System Troubleshooting and Diagnostics
5.2 Product Fault Management and Symptom-Directed Diagnosis
In cases of a new memory module used for repair or as part of system
installation, one may elect to replace the module rather than having
diagnostics map them out, even if the threshold has not been reached for
hard single-address errors.
• MULTIPLE ADDRESSES ( )—If the second occurrence of an error within
a footprint is at a different address (LOWEST ADDRESS not equal to
HIGHEST ADDRESS (
), MULTIPLE ADDRESSES will be set in STATUS
along with SCRUBBED. Scrubbing will not be attempted for this situation.
In most cases, the failing memory module should be replaced regardless of
the page mapout threshold.
If CRD BUFFER FULL is set in LOGGING REASON (
) (located in the
subpacket header) or PAGE MAPOUT THRESHOLD EXCEEDED is set in
SYSTAT ( ), the failing memory module should be replaced regardless of any
thresholds.
For all cases (except when SCRUBBED is the only flag set in STATUS) isolate
the offending memory by examining the translation in FOOTPRINT called
MEMORY ERROR STATUS (
): The memory module is identified by its
backplane position. In Example 5–6, memory module #3, slot 2, is identified as
the failing module.
The Memory SBE Reduction Subpacket header translates the MEMCON
register (
) for memory subsystem configuration information.
Unlike uncorrectable memory and CPU errors, the OpenVMS global counter, as
shown by the DCL command SHOW ERROR, is not incremented for correctable
ECC errors unless it results in an error log entry for reasons other than system
shutdown.
Note
If footprints are being generated for more than one memory module,
especially if they all have the same bit in error, the processor module,
backplane, or other component may be the cause.
System Troubleshooting and Diagnostics 5–25
