For the shared SCSI bus to work correctly, the SCSI host adapters must be configured correctly. As with SCSI specifications, each device on the bus must have a unique ID number. For proper operation, ensure that the host adapters are each set to a unique ID. For best results, set one adapter to ID 6 and the other adapter to ID 7 to ensure that the host adapters have adequate priority on the bus. Also, make sure that both adapters have the same firmware revision level. Because the shared SCSI bus is not used for booting the operating system, disable the BIOS on the adapter, unless otherwise directed by the hardware vendor.
Make sure that you connect only physical disk or hardware RAID devices to the shared bus. Devices other than these, such as tape drives, CD-ROM drives, or removable media devices, should not be used on the shared bus. You may use them on another bus for local storage.
Cables and termination are vital parts of the SCSI bus configuration, and should not be compromised. Cables need to be of high quality and within SCSI specifications. The total cable length on the shared SCSI bus needs to be within specifications. Cables supplied with complete certified systems should be correct for use with the shared SCSI bus. Check for bent pins on SCSI cable connectors and devices, and ensure that each cable is attached firmly.
Correct termination is also important. Terminate the bus at both ends, and use active terminators. Use of SCSI Y cables may allow disconnection of one of the nodes from the shared bus without losing termination. If you have terminators attached to each end of the bus, make sure that the controllers are not trying to also terminate the bus.
Make sure that all devices connected to the bus are rated for the type of controllers used. For example, do not attach differential SCSI devices to a standard SCSI controller. Verify that the controllers can each identify every disk device attached to the bus. Make sure that the configuration of each disk device is correct. Some newer smart devices can automatically terminate the bus or negotiate for SCSI IDs. If the controllers do not support this, configure the drives manually. A mixture of smart devices with others that require manual configuration can lead to problems in some configurations. For best results, configure the devices manually.
Also, make sure that the SCSI controllers on the shared bus are configured correctly and with the same parameters (other than SCSI ID). Data transfer rate and other parameter differences between the two controllers may encourage unpredictable behavior.
To add disk devices to the shared SCSI bus, you must properly shut down all equipment and both cluster nodes. This is necessary because the SCSI bus may be disconnected while adding the device or devices. Attempting to add devices while the cluster and devices are in use may induce failures or other serious problems that may not be recoverable. Add the new device or devices in the same way you add a device to a standard SCSI bus. This means you must choose a unique SCSI ID for the new device, and ensure that the device configuration is correct for the bus and termination scheme. Verify cable and termination before applying power. Turn on one cluster node, and use Disk Administrator to assign a drive letter and format each new device. Before turning on the other node, create a physical disk resource using Cluster Administrator. After you create the physical disk resource and verify that the resource will go online successfully, turn on the other cluster node and allow it to join the cluster. Allowing both nodes to be online without first creating a disk resource for the new device can lead to file system corruption, as both nodes may have different interpretations of disk structure.
A good procedure for verification of cable and termination integrity is to first use the SCSI host adapter utilities to determine whether the adapter can identify all disk devices on the bus. Perform this check with only one node turned on. Then, turn off the computer and perform the same check on the other system. If this initial check succeeds, the next step is to check drive identification from the operating system level, with only one of the nodes turned on. If MSCS is already installed, then the cluster service will need to be started for shared drives to go online. Check to make sure the shared drives go online. If the device fails, there may be a problem with the device, or perhaps a cable or termination problem.