SCSI Path DBIO Holdoff (and Other) Errors Continue After Service Action (G06.15 and Later RVUs)

SCSI Path DBIO Holdoff (or Other) Errors
Continue After Service Action (G06.15 and
Later RVUs)
Symptom
After a service action to resolve a ServerNet fabric error between a
controller and a processor is completed, events continue to be reported for
a period of time, even if the problem has already been fixed. These events
are event number 138 (%H8A) associated with the SCSI Module Driver
(SMD) subsystem.
Product Version or RVU
G06.15 and later.
Problem
Starting with G06.15, SMD has an enhancement to help reduce problems
when a fabric failure causes controllers to get errors on Direct Bulk I/O
(DBIO) transfers. (These are often reported as SCSI Controller DMA Error.)
When a failure during a DBIO transfer is believed to be a ServerNet fabric
failure, SMD starts a timer for the processor that was the target of the DBIO
request. This begins the DBIO holdoff period. The DBIO holdoff period is 5
minutes for G06.15 T1076AAV but is subject to change in future RVUs.
During the DBIO holdoff period for each processor, SMD rejects any DBIO
request specifying that processor as a target, returning a new syndrome to
the input/output process (IOP). This new syndrome appears in the events
posted by the IOP.
Fix
Wait until the DBIO holdoff period is over before attempting any additional
service actions to fix a ServerNet fabric error. Do not perform any service
action based on "SCSI DBIO Path Hold Off"; this event indicates that SMD
is attempting to improve overall system performance by refusing to send
DBIO requests to a controller.
Related Information
DBIO path holdoff is also described in the following:
Operator Messages Manual.●
"T1076G05 in G06.15 SCSI Module Driver" Software Release
Document, located in the G06.15 Software Archive library.
●