Skip Navigation Links | |
Exit Print View | |
Using Sun QFS and Sun Storage Archive Manager With Oracle Solaris Cluster Sun QFS and Sun Storage Archive Manager 5.3 Information Library |
1. Using SAM-QFS With Oracle Solaris Cluster
2. Requirements for Using SAM-QFS With Oracle Solaris Cluster
Example - Verifying Devices and Device Redundancy
Analyzing the Output From the Commands
3. Configuring Sun QFS Local Failover File Systems With Oracle Solaris Cluster
4. Configuring Sun QFS Shared File Systems With Oracle Solaris Cluster
5. Configuring SAM-QFS Archiving in an Oracle Solaris Cluster Environment (HA-SAM)
The following sections provide disk device requirements for clustered (shared) file systems and for failover (local) file systems.
To configure a shared file system, use raw device identifier (DID) devices. In the cldevice show command output, these appear as /dev/did/rdsk/dN devices. The Oracle Solaris Cluster nodes that share the file system must have access to each DID device through a host bus adapter (HBA) direct connection. All devices must be accessible to the file system from all nodes in the Oracle Solaris Cluster environment that mount the shared file system. For more information about DID devices, see the did(7) man page.
Note - Requirements for using device identifier are different for clients outside the cluster.
When you specify these devices in your mcf file, use the /dev/did/rdsk/dN devices from the cldevice show command output. For more information about using the cldevice command, see Example - Verifying Devices and Device Redundancy.
Note - The Sun QFS software supports using multi-owner disk sets in Solaris Volume Manager for Oracle Solaris Cluster to obtain redundancy.
To configure a local file system for failover, you must use highly available devices. You can use either of the following options:
Raw devices – Use Oracle Solaris Cluster global devices. Use the output from the cldevice show command to determine the names of the global devices and substitute global for did when specifying the devices in the mcf file. Global devices are accessible from all nodes in an Oracle Solaris Cluster environment, even if these devices are not physically attached to all nodes. If all nodes that have a hardware connection to the disk lose their connections, the remaining nodes cannot access the disk. File systems created on global devices are not necessarily highly available.
Volume manager - Use one of the following:
Solaris Volume Manager for Oracle Solaris Cluster for either shared or stand-alone Sun QFS configurations. Such devices are located in /dev/md.
Veritas Volume Manager (VxVM) for stand-alone QFS configurations only. Such devices are located in /dev/vx.
Use the clsetup utility to register volume-managed devices with the Oracle Solaris Cluster framework before you configure your file system.
Note - If you use a volume manager, use it only to provide redundancy. For performance reasons, do not use the volume manager to concatenate storage on separate devices. This practice causes the Sun QFS highly available file system to distribute I/O inefficiently across the component devices.
To confirm which devices in your Oracle Solaris Cluster environment are highly available, use the cldevice show | grep Device command . This command lists the paths of the devices in the DID configuration file. In the output from the cldevice show command, look for devices that have two or more DID devices listed with the identical DID device number. Such devices are highly available in an Oracle Solaris Cluster environment and can also be configured as global devices for a file system, even if they directly connect only to a single node.
I/O requests issued to global devices from a node other than the direct-attached node are issued over the Oracle Solaris Cluster interconnect. These single-node, global devices cease to be available when all nodes that have direct access to the device are unavailable.