How to ensure I/O prevention

After an XRF takeover, you must ensure that the failing active subsystem cannot update data sets used by the new active subsystem. If the new active subsystem can write to the databases before the failing active subsystem is prevented from changing this data, data integrity is compromised.

That is, you must ensure that the new active subsystem is in a state known as I/O prevention. I/O prevention can occur in one of the following ways, depending on the cause of the takeover:

• I/O prevention occurs automatically:
  • If IMS fails and z/OS® continues running
  • If you enter the /SWITCH SYSTEM command on the active subsystem and IMS issues abend U0604
• If z/OS or the CPC fails, ensure I/O prevention by resetting the active CPC.
• For all other causes of takeover, terminate the active IMS subsystem with the z/OS MODIFY command.

Because there is no communication between the z/OS for the active subsystem and the z/OS for the alternate subsystem, the operator for the active subsystem must inform the operator for the alternate subsystem when I/O prevention is complete.

I/O prevention is completed in one of the following ways:

• By receiving the following message from the z/OS availability manager:

  AVM006E  TELL OPERATOR AT BACKUP TO REPLY          "UNLOCK" TO MESSAGE AVM005A.
           I/O PREVENTION IS COMPLETE FOR
           SUBSYSTEM ssid,
           FAILING ACTIVE ELEMENT OF RSE rsename.

• By ensuring that the failed IMS control region and DL/I address space have terminated execution
• By ensuring that the CPC has failed
• By ensuring that the CPC has been reset
• By canceling IMS and waiting until all of the address spaces terminate
• By physically switching the DASD that contains the data sets that the active subsystem might access

If the availability manager's address space fails, IMS continues processing normally but takeover processing is degraded. You can choose to continue without the availability manager, however, you cannot reconnect IMS to the availability manager unless you terminate and restart IMS.