A TruCluster Server cluster uses shared SCSI buses, external storage shelves or redundant array of independent disks (RAID) controllers, and supports disk mirroring and fast file system recovery to provide high data availability and reliability.
This chapter discusses the following topics:
Shared SCSI bus configuration requirements (Section 3.1)
SCSI bus performance (Section 3.2)
SCSI bus device identification numbers (Section 3.3)
SCSI bus length (Section 3.4)
SCSI bus termination (Section 3.5)
UltraSCSI hubs (Section 3.6)
How to configure UltraSCSI hubs with RAID array controllers (Section 3.7)
This chapter:
Introduces SCSI bus configuration concepts
Describes requirements for the shared SCSI bus
Provides procedures for cabling TruCluster Server radial configurations using UltraSCSI hubs and:
Dual-redundant HSZ70 or HSZ80 RAID array controllers that are enabled for simultaneous failover
Dual-redundant HSZ70 or HSZ80 RAID array controllers that are enabled for multiple-bus failover
RAID Array 3000 (RA3000) with HSZ22 controller configured for active/active or active/passive mode.
Provides diagrams of TruCluster Server storage configurations using UltraSCSI hardware that is configured for radial connections
Note
Although the UltraSCSI BA356 might have been included in this chapter with the other UltraSCSI devices, it is not. The UltraSCSI BA356 is discussed in Chapter 10 with the configurations using external termination. It cannot be cabled directly to an UltraSCSI hub because it does not provide SCSI bus termination power (
termpwr
).
In addition to using only supported hardware, adhering to the requirements described in this chapter will ensure that your cluster operates correctly.
Chapter 10
contains additional information
about using SCSI bus signal converters, and also contains diagrams of TruCluster Server
configurations using UltraSCSI and non-UltraSCSI storage shelves and
RAID array controllers.
The chapter also covers
the older method of using external termination and covers
radial configurations with the DWZZH UltraSCSI hubs and non-UltraSCSI RAID
array controllers.
3.1 Shared SCSI Bus Configuration Requirements
A shared SCSI bus must adhere to the following requirements:
Only an external bus can be used for a shared SCSI bus.
SCSI bus specifications set a limit of 8 devices on an 8-bit (narrow) SCSI bus. The limit is 16 devices on a 16-bit SCSI bus (wide). See Section 3.3 for more information.
The length of each physical bus is strictly limited. See Section 3.4 for more information.
You can directly connect devices only if they have the same transmission mode (differential or single-ended) and data path (narrow or wide). Use a SCSI signal converter to connect devices with different transmission modes. See Section 10.1 for information about the DWZZA (BA350) or DWZZB (BA356) signal converters or the DS-BA35X-DA personality module (which acts as a differential to single-ended signal converter for the UltraSCSI BA356).
For each SCSI bus segment, you can have only two terminators, one at each end. A physical SCSI bus may be composed of multiple SCSI bus segments.
If you do not use an UltraSCSI hub, you must use trilink connectors and Y cables to connect devices to a shared bus, so you can disconnect the devices without affecting bus termination. See Section 10.2 for more information.
Be careful when performing maintenance on any device that is on a shared bus because of the constant activity on the bus. Usually, to perform maintenance on a device without shutting down the cluster, you must be able to isolate the device from the shared bus without affecting bus termination.
All supported UltraSCSI host adapters support UltraSCSI disks at UltraSCSI speeds in UltraSCSI BA356 shelves, RA7000 or ESA10000 storage arrays (HSZ70 and HSZ80), RA8000 or ESA12000 storage arrays (HSZ80 and HSG80), or the RAID Array 3000 (RA3000, with HSZ22 controller). Older, non-UltraSCSI BA356 shelves are supported with UltraSCSI host adapters and host RAID controllers as long as they contain no UltraSCSI disks.
Note
The RA3000 is supported only with the KZPBA-CB UltraSCSI host bus adapter.
UltraSCSI drives and fast wide drives can be mixed together in an UltraSCSI BA356 shelf. (See Chapter 10.)
Differential UltraSCSI adapters may be connected to either (or both) a non-UltraSCSI BA356 shelf (via a DWZZB-VW) and the UltraSCSI BA356 shelf (via the DS-BA35X-DA personality module) on the same shared SCSI bus. The UltraSCSI adapter negotiates maximum transfer speeds with each SCSI device. (See Chapter 10.)
The HSZ70 and HSZ80 UltraSCSI RAID controllers have a wide differential UltraSCSI host bus with a Very High Density Cable Interconnect (VHDCI) connector. HSZ70 and HSZ80 controllers will work with fast and wide differential SCSI adapters (for example, KZPSA-BB) at fast SCSI speeds.
Fast, wide SCSI drives (green StorageWorks building blocks (SBBs) with part numbers ending in -VW) may be used in an UltraSCSI BA356 shelf.
Do not use fast, narrow SCSI drives (green SBBs with part numbers ending in -VA) in any shelf that can assign the drive a SCSI ID greater than 7. They will not work.
The UltraSCSI BA356 requires a 180-watt power supply (BA35X-HH). It will not function properly with the older, lower-wattage BA35X-HF universal 150-watt power supply. (See Chapter 10.)
An older BA356 that has been retrofitted with a BA35X-HH 180-watt power supply and DS-BA35X-DA personality module is still only FCC certified for Fast 10 configurations. (See Chapter 10.)
Before you set up a SCSI bus, it is important that you understand a number of issues that affect the viability of a bus and how the devices that are connected to it operate. Specifically, bus performance is influenced by the following factors:
Transmission method (Section 3.2.2)
Data path (Section 3.2.3)
Bus speed (Section 3.2.4)
3.2.1 SCSI Bus Versus SCSI Bus Segments
An UltraSCSI bus may comprise multiple UltraSCSI bus segments. Each UltraSCSI bus segment comprises electrical conductors that may be in a cable or a backplane, and cable or backplane connectors. Each UltraSCSI bus segment must have a terminator at each end.
Up to two UltraSCSI bus segments may be coupled together with UltraSCSI
hubs or signal converters, increasing the total length of the UltraSCSI
bus.
3.2.2 Transmission Methods
Two transmission methods can be used in a SCSI bus:
Single-ended -- In a single-ended SCSI bus, one data lead and one ground lead are utilized for the data transmission. A single-ended receiver looks only at the signal wire as the input. The transmitted signal arrives at the receiving end of the bus on the signal wire somewhat distorted by signal reflections. The length and loading of the bus determine the magnitude of this distortion. This transmission method is economical, but is more susceptible to noise than the differential transmission method, and requires short cables. Devices with single-ended SCSI devices include the following:
BA350, BA356, and UltraSCSI BA356 storage shelves
Single-ended side of a SCSI signal converter or personality module
Differential -- Differential signal transmission uses two wires to transmit a signal. The two wires are driven by a differential driver that places a signal on one wire (+SIGNAL) and another signal that is 180 degrees out of phase (-SIGNAL) on the other wire. The differential receiver generates a signal output only when the two inputs are different. As signal reflections occur virtually the same on both wires, they are not seen by the receiver, because it only sees differences on the two wires.
This transmission method is less susceptible to noise than single-ended SCSI and enables you to use longer cables. Devices with differential SCSI interfaces include the following:
KZPBA-CB
KZPSA-BB
HSZ40, HSZ50, HSZ70, and HSZ80 controllers
Differential side of a SCSI signal converter or personality module
You cannot use the two transmission methods in the same SCSI bus segment. For example, a device with a differential SCSI interface must be connected to another device with a differential SCSI interface. If you want to connect devices that use different transmission methods, use a SCSI signal converter between the devices. The DS-BA35X-DA personality module is discussed in Section 10.1.2.2. See Section 10.1 for information about using the DWZZ* series of SCSI signal converters.
You cannot use a DWZZA or DWZZB signal converter at UltraSCSI speeds for TruCluster Server if there are any UltraSCSI disks on the bus, because the DWZZA or DWZZB will not operate correctly at UltraSCSI speed. The DS-BA35X-DA personality module contains a signal converter for the UltraSCSI BA356. It is the interface between the shared differential UltraSCSI bus and the UltraSCSI BA356 internal single-ended SCSI bus.
RAID array controller subsystems provide the function of a signal
converter, accepting the differential input and driving the
single-ended device buses.
3.2.3 Data Path
There are two data paths for SCSI devices:
Narrow -- Implies an 8-bit data path for SCSI-2. The performance of this mode is limited.
Wide -- Implies a 16-bit data path for SCSI-2 or UltraSCSI. This mode increases the amount of data that is transferred in parallel on the bus.
Bus speeds vary depending upon the bus clocking rate and bus
width, as listed in
Table 3-1.
Table 3-1: SCSI Bus Speeds
SCSI Bus | Transfer Rate (MHz) | Bus Width (Bytes) | Bus Bandwidth (Speed) MB/sec |
SCSI | 5 | 1 | 5 |
Fast SCSI | 10 | 1 | 10 |
Fast-Wide | 10 | 2 | 20 |
UltraSCSI | 20 | 2 | 40 |
UltraSCSI-II | 40 | 2 | 80 |
3.3 SCSI Bus Device Identification Numbers
On a shared SCSI bus, each SCSI device uses a device address and must have a unique SCSI ID (from 0 through 15). For example, each SCSI bus adapter and each disk in a single-ended storage shelf uses a device address.
SCSI bus adapters have a default SCSI ID that you can change by using console commands or utilities. For example, a KZPSA adapter has an initial SCSI ID of 7.
Note
If you are using a DS-DWZZH-05 UltraSCSI hub with fair arbitration enabled, SCSI ID numbering will change. (See Section 3.6.1.2.)
Use the following priority order to assign SCSI IDs to the SCSI bus adapters connected to a shared SCSI bus:
7-6-5-4-3-2-1-0-15-14-13-12-11-10-9-8
This order specifies that 7 is the highest priority, and 8 is the lowest priority. When assigning SCSI IDs, use the highest priority ID for member systems (starting at 7). Use lower priority IDs for disks.
Normal SCSI priority is not followed when using the DS-DWZZH-05 UltraSCSI hub with fair arbitration enabled because the DS-DWZZH-05 determines the SCSI ID of the next device to use the SCSI bus.
The SCSI ID for a disk in a BA350 storage shelf corresponds
to its slot location.
The SCSI ID for a disk in a BA356 or UltraSCSI
BA356 depends upon its slot location and the personality module SCSI
bus address switch settings.
3.4 SCSI Bus Length
There is a limit to the length of the cables in a shared SCSI bus. The total cable length for a SCSI bus segment is calculated from one terminated end to the other.
If you are using devices that have the same transmission method and data path (for example, wide differential), a shared bus will consist of only one bus segment. If you have devices with different transmission methods, you will have both single-ended and differential bus segments, each of which must be terminated only at both ends and must adhere to the rules on bus length.
Table 3-2
describes the maximum cable
length for a physical SCSI bus segment.
Table 3-2: SCSI Bus Segment Length
SCSI Bus | Bus Speed (MB/sec) | Maximum Cable Length |
Narrow, single-ended | 5 | 6 meters (19.7 feet) |
Narrow, single-ended fast | 10 | 3 meters (9.8 feet) |
Wide differential, fast | 20 | 25 meters (82 feet) |
Differential UltraSCSI | 40 | 25 meters [Footnote 7] |
Because of the cable length limit, you must plan your hardware
configuration carefully, and ensure that each SCSI bus meets the cable limit
guidelines.
In general, you must place systems and storage shelves as
close together as possible and choose the shortest possible cables for the
shared bus.
3.5 Terminating the Shared SCSI Bus When Using UltraSCSI Hubs
You must properly connect devices to a shared SCSI bus. In addition, you can terminate only the beginning and end of each bus segment (either single-ended or differential).
There are two rules for SCSI bus termination:
There are only two terminators for each SCSI bus segment. If you use an UltraSCSI hub, you only have to install one terminator.
If you do not use an UltraSCSI hub, bus termination must be external. External termination is covered in Section 10.2.
Notes
With the exception of the TZ885, TZ887, TL890, TL891, and TL892, tape devices can only be installed at the end of a shared SCSI bus. These tape devices are the only supported tape devices that can be terminated externally.
We recommend that tape loaders be on a separate, shared SCSI bus to allow normal shared SCSI bus termination for those shared SCSI buses without tape loaders.
Whenever possible, connect devices to a shared bus so that they can be isolated from the bus. This allows you to disconnect devices from the bus for maintenance purposes, without affecting bus termination and cluster operation. You also can set up a shared SCSI bus so that you can connect additional devices at a later time without affecting bus termination.
Most devices have internal termination. For example, the UltraSCSI KZPBA-CB and the fast and wide KZPSA-BB host bus adapters have internal termination. When using a KZPBA-CB or KZPSA-BB with an UltraSCSI hub, ensure that the onboard termination resistor SIPs have not been removed.
You will need to provide termination at the storage end of one SCSI bus segment. You will install an H8861-AA trilink connector on the HSZ70 or HSZ80 at the bus end. Connect an H8863-AA terminator to the trilink connector to terminate the bus.
Figure 3-1
shows a VHDCI trilink
connector (UltraSCSI), which you may attach to an HSZ70 or HSZ80.
Figure 3-1: VHDCI Trilink Connector (H8861-AA)
The DS-DWZZH series UltraSCSI hubs are UltraSCSI signal converters that provide radial connections of differential SCSI bus adapters and RAID array controllers. Each connection forms a SCSI bus segment with SCSI bus adapters or the storage unit. The hub provides termination for one end of the bus segment. Termination for the other end of the bus segment is provided by the folowing components:
Installed KZPBA-CB (or KZPSA-BB) termination resistor SIPs
External termination on a trilink connector that is attached to an HSZ40, HSZ50, HSZ70, or HSZ80
Note
The DS-DWZZH-03/05 UltraSCSI hubs cannot be connected to a StorageWorks BA35X storage shelf because the storage shelf does not provide termination power to the hub.
3.6.1 Using a DWZZH UltraSCSI Hub in a Cluster Configuration
The DS-DWZZH-03 and DS-DWZZH-05 UltraSCSI hubs are supported in a TruCluster Server cluster. They both provide radial connection of cluster member systems and storage, and are similar in the following ways:
Contain internal termination for each port; therefore, the hub end of each SCSI bus segment is terminated.
Note
Do not put trilinks on a DWZZH UltraSCSI hub because it is not possible to remove the DWZZH internal termination.
Require that termination power (
termpwr
) be provided by the SCSI bus
host adapters on each SCSI bus segment.
Note
The UltraSCSI hubs are designed to sense loss of termination power (such as a cable pull or
termpwr
not enabled on the host adapter) and shut down the applicable port to prevent corrupted signals on the remaining SCSI bus segments.
3.6.1.1 DS-DWZZH-03 Description
The DS-DWZZH-03:
Is an 8.9-centimeter (3.5-inch) StorageWorks building block (SBB).
Can be installed in:
A StorageWorks UltraSCSI BA356 storage shelf (which has the required 180-watt power supply).
The lower righthand device slot of the BA370 shelf within the RA7000 or ESA 10000 RAID array subsystems. This position minimizes cable lengths and interference with disks.
A non-UltraSCSI BA356 that has been upgraded to the 180-watt power supply with the DS-BA35X-HH option.
Uses the storage shelf only to provide its power and mechanical support. (It is not connected to the shelf internal SCSI bus.)
Has three Very High Density Cable Interconnect (VHDCI) differential SCSI bus connectors.
Does not use a SCSI ID.
DS-DWZZH-03 and DS-DWZZH-05 UltraSCSI hubs may be housed in the same storage shelf with disk drives. Table 3-3 provides the supported configurations.
Figure 3-2
shows a front view of the
DS-DWZZH-03 UltraSCSI hub.
Figure 3-2: DS-DWZZH-03 Front View
The differential symbol (and the lack of a single-ended symbol) indicates
that all three connectors are differential.
3.6.1.2 DS-DWZZH-05 Description
Is a 13.33-centimeter (5.25-inch) StorageWorks building block (SBB).
Has five Very High Density Cable Interconnect (VHDCI) differential SCSI bus connectors.
Uses SCSI ID 7 whether or not fair arbitration mode is enabled. Therefore, you cannot use SCSI ID 7 on the member systems' SCSI bus adapter.
The following section describes how to prepare the DS-DWZZH-05
UltraSCSI hub for use on a shared SCSI bus in more detail.
3.6.1.2.1 DS-DWZZH-05 Configuration Guidelines
The DS-DWZZH-05 UltraSCSI hub can be installed in:
A StorageWorks UltraSCSI BA356 shelf (which has the required 180-watt power supply).
A non-UltraSCSI BA356 that has been upgraded to the 180-watt power supply with the DS-BA35X-HH option.
Note
Dual power supplies are recommended for any BA356 shelf containing a DS-DWZZH-05 UltraSCSI hub in order to provide a higher level of availability between cluster member systems and storage.
The lower righthand device slot of the BA370 shelf within the RA7000 or ESA 10000 RAID array subsystems. This position minimizes cable lengths and interference with disks.
A DS-DWZZH-05 UltraSCSI hub uses the storage shelf only to provide its power and mechanical support (it is not connected to the shelf internal SCSI bus).
Note
When the DS-DWZZH-05 is installed, its orientation is rotated 90 degrees counterclockwise from what is shown in Figure 3-3 and Figure 3-4.
The maximum configurations with combinations of DS-DWZZH-03 and DS-DWZZH-05 UltraSCSI hubs, and disks in the same storage shelf containing dual 180-watt power supplies, are listed in Table 3-3.
Note
With dual 180-watt power supplies installed, there are slots available for six 8.9-centimeter (3.5-inch) SBBs or two 13.33-centimeter (5.25-inch) SBBs.
Table 3-3: DS-DWZZH UltraSCSI Hub Maximum Configurations
DS-DWZZH-03 | DS-DWZZH-05 | Disk Drives [Footnote 8] | Personality Module [Footnote 9] [Footnote 10] |
5 | 0 | 0 | Not Installed |
4 | 0 | 0 | Installed |
3 | 0 | 3 | Installed |
2 | 0 | 4 | Installed |
1 | 0 | 5 | Installed |
0 | 2 | 0 | Not Installed |
3 | 1 | 0 | Not Installed |
2 | 1 | 1 | Installed |
1 | 1 | 2 | Installed |
0 | 1 | 3 | Installed |
3.6.1.2.2 DS-DWZZH-05 Fair Arbitration
Although each cluster member system and storage controller that are connected to an UltraSCSI hub are on separate SCSI bus segments, they all share a common SCSI bus and its bandwidth. As the number of systems accessing the storage controllers increases, the adapter with the highest priority SCSI ID will probably obtain a higher proportion of the UltraSCSI bandwidth.
The DS-DWZZH-05 UltraSCSI hub provides a fair arbitration feature that overrides the traditional SCSI bus priority. Fair arbitration applies only to the member systems, not to the storage controllers (which are assigned higher priority than the member system host adapters).
You enable fair arbitration by placing the switch on the front of the
DS-DWZZH-05 UltraSCSI hub to the
Fair
position.
(See
Figure 3-4.)
Fair arbitration works as follows. The DS-DWZZH-05 UltraSCSI hub is assigned the highest SCSI ID, which is 7. During the SCSI arbitration phase, the hub, because it has the highest priority, captures the SCSI ID of all host adapters arbitrating for the bus. The hub compares the SCSI IDs of the host adapters requesting use of the SCSI bus, and then allows the device with the highest priority SCSI ID to take control of the SCSI bus. That SCSI ID is removed from the group of captured SCSI IDs prior to the next comparison.
After the host adapter has been serviced, if there are still SCSI IDs retained from the previous arbitration cycle, the next highest SCSI ID is serviced.
When all devices in the group have been serviced, the DS-DWZZH-05 repeats the sequence at the next arbitration cycle.
Fair arbitration is disabled by placing the switch on the front of the
DS-DWZZH-05 UltraSCSI hub in the
Disable
position.
(See
Figure 3-4.) With fair arbitration disabled, the SCSI
requests are serviced in the conventional manner; the highest SCSI ID
asserted during the arbitration cycle obtains use of the SCSI bus.
Note
Host port SCSI ID assignments are not linked to the physical port when fair arbitration is disabled.
The DS-DWZZH-05 reserves SCSI ID 7 regardless of whether fair arbitration is enabled or not.
3.6.1.2.3 DS-DWZZH-05 Address Configurations
The DS-DWZZH-05 has two addressing modes: wide addressing mode and narrow addressing mode. With either addressing mode, if fair arbitration is enabled, each hub port is assigned a specific SCSI ID. This allows the fair arbitration logic in the hub to identify the SCSI ID of the device participating in the arbitration phase of the fair arbitration cycle.
Caution
If fair arbitration is enabled, the SCSI ID of the host adapter must match the SCSI ID assigned to the hub port. Mismatching or duplicating SCSI IDs will cause the hub to hang.
SCSI ID 7 is reserved for the DS-DWZZH-05 whether fair arbitration is enabled or not.
Jumper W1, which is accessible from the rear of the DS-DWZZH-05 (see Figure 3-3), determines which addressing mode is used. The jumper is installed to select narrow addressing mode. If fair arbitration is enabled, the SCSI IDs for the host adapters are 0, 1, 2, and 3. (See the port numbers not in parentheses in Figure 3-4.) The controller ports are assigned SCSI IDs 4 through 6, and the hub uses SCSI ID 7.
If jumper W1 is removed, the host adapter ports assume SCSI IDs 12,
13, 14, and 15.
The controllers are assigned SCSI IDs 0 through 6.
The DS-DWZZH-05 retains the SCSI ID of 7.
Figure 3-3: DS-DWZZH-05 Rear View
Figure 3-4: DS-DWZZH-05 Front View
3.6.1.2.4 SCSI Bus Termination Power
Each host adapter that is connected to a DS-DWZZH-05 UltraSCSI hub port must
supply termination power (
termpwr
) to enable the termination resistors
on each end of the SCSI bus segment.
If the host adapter is
disconnected from the hub, the port is disabled.
Only the UltraSCSI
bus segment losing termination power is affected.
The remainder of
the SCSI bus operates normally.
3.6.1.2.5 DS-DWZZH-05 Indicators
The DS-DWZZH-05 has two indicators on the front panel.
(See
Figure 3-4.) The green LED indicates that power
is applied to the hub.
The yellow LED indicates that the SCSI bus is
busy.
3.6.1.3 Installing the DS-DWZZH-05 UltraSCSI Hub
To install the DS-DWZZH-05 UltraSCSI hub, follow these steps:
Remove the W1 jumper to enable wide addressing mode. (See Figure 3-3.)
If fair arbitration is to be used, ensure that the switch on the
front of the DS-DWZZH-05 UltraSCSI hub is in the
Fair
position.
Install the DS-DWZZH-05 UltraSCSI hub in a UltraSCSI BA356, non-UltraSCSI BA356 (if it has the required 180-watt power supply), or BA370 storage shelf.
3.7 Preparing the UltraSCSI Storage Configuration
A TruCluster Server cluster provides you with high data
availability through the Cluster File System (CFS), the device
request dispatcher, service failover through the cluster
application availability (CAA) subsystem, disk mirroring, and
fast file system recovery.
TruCluster Server supports mirroring of
the member-specific boot disks and the cluster quorum disk
through hardware RAID only.
You can mirror the clusterwide root
(/
),
/usr
and
/var
file systems, the data disks, and the
swap disk using the Logical Storage Manager (LSM) technology.
You must determine the storage configuration that will meet your
needs.
Mirroring disks across two shared buses provides the most
highly available data.
To determine the supported storage shelves, disk devices, and RAID array controllers, see the AlphaServer options list for your system at the following URL: http://www.compaq.com/alphaserver/products/options.html
Disk devices that are used on the shared bus must be installed in a supported storage shelf or behind a RAID array controller. Before you connect a storage shelf to a shared SCSI bus, you must install the disks in the unit. Before connecting a RAID array controller to a shared SCSI bus, install the disks and configure the storagesets. For detailed information about installation and configuration, see your storage shelf (or RAID array controller) documentation.
Note
The following sections mention only the KZPBA-CB UltraSCSI host bus adapter because it is needed to obtain UltraSCSI speeds for UltraSCSI configurations. The KZPSA-BB host bus adapter may be used in any configuration in place of the KZPBA-CB without any cable changes providing that it is supported on the member system and storage device. Be aware though, the KZPSA-BB is not an UltraSCSI device and therefore only works at fast-wide speed (20 MB/sec).
The following sections describe how to prepare and install
cables for storage configurations on a shared SCSI bus using
UltraSCSI hubs and the HSZ70 and HSZ80 RAID array controllers, or
the RAID Array 3000.
3.7.1 Configuring Radially Connected TruCluster Server Clusters with UltraSCSI Hardware
Radial configurations with RAID array controllers allow you to take advantage of the benefits of hardware mirroring, and to achieve a no-single-point-of-failure (NSPOF) cluster. Typical RAID array storage subsystems used in TruCluster Server cluster configurations are:
RA7000 or ESA10000 with HSZ70 controller
RA7000 or ESA10000 with HSZ80 controller
RA8000 or ESA12000 with HSZ80 controller
RA3000 with HSZ22 controller
Note
You cannot achieve a NSPOF configuration with a RA3000.
When used with TruCluster Server, one advantage of using a RAID
array controller is the ability to hardware mirror the clusterwide
root (/
) file system, member system boot disks,
swap disk, and quorum disk.
HSZ70 or HSZ80
When used in a dual-redundant configuration, Tru64 UNIX Version 5.1A supports both transparent failover, which occurs automatically, without host intervention, and multiple-bus failover, which requires host intervention for some failures.
Note
Enable mirrored cache for dual-redundant configurations to further ensure the availability of unwritten cache data.
Use transparent failover if you only have one shared SCSI bus. Both controllers are connected to the same host and device buses, and either controller can service all of the units if the other controller fails.
Transparent failover compensates only for a controller failure, and not for failures of either the SCSI bus or host adapters and is therefore not a NSPOF configuration.
Note
Set each controller to transparent failover mode before configuring devices (
SET FAILOVER COPY = THIS_CONTROLLER
).
To achieve a NSPOF configuration, you need multiple-bus failover and two shared SCSI buses.
You may use multiple-bus failover (SET MULTIBUS_FAILOVER
COPY = THIS_CONTROLLER
) to help achieve a NSPOF
configuration if each host has two shared SCSI buses to the array
controllers.
One SCSI bus is connected to one controller and the
other SCSI bus is connected to the other controller.
Each member
system has a host bus adapter for each shared SCSI bus.
The load can
be distributed across the two controllers.
In case of a host adapter
or SCSI bus failure, the host can redistribute the load to the
surviving controller.
In case of a controller failure, the surviving
controller will handle all units.
Notes
Multiple-bus failover does not support device partitioning with the HSZ70 or HSZ80.
Partitioned storagesets and partitioned single-disk units cannot function in multiple-bus failover dual-redundant configurations. Because they are not supported, you must delete your partitions before configuring the HSZ70 or HSZ80 controllers for multiple-bus failover.
Device partitioning is supported with HSG60 and HSG80 array controllers with ACS Version 8.5.
Multiple-bus failover does not support tape drives or CD-ROM drives.
RA3000
The RA3000 uses either active/active or active/passive mode and does not support transparent or multiple-bus failover.
In the active/active mode, the top controller sees one host port as active, while the other controller sees the other host port as active. The controllers see their non-active host ports as passive. If one of the controllers fails, the surviving controller sees both host ports as active.
In the active/passive mode, the primary controller sees both host ports as active. The other controller sees both host ports as passive. If the primary controller fails, the remaining controller takes over and sees both host ports as active.
The following sections describe how to cable the HSZ70, HSZ80, or
RA3000 for TruCluster Server configurations using an UltraSCSI hub.
See
Chapter 9
and
Chapter 10
for information on configurations
using external termination.
See
Chapter 6
for information regarding Fibre Channel
storage.
3.7.1.1 Preparing an HSZ70 or HSZ80 for a Shared SCSI Bus Using Transparent Failover Mode
When using transparent failover mode:
Both controllers of an HSZ70 are connected to the same shared SCSI bus
For an HSZ80:
Port 1 of controller A and Port 1 of controller B are on the same SCSI bus.
If used, Port 2 of controller A and Port 2 of controller B are on the same SCSI bus.
HSZ80 targets assigned to Port 1 cannot be seen by Port 2.
To cable a dual-redundant HSZ70 or HSZ80 for transparent failover in a TruCluster Server configuration using a DS-DWZZH-03 or DS-DWZZH-05 UltraSCSI hub, see Figure 3-5 (HSZ70) or Figure 3-6 (HSZ80) and follow these steps:
You will need two H8861-AA VHDCI trilink connectors. Install an H8863-AA VHDCI terminator on one of the trilinks.
Attach the trilink with the terminator to the controller that you want to be on the end of the shared SCSI bus. Attach an H8861-AA VHDCI trilink connector to:
HSZ70 controller A and controller B
HSZ80 Port 1 (2) of controller A and Port 1 (2) of controller B
Note
You must use the same port on each HSZ80 controller.
Install a BN37A cable between the trilinks on:
HSZ70 controller A and controller B
HSZ80 controller A Port 1 (2) and controller B Port 1 (2)
The BN37A-0C is a 30-centimeter (11.8-inch) cable and the BN37A-0E is a 50-centimeter (19.7-inch) cable.
Install the DS-DWZZH-03 or DS-DWZZH-05 UltraSCSI hub in an UltraSCSI BA356, non-UltraSCSI BA356 (with the required 180-watt power supply), or BA370 storage shelf. (See Section 3.6.1.1 or Section 3.6.1.2.)
If you are using a:
DWZZH-03: Install a BN37A cable between any DWZZH-03 port and the open trilink connector on HSZ70 controller A (B) or HSZ80 controller A Port 1 (2) or controller B Port 1 (2).
DWZZH-05:
Verify that the fair arbitration switch is in the
Fair
position to enable fair arbitration.
(See
Section 3.6.1.2.2.)
Ensure that the W1 jumper is removed to select wide addressing mode. (See Section 3.6.1.2.3.)
Install a BN37A cable between the DWZZH-05 controller port and the open trilink connector on HSZ70 controller A (B) or HSZ80 controller A Port 1 (2) or controller B Port 1 (2).
When the KZPBA-CB host bus adapters in each member system are installed, connect each KZPBA-CB to a DWZZH port with a BN38C (or BN38D) HD68 to VHDCI cable. Ensure that the KZPBA-CB SCSI ID matches the SCSI ID that is assigned to the DWZZH-05 port it is cabled to (12, 13, 14, and 15).
Figure 3-5
shows a two-member
TruCluster Server configuration with a radially connected dual-redundant
HSZ70 RAID array controller configured for transparent failover.
Figure 3-5: Shared SCSI Bus with HSZ70 Configured for Transparent Failover
Table 3-4
lists the components
that are used to create the clusters that are shown in
Figure 3-5,
Figure 3-6,
Figure 3-7, and
Figure 3-8.
Table 3-4: Hardware Components Shown in Figure 3-5 Through Figure 3-8
Callout Number | Description |
1 | BN38C cable [Footnote 11] |
2 | BN37A cable [Footnote 12] |
3 | H8861-AA VHDCI trilink connector |
4 | H8863-AA VHDCI terminator [Footnote 12] |
Figure 3-6
shows a two-member
TruCluster Server configuration with a radially connected dual-redundant
HSZ80 RAID array controller configured for transparent failover.
Figure 3-6: Shared SCSI Bus with HSZ80 Configured for Transparent Failover
Table 3-4
lists the components
that are used to create the cluster that is shown in
Figure 3-6.
3.7.1.2 Preparing a Dual-Redundant HSZ70 or HSZ80 for a Shared SCSI Bus Using Multiple-Bus Failover
Multiple-bus failover is a dual-redundant controller configuration in which each host has two paths (two shared SCSI buses) to the array controller subsystem. The hosts have the capability to move LUNs from one controller (shared SCSI bus) to the other. If one host adapter or SCSI bus fails, the hosts can move all storage to the other path. Because both controllers can service all of the units, either controller can continue to service all of the units if the other controller fails. Therefore, multiple-bus failover can compensate for a failed host bus adapter, SCSI bus, or RAID array controller, and can, if the rest of the cluster has necessary hardware, provide a NSPOF configuration.
Note
Each host (cluster member system) requires at least two KZPBA-CB host bus adapters.
Although both the HSZ70 and HSZ80 have multiple-bus failover, they operate differently:
HSZ70: Only one controller (or shared SCSI bus) is active for the units that are preferred (assigned) to it. If all units are preferred to one controller, then all units are accessed through one controller. If a controller detects a problem, all of its units fail over to the other controller. If the host detects a problem with the host bus adapter or SCSI bus, the host initiates the failover to the other controller (and SCSI bus).
HSZ80: Both HSZ80 controllers can be active at the same time. If the host detects a problem with a host bus adapter or SCSI bus, the host initiates the failover to the other controller. If a controller detects a problem, all of its units fail over to the other controller.
Also, the HSZ80 has two ports on each controller. If multiple-bus failover mode is enabled, the targets assigned to any one port are visible to all ports unless access to a unit is restricted to a particular port (on a unit-by-unit basis).
To cable an HSZ70 or HSZ80 for multiple-bus failover in a TruCluster Server configuration using DS-DWZZH-03 or DS-DWZZH-05 UltraSCSI hubs (you need two hubs), see Figure 3-7 (HSZ70) and Figure 3-8 (HSZ80) and follow these steps:
Install an H8863-AA VHDCI terminator on each of two H8861-AA VHDCI trilink connectors.
Install H8861-AA VHDCI trilink connectors (with terminators) on:
HSZ70 controller A and controller B
HSZ80 controller A Port 1 (2) and controller B Port 1 (2)
Note
You must use the same port on each HSZ80 controller.
Install the DS-DWZZH-03 or DS-DWZZH-05 UltraSCSI hub in a DS-BA356, BA356 (with the required 180-watt power supply), or BA370 storage shelf. (See Section 3.6.1.1 or Section 3.6.1.2.)
If you are using a:
DS-DWZZH-03: Install a BN37A VHDCI to VHDCI cable between the trilink connector on controller A (HSZ70) or controller A Port 1 (2) (HSZ80) and any DS-DWZZH-03 port. Install a second BN37A cable between the trilink on controller B (HSZ70) or controller B Port 1 (2) (HSZ80) and any port on the second DS-DWZZH-03.
DS-DWZZH-05:
Verify that the fair arbitration switch is in the
Fair
position to enable fair arbitration.
(See
Section 3.6.1.2.2.)
Ensure that the W1 jumper is removed to select wide addressing mode. (See Section 3.6.1.2.3.)
Install a BN37A cable between the DWZZH-05 controller port and the open trilink connector on HSZ70 controller A or HSZ80 controller A Port 1 (2).
Install a second BN37A cable between the second DWZZH-05 controller port and the open trilink connector on HSZ70 controller B or HSZ80 controller B Port 1 (2).
When the KZPBA-CBs are installed, use a BN38C (or BN38D) HD68-to-VHDCI cable to connect the first KZPBA-CB on each system to a port on the first DWZZH hub. Ensure that the KZPBA-CB SCSI ID matches the SCSI ID that is assigned to the DWZZH-05 port it is cabled to (12, 13, 14, and 15).
Install BN38C (or BN38D) HD68-to-VHDCI cables to connect the second KZPBA-CB on each system to a port on the second DWZZH hub. Ensure that the KZPBA-CB SCSI ID matches the SCSI ID that is assigned to the DWZZH-05 port it is cabled to (12, 13, 14, and 15).
Figure 3-7
shows a two-member
TruCluster Server configuration with a radially connected dual-redundant
HSZ70 configured for multiple-bus failover.
Figure 3-7: TruCluster Server Configuration with HSZ70 in Multiple-Bus Failover Mode
Table 3-4 lists the components that are used to create the cluster that is shown in Figure 3-7.
Figure 3-8
shows a two-member
TruCluster Server configuration with a radially connected dual-redundant
HSZ80 configured for multiple-bus failover.
Figure 3-8: TruCluster Server Configuration with HSZ80 in Multiple-Bus Failover Mode
Table 3-4
lists the components
that are used to create the cluster that is shown in
Figure 3-8.
3.7.1.3 Preparing an RA3000 for Use on a Shared SCSI Bus with an UltraSCSI Hub
The RAID Array 3000 (RA3000) is a low-end, standalone UltraSCSI RAID subsystem. It supports RAID levels 0, 1, 0+1, 4, 5, and just a bunch of disks (JBOD) disks.
The RA3000 storage subsystem has fully redundant components to eliminate single points of failure. It comes with a standard uninterruptible power supply (UPS) for cache data protection during power outages.
The RA3000 uses the dual-ported HSZ22 controller. Optional dual redundant controllers with mirrored write-back cache provide maximum data integrity.
The StorageWorks Command Console (SWCC) V2.2 (or higher) client graphical user interface (GUI) runs on a Microsoft Windows NT V4.0 Service Pack 4 (or later) or Windows 2000 PC connected directly to the RA3000 by a serial line.
After the first virtual disk has been created, you can also communicate with your RAID Array 3000 over a TCP/IP network provided the V2.2 (or higher) SWCC Agent has been installed on the Tru64 UNIX member system.
The RA3000 is available as:
DS-SWXRA-GH -- A rackmount subsystem (standard RETMA or metric cabinet) containing a controller shelf with one HSZ22 controller, an uninterruptible power supply (UPS), two host I/O modules, a device I/O module, and one 6-slot device expansion shelf. Up to three additional expansion shelves (DS-SWXRA-GN) may be added to provide a maximum of 24 storage devices.
DS-SWXRA-GA -- A deskside pedestal subsystem that includes one HSZ22 controller. The base pedestal accommodates up to seven storage devices. The included battery backup subsystem is a free-standing UPS. An expansion pedestal option (DS-SWXRA-GD) increases the storage capacity of the subsystem to 14 storage devices.
A second HSZ22 controller option (DS-HSZ22-AA) can be added to either RA3000 subsystem.
To cable a RA3000 in a TruCluster Server configuration using a DS-DWZZH-03 or DS-DWZZH-05 UltraSCSI hub, follow the steps in Table 3-5.
Notes
All configuration illustrations assume that a second, redundant HSZ22 controller is installed to achieve active/active or active/passive failover.
See the RA3000 documentation for information about configuring the storage devices.
Table 3-5: Installing Cables for RA3000 Radial Configuration with a DWZZH UltraSCSI Hub
Action | Refer to: |
Install a BN38C HD68 to VHDCI cable between each KZPBA-CB UltraSCSI host adapter and a DWZZH port. The DWZZH accepts the VHDCI connector. You may use a BN38E-0B technology adapter cable with a BN37A cable instead of the BN38C cable. [Footnote 13] [Footnote 14] | Figure 3-9 through Figure 3-12 |
Install BN37A cables: [Footnote 15] | -- |
RA3000 controller shelf with active/passive failover: Install a BN37A cable between any DWZZH-03 port or the DWZZH-05 controller port and the RA3000 controller shelf Host 0 I/O module Host In port. | Figure 3-9 |
RA3000 pedestal with active/passive failover: Install a BN37A cable between any DWZZH-03 port or the DWZZH-05 controller port and the RA3000 pedestal Host 0 port. | Figure 3-10 |
RA3000 pedestal with active/active or active/passive failover: Install a BN37A cable between the DWZZH-05 controller port and the RA3000 pedestal Host 0 port. Install a second BN37A cable between a DWZZH-05 host port and the RA3000 pedestal Host 1 port. | Figure 3-11 |
RA3000 controller shelf with active/active or active/passive failover: Install a BN37A cable between any DWZZH-03 port or the DWZZH-05 controller port and the RA3000 controller shelf Host 0 I/O module Host In port. Install a BN37A-0E 50-centimeter (19.7-inch) cable between the Host 0 I/O module Host Out port and the Host 1 I/O module Host In port. | Figure 3-12 |
|
In the configuration shown in Figure 3-9 through Figure 3-12, it is assumed that the RA3000 contains two HSZ22 controllers.
Figure 3-9
shows a four-member
TruCluster Server configuration and an RA3000 controller shelf with
active/passive failover radially connected to a DS-DWZZH-05 UltraSCSI
hub.
Table 3-6
describes the callouts.
Figure 3-9: TruCluster Server Configuration with an RA3000 Controller Shelf with Active/Passive Failover
Figure 3-10
shows a four-member
TruCluster Server configuration and an RA3000 pedestal with
active/passive failover radially connected to a DS-DWZZH-05 UltraSCSI
hub.
Table 3-6
describes the callouts.
Figure 3-10: TruCluster Server Configuration with an RA3000 Pedestal with Active/Passive Failover and a DWZZH-05 UltraSCSI Hub
Figure 3-11 shows a two-member TruCluster Server configuration and an RA3000 pedestal with active/active or active/passive failover radially connected to a DS-DWZZH-05 UltraSCSI hub. This configuration uses independent connections to the two pedestal host ports to increase the available bandwidth to the RA3000 controllers. Table 3-6 describes the callouts.
Note
If you connect a DWZZH-05 host port to an RA3000 pedestal host port to provide active/active failover, you must disable fair arbitration on the DWZZH-05 by placing the fair arbitration switch in the
DISABLE
position.
Figure 3-11: TruCluster Server Configuration with an RA3000 Pedestal with Active/Active or Active/Passive Failover
Figure 3-12
shows a four-member
TruCluster Server configuration and an RA3000 controller shelf with
active/active or active/passive failover radially connected to a
DS-DWZZH-05 UltraSCSI hub.
Table 3-6
describes the callouts.
Figure 3-12: TruCluster Server Configuration with an RA3000 Controller Shelf with Active/Active or Active/Passive Failover
Table 3-6: Hardware Components Used in the Configurations Shown in Figure 3-9 through Figure 3-12
Callout Number | Description |
1 | BN38C HD68-to-VHDCI cable. [Footnote 16] A BN38E-0B technology adapter cable may be connected to a BN37A cable and used in place of a BN38C cable. [Footnote 17] |
2 | BN37A VHDCI cable [Footnote 18] |
3 | BN37A-0E 50-centimeter (19.7-inch) VHDCI cable [Footnote 19] |