A single-initiator SCSI bus has only one node connected to it,
and provides host isolation and better performance than a
multi-initiator bus. Single-initiator buses ensure that each node
is protected from disruptions due to the workload, initialization,
or repair of the other nodes.
When using a single- or dual-controller RAID array that has
multiple host ports and provides simultaneous access to all the
shared logical units from the host ports on the storage enclosure,
the setup of the single-initiator SCSI buses to connect each
cluster node to the RAID array is possible. If a logical unit can
fail over from one controller to the other, the process must be
transparent to the operating system. Note that some RAID
controllers restrict a set of disks to a specific controller or
port. In this case, single-initiator bus setups are not
possible.
A single-initiator bus must adhere to the requirements described
in Section A.3.1
SCSI Configuration Requirements.
To set up a single-initiator SCSI bus configuration, perform the
following steps:
-
Enable the onboard termination for each host bus adapter.
-
Enable the termination for each RAID controller.
-
Use the appropriate SCSI cable to connect each host bus adapter
to the storage enclosure.
Setting host bus adapter termination is done in the adapter BIOS
utility during system boot. To set RAID controller termination,
refer to the vendor documentation. Figure A-3
shows a configuration that uses two single-initiator SCSI
buses.
Figure
A-4 shows the termination in a single-controller RAID array
connected to two single-initiator SCSI buses.
Figure
A-5 shows the termination in a dual-controller RAID array
connected to two single-initiator SCSI buses.
SCSI devices must adhere to a number of configuration
requirements to operate correctly. Failure to adhere to these
requirements adversely affects cluster operation and resource
availability.
The following is an overview of SCSI configuration
requirements:
-
Buses must be terminated at each end. Refer to Section A.3.2
SCSI Bus Termination for more information.
-
Buses must not extend beyond the maximum length restriction for
the bus type. Internal cabling must be included in the length of
the SCSI bus. Refer to Section A.3.3 SCSI
Bus Length for more information.
-
All devices (host bus adapters and disks) on a bus must have
unique SCSI identification numbers. Refer to Section A.3.4 SCSI
Identification Numbers for more information.
-
The Linux device name for each shared SCSI device must be the
same on each cluster system. For example, a device named /dev/sdc on one cluster system must be named
/dev/sdc on the other cluster system. One
way to ensure that devices are named the same is by using identical
hardware for both cluster systems.
Use the system's configuration utility to set SCSI
identification numbers and enable host bus adapter termination.
When the system boots, a message is displayed describing how to
start the utility. For example, the utility prompts the user to
press [Ctrl]-[A], and follow the prompts to perform a particular
task. To set storage enclosure and RAID controller termination,
refer to the vendor documentation. Refer to Section A.3.2
SCSI Bus Termination and Section A.3.4 SCSI
Identification Numbers for more information.
A SCSI bus is an electrical path between two terminators. A
device (host bus adapter, RAID controller, or disk) attaches to a
SCSI bus by a short stub, which is an
unterminated bus segment that usually must be less than 0.1 meter
in length.
Buses must have only two terminators located at opposing ends of
the bus. Additional terminators, terminators that are not at the
ends of the bus, or long stubs cause the bus to operate
incorrectly. Termination for a SCSI bus can be provided by the
devices connected to the bus or by external terminators, if the
internal (onboard) device termination can be disabled.
Testing has shown that external termination on HBAs that run at
speeds greater than 80MB/second does not work reliably.
When disconnecting a device from a single-initiator SCSI bus
follow these guidelines:
-
Unterminated SCSI cables must not be connected to an operational
host bus adapter or storage device.
-
Connector pins must not bend or touch an electrical conductor
while the SCSI cable is disconnected.
-
To disconnect a host bus adapter from a single-initiator bus,
first disconnect the SCSI cable from the RAID controller and then
from the adapter. This ensures that the RAID controller is not
exposed to any erroneous input.
-
Protect connector pins from electrostatic discharge while the
SCSI cable is disconnected by wearing a grounded anti-static wrist
guard and physically protecting the cable ends from contact with
other objects.
-
Do not remove a device that is currently participating in any
SCSI bus transactions.
To enable or disable an adapter's internal termination, use the
system BIOS utility. When the system boots, a message is displayed
describing how to start the utility. For example, many utilities
prompt users to press [Ctrl]-[A]. Follow the
prompts for setting the termination. At this point, it is also
possible to set the SCSI identification number, as needed, and
disable SCSI bus resets. Refer to Section A.3.4 SCSI
Identification Numbers for more information.
To set storage enclosure and RAID controller termination, refer
to the vendor documentation.
A SCSI bus must adhere to length restrictions for the bus type.
Buses that do not adhere to these restrictions do not operate
properly. The length of a SCSI bus is calculated from one
terminated end to the other and must include any cabling that
exists inside the system or storage enclosures.
A cluster supports LVD (low voltage differential) buses. The
maximum length of a single-initiator LVD bus is 25 meters. The
maximum length of a multi-initiator LVD bus is 12 meters. According
to the SCSI standard, a single-initiator LVD bus is a bus that is
connected to only two devices, each within 0.1 meter from a
terminator. All other buses are defined as multi-initiator
buses.
Do not connect any single-ended devices to an LVD bus; doing so
converts the bus single-ended, which has a much shorter maximum
length than a differential bus.
Each device on a SCSI bus must have a unique SCSI identification
number. Devices include host bus adapters, RAID controllers, and
disks.
The number of devices on a SCSI bus depends on the data path for
the bus. A cluster supports wide SCSI buses, which have a 16-bit
data path and support a maximum of 16 devices. Therefore, there are
sixteen possible SCSI identification numbers that can be assigned
to the devices on a bus.
In addition, SCSI identification numbers are prioritized. Use
the following priority order to assign SCSI identification
numbers:
7 - 6 - 5 - 4 - 3 - 2 - 1 - 0 - 15 - 14 - 13 - 12 - 11 - 10 - 9 - 8
|
The previous order specifies that 7 is the highest priority, and
8 is the lowest priority. The default SCSI identification number
for a host bus adapter is 7, because adapters are usually assigned
the highest priority. It is possible to assign identification
numbers for logical units in a RAID subsystem by using the RAID
management interface.
To modify an adapter's SCSI identification number, use the
system BIOS utility. When the system boots, a message is displayed
describing how to start the utility. For example, a user may be
prompted to press [Ctrl]-[A] and follow the prompts for setting the SCSI
identification number. At this point, it is possible to enable or
disable the adapter's internal termination, as needed, and disable
SCSI bus resets. Refer to Section A.3.2
SCSI Bus Termination for more information.
The prioritized arbitration scheme on a SCSI bus can result in
low-priority devices being locked out for some period of time. This
may cause commands to time out, if a low-priority storage device,
such as a disk, is unable to win arbitration and complete a command
that a host has queued to it. For some workloads, it is possible to
avoid this problem by assigning low-priority SCSI identification
numbers to the host bus adapters.
