An erroneous partner crash detection occurs if a partner is declared as crashed without justification, either by systems support by answering message MCS1100 with “MTERM” or “CRASH” or by the system itself (see “MSCF configuration parameter RECOVERY-START” (Global control parameters)), even though the processor is still active. Since the global resources of the partner can be reused on the remaining processors, inconsistencies may arise. If a processor discovers when the MSCF connection is reestablished or by reading the vital-sign messages that a partner was falsely declared as crashed, configuration management disconnects from the network and HIPLEX MSCF terminates.
If two master processors have nonetheless been set up as a result of an error, the integrity of the shared pubset is no longer guaranteed; systems support must export the pubset immediately from all network sharers.
If data inconsistencies arise on the global resources, shared pubsets etc., it may be necessary to revert to an older backup version before MSCF is restarted.