The conditions for the system switchover facility when ClusterPerfect is used are as follows:

• There must be at least two connectable fiber optic cables for the shared disk unit (CR80).
• Only a 1-to-1 switchover configuration can be used. A mutual system switchover configuration cannot be used.

(a) System configuration able to perform system switchovers

The following figure shows a system configuration that can perform system switchovers.

Figure 26-42 System configuration able to perform system switchovers

Explanation

The fiber optic cable of each server machine is connected to the two fiber optic cables that connect to the shared disk unit (CR80). Locks operate correctly with this system configuration.

(b) System configurations unable to perform system switchovers

Figure 26-43 A system configuration unable to perform system switchovers (1/2) and Figure 26-44 A system configuration unable to perform system switchovers (2/2) show system configurations that cannot perform system switchovers.

Figure 26-43 A system configuration unable to perform system switchovers (1/2)

Explanation

If an optical hub is placed between a CR80 unit and the server machines, even if server machine 1 is in reserved status, its reserved status is released when server machine 2 restarts. This occurs because server machine 2 generates a loop initialization primitive (LIP) that affects server machine 1.

Figure 26-44 A system configuration unable to perform system switchovers (2/2)

Explanation

• Because a hub is being used, the LIP that is generated when the system restarts releases the lock, causing the system to operate incorrectly.
• If a switch were used, the LIP would function independently on each server, so the system would operate properly.
• When a fiber hub is used, the LIP causes a lock error.
  

  Note

  LIPs are used on all devices that are connected in a loop. Therefore, locking might not function properly when two CR80 units are connected, or when multiple server machines and one CR80 unit are connected via a hub.

(a) When IP addresses are inherited

When IP addresses are inherited, set a logical IP address and host name that correspond to the logical IP address (host name of the IP address that is set in the IP type resource). Also, set up this logical IP address so it can be replaced on multiple server machines using the system switchover facility. The following figure shows a network configuration example for inheriting IP addresses.

Figure 26-45 Network configuration example when IP addresses are inherited (using ClusterPerfect)

Explanation

• Set a logical IP address (Service IP: 192.168.1.1) and the host name (logicalhost) that correspond to this logical IP address.
• HiRDB is installed on both server machine 1 and server machine 2.
• ClusterPerfect, which monitors server machine 1 and server machine 2, is installed on the MMC machine.
• /mirror1 is the mount point.
• Create HiRDB system files on the shared disk unit (/mirror1).

(b) When IP addresses are not inherited

When IP addresses are not inherited, set a different IP address and host name in the primary system and secondary system. To do this, you must specify the host name of the secondary system in the pdunit -c operand of the HiRDB system common definition. The following figure shows a network configuration example for when IP addresses are not inherited.

Figure 26-46 Network configuration example when IP addresses are not inherited (using ClusterPerfect)

Explanation

• HiRDB is installed in server machine 1 and server machine 2.
• ClusterPerfect, which monitors server machine 1 and server machine 2, is installed on the MMC machine.
• /mirror1 is the mount point.
• Create HiRDB system files on the shared disk unit (/mirror1).

Use the DNCWARE design support system in ClusterPerfect to create and register abnormal termination, startup, termination, and takeover scenarios. For details about creating and registering scenarios, see the ClusterPerfect documentation. The content of each scenario is described below.

(a) Abnormal termination scenario

The actions of the abnormal termination scenario are as follows:

(b) Startup scenario

The actions of the startup scenario are as follows:

(c) Termination scenario

The actions of the termination scenario are as follows:

(d) Takeover scenario

The actions of the takeover scenario are as follows:

This subsection provides examples of shells used to set up HiRDB scenarios.

(a) HiRDB single server configuration

The following examples show shells used to set up HiRDB single server configuration scenarios.

Start

#!/bin/sh
PDDIR=/hirdb_x
PATH=/bin:/usr/bin:/usr/ucb:$PDDIR/bin
PDCONFPATH=$PDDIR/conf
SHLIB_PATH=$PDDIR/lib
LD_LIBRARY_PATH=$PDDIR/lib
export PATH PDDIR PDCONFPATH SHLIB_PATH LD_LIBRARY_PATH
# single
$PDDIR/bin/pdstart
exit 0

Terminate

#!/bin/sh
PDDIR=/hirdb_x
PATH=/bin:/usr/bin:/usr/ucb:$PDDIR/bin
PDCONFPATH=$PDDIR/conf
SHLIB_PATH=$PDDIR/lib
LD_LIBRARY_PATH=$PDDIR/lib
export PATH PDDIR PDCONFPATH SHLIB_PATH LD_LIBRARY_PATH
# single
$PDDIR/bin/pdstop -f
exit 0

Mount the shared disk

#!/bin/sh
/usr/local/DNCWARE/bin/genresv /dev/sdb
sleep 10
/usr/local/DNCWARE/bin/hadkresv /dev/sdb
mount /mirror1

(b) HiRDB parallel server configuration

The following examples show shells used to set up HiRDB parallel server configuration scenarios.

Start

#!/bin/sh
PDDIR=/hirdb_x
PATH=/bin:/usr/bin:/usr/ucb:$PDDIR/bin
PDCONFPATH=$PDDIR/conf
SHLIB_PATH=$PDDIR/lib
LD_LIBRARY_PATH=$PDDIR/lib
export PATH PDDIR PDCONFPATH SHLIB_PATH LD_LIBRARY_PATH
# parallel
$PDDIR/bin/pdstart -q
exit 0

Terminate

#!/bin/sh
PDDIR=/hirdb_x
PATH=/bin:/usr/bin:/usr/ucb:$PDDIR/bin
PDCONFPATH=$PDDIR/conf
SHLIB_PATH=$PDDIR/lib
LD_LIBRARY_PATH=$PDDIR/lib
export PATH PDDIR PDCONFPATH SHLIB_PATH LD_LIBRARY_PATH
# parallel
$PDDIR/bin/pdstop -z
exit 0

Mount the shared disk

#!/bin/sh
/usr/local/DNCWARE/bin/genresv /dev/sdb
sleep 10
/usr/local/DNCWARE/bin/hadkresv /dev/sdb
mount /mirror1