Nonstop Database, HiRDB Version 9 System Operation Guide
When a system switchover for the system manager unit occurs while there is a stopped unit in the system, it is not possible to start the system manager unit at the switching destination. As a result, the system switchover for the system manager unit fails. However, the system switchover for the system manager unit will not fail if the only stopped unit is a recovery-unnecessary front-end server unit.
If you take either of the following actions, system switchovers will not fail, even when there is a stopped unit in the system:
The required conditions and operation timing restrictions depend on which of these actions you use. Choose the action that is appropriate for your system.
When notwait is specified in the pd_ha_mgr_rerun operand, HiRDB does not wait to receive a startup processing completion notification from each unit during a system switchover for the system manager unit (at the time of startup processing at the switching destination). The following table describes how HiRDB processing differs depending on the value specified in the pd_ha_mgr_rerun operand.
Table 26-41 Processing by HiRDB depending on the value specified in the pd_ha_mgr_rerun operand
Item | pd_ha_mgr_rerun value | |
---|---|---|
wait (default value) | notwait | |
Whether a system switchover for the system manager unit can occur when there is a stopped unit. | The system switchover cannot be executed (system switchover fails). | The system switchover can be executed.#1, #2 |
When notwait is specified in the pd_ha_mgr_rerun operand, the system must be configured so that all the conditions listed below are satisfied. If you cannot satisfy the conditions, see (2) Using reduced activation.
Use the following procedure to specify the environment settings.
When a system switchover for the system manager unit is complete, the KFPS05210-I message (system startup completion message) is output, even if the HiRDB operating environment is not complete (for example, no front-end server is active). Consequently, if jobs cannot be executed after completion of a system switchover for the system manager unit, a UAP could terminate in an error. For this reason, use the pdls command to check the operating status of each server.
If a system switchover for the system manager unit occurs while HiRDB is starting (or terminating), the system manager unit starts (or terminates) regardless of the operating status of other units. Therefore, units other than the system manager unit might not be able to start (or terminate). If this happens, you must forcibly terminate HiRDB.
You can avoid this situation by taking the following actions when you start or terminate HiRDB:
After a system switchover, you can execute the pdls -d svr command for a unit that was stopped at the time of the system switchover for the system manager unit. STOP(N), indicating normal termination status following execution of the pdstop command, will be displayed as long as the unit remains stopped.
If a network error causes the unit to remain in STOP(N) status, even though it has actually started, first resolve the network error, terminate the unit, and then restart it.
When there is a stopped unit in the system, system switchovers for the system manager unit can be performed if the following operands are specified:
The following table describes how to specify these operands, and it shows the actions HiRDB takes during a system switchover.
Table 26-42 Operands related to reduced activation and actions HiRDB takes during a system switchover
Condition | Actions HiRDB takes when a system switchover occurs for the system manager unit | |
---|---|---|
pd_start_level specification value | pd_start_skip_unit specification | |
0 (default value) |
-- | If there is a stopped unit, a system switchover for the system manager unit fails. |
1 | Not specified | Even if there is a stopped unit, a system switchover for the system manager unit executes. However, a wait occurs for receipt of a startup processing completion notification from each unit. This notification wait time is specified in the pd_reduced_check_time operand (the default value is 20 minutes). |
Specified | If the stopped unit is specified in the pd_start_skip_unit operand, a system switchover for the system manager unit executes. No notification wait time occurs. |
When you specify the pd_start_skip_unit operand, you must terminate HiRDB. Therefore, consider (1) Specifying the pd_ha_mgr_rerun operand if your system must remain active around the clock.
All Rights Reserved. Copyright (C) 2011, 2015, Hitachi, Ltd.