1.6.1 Required hardware
This subsection describes the hardware required in order to run HA Monitor. For details about the hardware configuration, see 5.5 Evaluation of hardware configuration.
The following figure shows an example of a hardware configuration.
|
|
|
- Organization of this subsection
(1) Processors
You use one server machine in each of the active and standby systems to execute actual jobs.
Processor requirements for using HA Monitor are as follows:
-
BladeSymphony Integrated Service Platform servers
If you use BladeSymphony, you can install multiple processors on a single chassis. HA Monitor enables you to apply a hot-standby configuration to all the processors in the same chassis.
-
Hitachi Advanced Server HA8000 series servers
-
PC/AT-compatible systems from other companies
If you will be resetting hosts, you must not use BladeSymphony or HA8000 together with PC/AT-compatible systems from other companies.
For details about the functions supported by each shared disk data protection method and the function for controlling hot standby based on the availability of LAN communications, see 3.1 List of functions supported by HA Monitor.
- Note
-
If you use BladeSymphony with Hitachi server virtualization (Virtage), you can use logical partitions (LAPRs) in the processor as hot-standby configuration units.
Regardless of the model of server used, you can run multiple virtual systems on a single processor by using VMware ESXiTM. In this case, you can define each virtual system as a hot-standby switchover configuration unit.
- Note
-
With BladeSymphony and HA8000, a shared disk is protected from data corruption by host reset or SCSI reservation for shared disk. In PC/AT-compatible systems from other companies, a shared disk is protected from data corruption by SCSI reservation for shared disk. If neither host reset nor SCSI reservation for shared disk is available, the function for controlling hot standby based on the availability of LAN communications is used. For details about host reset, see 2.3.5 Host reset. For details about SCSI reservation for shared disk, see 2.3.6 SCSI reservation for shared disk. For details about the function for controlling hot standby based on the availability of LAN communications, see 2.3.7 Function for controlling hot standby based on the availability of LAN communications.
(2) Monitoring path
The monitoring path is used to connect the processors in the active and standby systems and for the systems to monitor each other and exchange information required for hot standby.
You can use a TCP/IP LAN as a LAN dedicated as the monitoring path (monitoring LAN). You must use a LAN that supports broadcast communication because HA Monitor uses a TCP/IP LAN to broadcast. Although you can use a TCP/IP LAN in any hot-standby configuration, you need a dedicated IP address and port number.
If you use HA Monitor Extension to change the maximum number of hosts in a single hot-standby configuration, use a LAN that is capable of multicasting. For details about HA Monitor Extension, see 3.7 Functions of HA Monitor Extension.
(3) Reset path
The reset path is used to send host input/output reset instructions to a failure management processor in the event of a failure in the active system. A reset path is not needed if you do not perform host reset. A failure management processor is a hardware device that enables you to specify the system configuration and to control the CPU. It is an independent hardware device that is installed in the system separately from the CPU. Its name depends on the system you use, such as an SVP, BMC, or management server, but its functions are the same. In this manual, this hardware is referred to collectively as a failure management processor.
For notes about the reset path, see 5.5.4 Reset path configuration.
(4) Shared disk
A shared disk is used to store data that is shared by the active and standby system servers. A shared disk is mandatory when you use SCSI reservation for shared disk. Data on the shared disk is inherited during hot standby. All hosts that use data on a shared disk must be connected to the shared disk. HA Monitor can control the following three types of shared disks:
-
A volume group for which LVM is used
-
A file system directly created in a disk or partition that does not use LVM
HA Monitor supports a shared disk that has one of the following path configurations:
-
Configuration in which the path is not made redundant (single path configuration)
-
Configuration in which the path is made redundant by using DMMP
-
Configuration that is seen as a single path configuration from the OS of the host on which HA Monitor is started
For example, this configuration applies if the path is made redundant by the host OS in a virtualization environment.
-
-
A disk or device connected via HABooster
Supported shared disks include Hitachi disk array systems and shared datastores built with VMware vSAN.
HA Monitor does not handle disk and I/O path failures. To handle them, you must use some method other than HA Monitor to prevent a shared disk from becoming a single point of failure. For details about preventing a single point of failure, see 5.5.1 Shared disk configuration.
To use SCSI reservation for data protection, the following conditions must be met:
-
The storage device supports SCSI-3 Persistent Reservation and has a RAID controller.
A host-base RAID with individual RAID controller is not supported. Such a RAID can be accessed from the standby system, even if it is reserved in the active system.
If you use a shared datastore built with VMware vSAN, you can use either of the following methods to access it:
-
Native access (vSAN native VMDK)#
-
Access via the vSAN iSCSI target service
- #
-
This method cannot be used if SCSI reservation is used for data protection.
(5) Local disk
A local disk is used as a host-specific disk that is not shared with other hosts. You store on the local disk OS files and HA Monitor's various files. You start HA Monitor from the local disk, not the shared disk.
For the local disk also, we recommend that you employ a configuration that can prevent a single point of failure, in the same manner as with a shared disk.
In the event of a shared disk or local disk failure, applications can no longer be run because files cannot be accessed. In such a case, the server and OS will not terminate abnormally but hot standby might fail. To prevent this, consider generating an OS panic by using OS and driver functions to perform hot standby in the event of any I/O error on shared or local disks.
(6) Console
You use the console to execute HA Monitor commands and to check messages that are displayed.
(7) LAN
LANs are used to connect the active and standby systems, as well as to connect workstations and PCs that are clients. Use of a LAN is mandatory when you use the function for controlling hot standby based on the availability of LAN communications. Among the LANs that support TCP/IP communications, you can use any Ethernet LAN supported by the OS.
HA Monitor does not directly process LAN failures in HUB and LAN adaptors. For these failures, use some other method and employ a configuration that can prevent a LAN from becoming a single point of failure. For details about preventing a LAN from becoming a single point of failure, see 5.5.2 LAN configuration.
(8) Management server
To perform host reset with model HA8000xN or later, management servers with HCSM installed are required, depending on the number of monitored hosts. The connection topology of management servers or the number of monitored hosts differs depending on the HA Monitor operating environment. Examples of possible configurations are as follows:
-
One management server is connected to a hot-standby switchover configuration. One management server is used both for central server management and as a failure management processor for HA Monitor.
-
A management server dedicated to HA Monitor is connected to a hot-standby switchover configuration, in addition to a management server for central server management.
-
One management server is connected to multiple hot-standby switchover configurations.
-
One management server is connected to each hot-standby switchover configuration.