!!ACHINE CHECK HANDLER (!tCH) You can set the recording mode to record errors corrected by processor
retry (logically termed as CPU retry) and Error correction Code (ECC)
wi th the SET !tODE command,. In attached Frocesscr applica tions,
recording .ode can be set for either or both prccessors. For processor
retry, the default setting is record mode. !Qte: The SET !!ODE !!AIN command is invalid for 3031, 3032, and 3033 processors. When processor retry or ECC succeed in correcting errors, and the
processor 1S 1n record mode, the machine check handler records the
error. When processor retry or ECC fail, the machine check handler: Attempts to isolate the failure to one page frame and makes that page
frame invalid or unavailable for paging. Attempts to isolate the failure to one virtual machine and logs off
or resets that virtual machine. Attempts to isolate the failure to portions of the system and to
continue system operation in degraded mode. Abnormally terminates the system when recovery is not possible; or,
if V8/370 is operating in attached processor mode and the malfunction
is isolated to the attached processor and to a particular virtual
lIachine, then, system operation continues in uniprocessor Jlode,. MCH records an error whenever any of the following conditions occur: Processor retry occurred'. ECC corrected datal. Hardware reported a buffer or DLAT (Data Look Aside Table) error. Multiple-bit storage failure. External da,age. Storage protection feature damage. Timer error. System damage. Instruction processor damage. CHANNEL CHECK HANDLER CCCH) Whenever a channel control check, channel data check, or interface
control check occurs, the channel check handler (CCH) constructs an
error record and records the results in an IOERBLOK. The error recovery
procedures use this IOERBLOK to retry the error. Recovery is not attempted for channel errors associated with virtual machine I/O events.
1 VM/370 records these errors only under specific conditions,. The
conditions for recording these errors are detailed in the !nd Section 1. Intro. to Operational Ctrl. of the VM/370 System 3
I/O Error Recording and SVC 76 V8/370 maintains an error recording area that caFtures I/O, CCH, and BCB error records. Device and control unit detected unit checks during V8/370 spooling, paging, and virtual machine I/O errors generate the I/O records. V8/370 and the virtual machine's LOGREC data set contain recorded I/O errors; this double recording occurs when the virtual machine's
operating system does not invoke SVC 76.
If the virtual machine operating system invokes SVC 76 and passes the
correct parameters to V8/310, V8/370 records the error in its own error
recording area. VM/370 then passes control back to the virtual machine operating system, thus bypassing virtual machine error recording
facilities. VM/370 Recovery Features
The V!/370 recovery features are described more fully in the VBL11Q and RECORDING FACILITIES
The OS/VS Environmental Recording, Editing, and printing program (EREP) is executed when the CMS CPEREP command is invoked. The output of the CPEREP command consists of printed reports whose content depends upon
the specified (or defaulted) CPEREP operands and upon the input system error records. The reports generated by CPEREP have the salle format as
those generated on an OS/VS system. The input system error records aay be froa the V8/370 error recording area or fre. a history tape. The
history tape may have been produced earlier by CPEREP froa the V8/370 error recording area data or by an OS/VS system from SYS1.LOGREC data. Unlabeled tapes produced on OS/VS systems by OS/VS EREP and on V!/370 systeas by CPEREP are compatible and can be transported between systeas.
Data froll both systems can also be accumulated on the same tape,. For aore details on CPEREP, refer to the following publications: VBL11Q gnd l!!Q! and the f!inting If the facilities of an IB! 3850 Mass Storage System (!SS) are used
with VM/370 virtual machine operations and !SS errors are reflected to V8/370's error recording area, CPEREP aust be invoked so that 8SS-related errors recorded in the error recording area can be collected
on an accumulation (ACC=YES) tape for further processing by the VS System Data Analyzer Program (SDA). Because ess logged-out data is
voluminous and the interrelationships of !SS components are complex, it
is imperative that this service program be used to effectively diagnose
and isolate mass storage problems. 4 V8/370 Operator's Guide
Previous Page Next Page