SUNW.HAStorage describes a resource type that defines resources in a resource group to synchronize the actions between the cluster file system, global devices, and relevant data services.

There is no direct synchronization between resource groups and disk device groups (and the cluster file system). As a result, during a cluster reboot or failover, an attempt to start a data service can occur while its dependent global devices or cluster file systems are still unavailable. Consequently, the data service's START method might timeout and the service is not started on the cluster.

SUNW.HAStorage is a resource type that specifically monitors the storage device services. You add a resource of this type to resource groups containing other resources and set up dependencies between the other resources and the HAStorage resource. The HAStorage resource continually tests the availability of the global devices, device groups, and the cluster file system. The dependencies ensure that the data service resources does not attempt to start until the device services are available.

When a data service resource is set up with a "strong dependency" upon a SUNW.HAStorage resource, the data service resources are not started before all dependent global devices and cluster file systems become available.

Multiple SUNW.HAStorage resources can be set up within a cluster to obtain finer granularity of the service monitoring checks. Device services that the data service needs to check and wait for but not depend upon to be online can be defined in a separate resource, and a "weak dependency" can be set up from the data resource to the device resource.

In this case, the data service resource waits for the resource to check if the device services are all available. If not, even if the SUNW.HAStorage START method times out, the data service can still be brought online. This feature is useful to some data services. For example, assume a Web server depends on ten cluster file systems. If only one file system isn't ready within the timeout period, the Web service should still go online since it still can provide 90 percent of the services.

Two extension properties are associated with the SUNW.HAStorage resource type: ServicePaths and AffinityOn.

ServicePaths

Contains valid global device group names, paths to global devices, or cluster file system mount points that are to be checked. They are defined in the format of

paths[,...].

A typical example of a global device group is nfs-dg. A path to a global device is a valid device path in the global device namespace, such as /dev/global/dsk/d5s2, /dev/global/dsk/d1s2, or /dev/global/rmt/0. A cluster file system mount point is a valid global mount point defined in /etc/vfstab on all cluster nodes of the cluster. You can define a global device group, a global device path, and a cluster file system mount point in one SUNW.HAStorage resource.

AffinityOn

A boolean flag that specifies whether the SUNW.HAStorage resource needs to do an affinity switchover for the global devices and cluster file systems defined in ServicePaths.

When AffinityOn is set to False, the SUNW.HAStorage resource passively waits for the specified global services to become available. As a result, the primary of each online global service might not be the same node that is the primary of the resource group.

The purpose of an affinity switchover is to enhance performance by having data services and their dependent global services run on the same node. For each global service, the SUNW.HAStorage resource attempts affinity switchover only once. If switchover fails, nothing is affected and the availability check occurs normally.

The default value for ServicePaths is the empty string. The default value for AffinityOn is True. Both extension properties can be changed at any time when the resource group is offline.

For scalable service resources, the setting of the AffinityOn flag is ignored and no affinity switchover can be done. There is no benefit to switching over the disk device services because the scalable data service can be running on multiple nodes simultaneously.

rt_reg(4)

SUNW.HAStorage specifies resources that check and wait for the specified global devices, device group, and cluster file systems to become available. The checking is only meaningful when data service resources (application resources) in the same resource group are set up with the correct dependency upon the SUNW.HAStorage resources. Otherwise, no synchronization is done.

Avoid configuring two different SUNW.HAStorage resources in different resource groups with their ServicePaths property referencing the same global resource and with both AffinityOn flags set to True. When the cluster is booting or during a switchover, the resource groups might end up mastered on two different nodes. Both of the SUNW.HAStorage resources would attempt to do an affinity switchover of the same device group, resulting in a race condition. In this case, redundant switchovers would occur and the device group might not end up being mastered by the most preferred node.

The waiting time for global services to become available is specified by the Prenet_Start_Timeout property in SUNW.HAStorage. The time is tunable with a default value of 30 minutes (1,800 seconds).