To use reduced activation, 1 must be specified in the pd_start_level operand. HiRDB will start using only the available units if there are units that cannot be started.

When reduced activation takes effect, the KFPS05217-I message is issued. The pdls command can be used to determine the units that could not be started.

Remarks

During startup, HiRDB waits for up to 20 minutes to receive a startup message from all units. Therefore, there is a delay of 20 minutes before reduced activation is executed. The maximum of 20 minutes to wait to receive a startup message can be changed in the pd_reduced_check_time operand of the system common definition.

The following procedure shows how to return HiRDB to the normal mode after it has run in reduced activation.

Procedure

To return to normal operation mode:

1. Use the pdls command to check and eliminate the error in the unit that cannot be started.
   pdls -d svr
   
2. Use the pdstart -u command to start up a unit that is not active:
   pdstart -u unt1
   When HiRDB returns to the normal operation mode, the KFPS05218-I message is issued.
   

In the following cases, reduced activation is not available; the cause of the error must be eliminated in order to start HiRDB:

• The unit to be started does not have either of the following servers:
  
  • System manager
  • Dictionary server
• The unit to be started has none of the following servers:
  
  • Front-end server
  • Back-end server
• The pdstart -i command is being used to start HiRDB.
• HiRDB is being started for the first time after being upgraded.

If an attempt is made to access a table at an inactive unit, an SQL error results. If a table is row-partitioned, accessing data stored at an inactive unit results in an SQL error.

If reduced activation is used during a restart, a transaction related to the erroneous unit might be placed in uncompleted status. The transaction will be determined automatically when reduced activation is released and normal operation is started. If there is a transaction in uncompleted status, neither normal termination nor planned termination can be executed on HiRDB. In such a case, the pdstop -f command must be used to terminate HiRDB forcibly.

Special care is needed when the front-end server to which a client user is to be connected has been determined; that is, when the following operands have been specified in the client environment definition:

• PDFESHOST
• PDSERVICEGRP

If the front-end server specified in these operands is not running, no UAPs can be executed. To enable UAPs to execute, either delete these operands or specify a front-end server that is running.

(a) HiRDB used at the extracting side

• Data extraction processing is not performed on a back-end server in an inactive unit. For such a case, once the unit is running normally again, use the hdestart command of HiRDB Datareplicator to restart data extraction processing. If, however, normal operations resume when HiRDB is restarted after it had been terminated while it was in reduced activation mode, you do not need to enter the hdestart command. Data extraction processing will restart automatically in such a case.
• If data extraction processing stops during reduced activation, execute the pdrplstop command. Avoid executing the pdrplstop -f command, if possible. When you execute the pdrplstop command, data execution processing will stop on completion if there are incomplete extraction status system log files in the unit that has been returned to normal operation. However, if you execute the pdrplstop -f command, data extraction processing stops only in conjunction with the entire system, even if there are incomplete extraction status system log files in the unit that has been returned to normal operation.

(b) HiRDB used at the target side

• If an attempt is made to reflect data in an inactive unit, an error results and the reflection processing is canceled. In such a case, data reflection processing must be restarted with HiRDB Datareplicator's hdsstart command after reduced activation is released and normal operation is started.
• Re-create the target database based on the database in the unit that is inactive due to reduced activation. In such as case, however, mismatches might arise between the extracted database and the target database.

If reduced activation is to be used for a subsequent HiRDB startup because a unit error cannot be corrected, the name of the unit that cannot be started should be specified in the pd_start_skip_unit operand. Normally during startup, HiRDB waits for up to 20 minutes to receive a startup message from each unit. Therefore, there is a delay of 20 minutes before reduced activation is executed. If the name of a unit that cannot be started is specified in the pd_start_skip_unit operand, HiRDB does not wait for a unit startup message from that unit, which can save a considerable amount of time that HiRDB would otherwise spend waiting for a startup message, thereby reducing the time required for reduced activation.

Example

1. 1 is specified in the pd_start_level operand in the system common definition.
2. HiRDB executed reduced activation because a unit could not be started due to an error.
3. When the application terminated, the pdstop command was entered to terminate HiRDB. The error had not been corrected at this point.
4. The following operand was specified:
   pd_start_skip_unit=name-of-unit-that-cannot-be-started
   
5. To start the application, the pdstart command was entered to start HiRDB. The HiRDB startup time is reduced as compared to when the pd_start_skip_unit is not specified.
   

   Note

   To return to normal operation, delete the pd_start_skip_unit operand. If you do not delete this operand, HiRDB will again be placed in reduced activation mode the next time it is started.

Recovery-unnecessary front-end server units start independently in the reduced mode, regardless of the value set in the pd_start_level operand. Even if you specify a unit name in the pd_start_skip_unit operand, that specification is ignored and the recovery-unnecessary front-end server units are started. If any of the recovery-unnecessary front-end server units cannot be started during HiRDB startup, HiRDB is started without those units.