

Solaris Volume Manager for Sun Cluster device id support for multi-owner disk sets is not available. Therefore, importing multi-owner disk sets from one system to another is not supported at this time.


Solaris Volume Manager for Sun Cluster creates the same components that you can create with Solaris Volume Manager, including stripes, concatenations, mirrors, soft partitions, and hot spares. Solaris Volume Manager for Sun Cluster does not support RAID-5 volumes and transactional volumes.

The following figure shows the association between the software and the shared storage in a typical cluster configuration.

Figure 4–1 Sample Cluster Configuration

The diagram titled Sample Cluster Configuration shows
the association between the software and the shared storage in a typical cluster
configuration.

Each node has local storage as well as at least one path to shared storage. The multi-owner disk sets in the cluster are managed by Solaris Volume Manager for Sun Cluster, which is part of the Solaris Operating System (Solaris OS).

Prerequisite: Required Software Components for Multi-Owner Disk Set Functionality

To use Solaris Volume Manager for Sun Cluster, the following software must be installed in addition to the Solaris OS:


Note –

For information on setting up Sun Cluster and Oracle Real Application Clusters software, see Sun Cluster Software Installation Guide for Solaris OS. and Sun Cluster Data Service for Oracle Real Application Clusters Guide for Solaris OS.


Multi-Owner Disk Set Concepts

The storage managed by Solaris Volume Manager for Sun Cluster is grouped into multi-owner disk sets. Multi-owner disk sets allow multiple nodes to share ownership of a disk set and to simultaneously write to the shared disks. An instance of an application such as Oracle Real Application Clusters runs on each node in the cluster, so multi-owner disk sets provide scalability. Since each instance of the application directly accesses the shared storage, multi-owner disk sets also enhance the performance of the application.


Note –

Multi-owner disk set functionality is enabled only in a Sun Cluster environment. Nodes are the physical machines that are part of a Sun Cluster system.


Each multi-owner disk set is associated with a list of nodes. These nodes share ownership of the disk set. The following metaset -s disk-set command shows the output for a multi-owner disk set.


# metaset -s blue

Multi-owner Set name = blue, Set number = 1, Master = nodeone

Host                Owner          Member
  nodeone           multi-owner   Yes 
  nodetwo           multi-owner   Yes 

Drive    Dbase

d9       Yes  

d13      Yes  

This output shows nodeone and nodetwo in the list of nodes that share ownership of the disk set. Additionally, nodeone is designated as the master node.

Each multi-owner disk set has a master node. After a disk set is created, the node that adds the first disk to the disk set becomes the master node of the disk set. The master node creates, deletes, and updates the state database replicas in the disk set.


Note –

For more information on state database replicas, see Chapter 6, State Database (Overview).


Solaris Volume Manager for Sun Cluster can support disk sets with different, yet overlapping, node lists. Because each disk set has a master node, multiple masters can exist simultaneously on the same cluster.

The following output from the metaset command shows that nodeone becomes the master node when the first disk is added to the disk set.


nodeone# metaset -s red
Multi-owner Set name = red, Set number = 1, Master = 

Host                Owner          Member
  nodeone                           Yes 
  nodetwo                           Yes 
nodeone# metaset -s red -a /dev/did/dsk/d9
nodeone# metaset -s red

Multi-owner Set name = red, Set number = 1, Master = nodeone 

Host                Owner          Member
  nodeone           multi-owner     Yes 
  nodetwo           multi-owner     Yes 

Drive     Dbase

d9        Yes  

Solaris Volume Manager for Sun Cluster can support disk sets with different, yet overlapping, node lists. Because each disk set has a master node, multiple masters can exist simultaneously on the same cluster.

Tasks Associated With Multi-Owner Disk Sets


Caution – Caution –

Before configuring multi-owner disk sets, you must have the following software installed, in addition to the Solaris OS:

For information on setting up Sun Cluster and Oracle Real Application Clusters software, see Sun Cluster Software Installation Guide for Solaris OS. and Sun Cluster Data Service for Oracle Real Application Clusters Guide for Solaris OS.


Solaris Volume Manager for Sun Cluster generally uses the same set of Solaris Volume Manager commands to perform tasks associated with disk sets. Some command options unique to multi-owner disk sets have been added to the metaset command. For example, the task to create a multi-owner disk set requires the -M to the metaset command. The following output shows you how to create a multi-owner disk set using the metaset -s diskset-name -a -M -h hostname command.


# metaset -s red -a -M -h nodeone
# metaset
Multi-owner Set name = red, Set number = 1, Master = 

Host                Owner          Member
  nodeone                          Yes 

In addition, some of the metaset command options, such as the commands to take and release disk sets, are not used with multi-owner disk sets. For more information, see the metaset(1M) man page.

Another example of how tasks differ in a Sun Cluster environment occurs when working with disks. Sun Cluster assigns each disk a unique device ID (DID) number. Rather than using the cntndn format to identify a disk, use the Sun Cluster DID path name, /dev/did/dsk/dN. The variable N is the device number assigned by Sun Cluster.

The following output shows you how to add a disk to a multi-owner disk set using the metaset -s diskset-name -a disk-name command and the Sun Cluster DID path name to identify the disk.


nodeone# metaset -s red
Multi-owner Set name = red
Multi-owner Set name = red, Set number = 1, Master = 

Host                Owner          Member
  nodeone                           Yes 
  nodetwo                           Yes 
nodeone# metaset -s red -a /dev/did/dsk/d13
nodeone# metaset -s red
Multi-owner Set name = red, Set number = 1, Master = nodeone

Host                Owner          Member
  nodeone            multi-owner   Yes

Drive Dbase

d13   Yes

For information on creating multi-owner disk sets for the Oracle Real Application Clusters, see Creating a Multi-Owner Disk Set in Solaris Volume Manager for Sun Cluster for the Oracle Real Application Clusters Database in Sun Cluster Data Service for Oracle Real Application Clusters Guide for Solaris OS.

For tasks that are associated with disk sets, see Chapter 19, Disk Sets (Tasks).

Solaris Volume Manager for Sun Cluster Configuration

Solaris Volume Manager for Sun Cluster supports the following configuration:

Many of the extension properties for Sun Cluster Support for Oracle Real Application Clusters specify timeouts for steps in reconfiguration processes. For further information about setting timeouts, refer to Tuning Sun Cluster Support for Oracle Real Application Clusters in Sun Cluster Data Service for Oracle Real Application Clusters Guide for Solaris OS.

RAID–1 (Mirror) Volumes in Multi-Owner Disk Sets

A RAID–1 volume, or mirror, created in a multi-owner disk set functions identically to a RAID-1 volume in a Solaris Volume Manager shared disk set. However, RAID-1 volumes in multi-owner disk sets have some additional features.

Mirror Ownership With Multi-Owner Disk Sets

The concept of mirror ownership is unique to multi-owner disk sets. Unlike a RAID-1 volume in a Solaris Volume Manager shared disk set, a RAID-1 volume in a multi-owner disk set usually has an owner associated with it. The ownership of the mirror volume is chosen by the volume manager. The owner of the volume is one of the nodes designated in the node list for the disk set. Only the owner of the RAID-1 volume can write to the volume. If a non-owner node wants to write to the volume, the ownership switches to the node doing the write operation. The following output from the metastat -s diskset-name command shows nodeone as the owner of the RAID-1 volume, d24.


# metastat -s red
red/d24: Mirror
    Submirror 0: red/d20
      State: Okay
    Submirror 1: red/d21
      State: Okay
    Pass: 1
    Read option: roundrobin (default)
    Write option: parallel (default)
    Resync option: optimizedresync
    Owner: nodeone
    Size: 825930 blocks (403 MB)

Data Management and Recovery Processes

As with RAID-1 volumes in Solaris Volume Manager, RAID-1 volumes in Solaris Volume Manager for Sun Cluster perform operations to ensure consistent data. Solaris Volume Manager for Sun Cluster provides RAID-1 volumes with two options for data management and recovery.

Optimized Resynchronization for Solaris Volume Manager for Sun Cluster

Optimized resynchronization in Solaris Volume Manager for Sun Cluster functions identically to optimized resynchronization in Solaris Volume Manager. However, in a multi-owner disk set, a RAID-1 volume with the resynchronization option set to optimized resynchronization always has a mirror owner. The following output from the metastat -s diskset-name command shows the resynchronization option set to optimizedresync (for optimized resynchronization).


# metastat -s red
red/d24: Mirror
    Submirror 0: red/d20
      State: Okay
    Submirror 1: red/d21
      State: Okay
    Pass: 1
    Read option: roundrobin (default)
    Write option: parallel (default)
    Resync option: optimizedresync
    Owner: nodeone
    Size: 825930 blocks (403 MB)

For more information on optimized resynchronization, see Optimized Resynchronization.

Application-Based Recovery and Directed Mirror Reads

To optimize data recovery in Solaris Volume Manager for Sun Cluster, applications such as Oracle Real Application Clusters require the ability to manage and control the recovery of data. Enabling an application to control the recovery improves the performance of the recovery. The ioctls DKIOGETVOLCAP, DKIOSETVOLCAP, and DKIODMR provide support for an application's data management recovery in a cluster environment. These ioctls provide an application with the following capabilities:

For more information on the ioctls used with application-based data management recovery, see the dkio(7I) man page.

A RAID-1 volume with the resynchronization option set to application-based recovery only has a mirror owner during the application-based recovery process. The following output from the metastat -s diskset-name command shows a RAID-1 volume in a normal state. The resynchronization option is set to application-based recovery. There is no mirror owner.


# metastat -s red
red/d24: Mirror
    Submirror 0: red/d20
      State: Okay
    Submirror 1: red/d21
      State: Okay
    Pass: 1
    Read option: roundrobin (default)
    Write option: parallel (default)
    Resync option: application based
    Owner: None
    Size: 825930 blocks (403 MB)