Configuration File Description

Separate: job isolation
SubHealth: job subhealth

Resumable training ignores collateral faults caused by special faults in associated scenarios. ClusterD obtains special faults and process them based on the associated fault policy configured in the relationFaultCustomization.json and faultDuration.json files.

relationFaultCustomization.json and faultDuration.json are system configuration files. Do not modify them unless otherwise required.

**Table 1** relationFaultCustomization file description
Parameter	Description	Value
TriggerFault	Collateral fault code. Currently, fault codes configured in faultCode.json and SwitchFaultCode.json are supported.	String
RelationFaults	List of faults to be associated, which can be one or more fault codes. Currently, fault codes configured in faultCode.json and SwitchFaultCode.json are supported.	String list
FaultStrategy	Processing policy of a job when the associated fault is successfully matched. Separate: job isolation SubHealth: job subhealth	String
Note: When a fault configured by RelationFaults occurs, ClusterD adds the fault to the fault code queue to be processed. If the fault corresponding to TriggerFault occurs within the interval configured by TimeOutInterval, a job is processed based on the configured FaultStrategy. If the interval exceeds the value of TimeOutInterval, the interconnect device fault is processed using the SubHealth policy. If a processor fault or parameter plane network fault occurs, the fault is ignored.

**Table 2** faultDuration.json file description
Parameter	Description	Value
FaultCode	Fault code. Currently, fault codes configured in faultCode.json and SwitchFaultCode.json are supported.	String
FaultType	Fault type: faultDevice: processor or parameter plane network fault faultSwitch: interconnect device fault	String
TimeOutInterval	Maximum association time of a fault code, in seconds.	Integer

Parent topic: Associated Faults