The following sections describe setting up SCSI and Fibre Channel devices that the cluster systems share. For VCS I/O fencing, the data disks must support SCSI-3 persistent reservations. You need to configure a coordinator disk group that supports SCSI-3 PR and verify that it works.
See also the Veritas Cluster Server User's Guide for a description of I/O fencing.
When SCSI devices are used for storage shared between nodes, the SCSI address, or SCSI initiator ID, of each node must be unique. Since each node typically has the default SCSI address of "7," the addresses of one or more nodes must be changed to avoid a conflict. In the following example, two nodes share SCSI devices. The SCSI address of one node is changed to "5" by using nvedit
commands to edit the nvramrc
script.
Refer to the documentation shipped with the host adapters, the storage, and the systems.
...b) /sbus@6,0/QLGC,isp@2,10000/sd
The example output shows the path to one host adapter. You must include the path information, excluding the "/sd" directory, in the nvramrc
script. The path information varies from system to system.
nvramrc
script on to change the scsi-initiator-id to 5. (The Solaris OpenBoot 3.x Command Reference Manual contains a full list of nvedit
commands and keystrokes.) For example:
As you edit the script, note the following points:
nvedit
commands.
In this example, edit the nvramrc script as follows:
1: cd /sbus@6,0/QLGC,isp@2,10000
nvramrc
script. The changes you make are temporary until you store them.
If you are not sure of the changes you made, you can re-edit the script without risk before you store it. You can display the contents of the nvramrc
script by entering:
You can re-edit the file to make corrections:
nvramrc
script on the node.
...b) /sbus@6,0/QLGC,isp@2,10000/sd
show-disks
command to find the paths for the host adapters. Then, display the properties for that paths. For example:
...b) /sbus@6,0/QLGC,isp@2,10000/sd
If you have more than two systems sharing the SCSI bus, use the same procedure, making sure that the storage devices have power before any of the systems, and that only one node is running at one time until each node's address is set to a unique value.
format(1m)
command to verify that each system can see all shared devices.