IBM 86655RY - Netfinity 7600 - 8665 Hardware Maintenance Manual page 217

received from the target device (hard disk drive, CD ROM, tape drive, etc.) since the
Device Error Table was last cleared. There are many types of SCSI check conditions
that can be received by the ServeRAID controller. Some of the check conditions
indicate errors while others indicate unexpected (but not error) conditions such as the
command queue on a drive being temporarily full.
How should soft events be handled?: An occasional soft event in the absence of PFA,
parity and hard event entries is usually not a problem. Check for cables being
properly seated, bent pins, pushed pins, damaged cables and proper termination of
the SCSI bus. No further action is necessary if there are only a small number of
random events.
Is there a threshold guide line to follow before replacing the drive?: As with hard errors, a
pattern of entries for a particular drive or channel may indicate a problem. There is
not an absolute threshold for soft errors and when drives should be replaced. The
ServeRAID controller internally filters the types of soft errors and will mark a disk
drive defunct when appropriate.
In the Device Error Table what is the meaning of parity events?: The parity event entry in
the device event table is the number of parity or ECC errors on the SCSI bus found by
the ServeRAID controller since the last time the Device Event Table was cleared.
Parity errors found by the targets (hard disk drive, tape drive, etc.) are reported as soft
events.
How should parity events be handled?: Check for cables being properly seated, bent
pins, pushed pins, damaged cables and proper termination of the SCSI bus. If some or
all of the devices are operating at Fast or Ultra speeds, then ensure that the maximum
cable lengths for the SCSI interface are not exceeded. No further action is necessary if
there are only a small number of random events. A large number of events on a
particular channel or to a particular target may require replacement of the backplane
or SCSI cabling. It is possible but unlikely that the ServeRAID controller has caused
the parity errors.
In the Device Event Table what is the meaning of miscellaneous
events?: Miscellaneous events are all entries that are not parity, soft, hard or PFA
entries. Miscellaneous events are very often target (hard drive, tape drive, etc.)
problems.
How should miscellaneous events be handled?: Some common events that cause
miscellaneous errors are selection time-out when accessing the drive, unexpected
SCSI bus free detected by the SCSI I/O processor or SCSI phase error. Check for cables
being properly seated, bent pins, pushed pins, damaged cables and proper
termination of the SCSI bus. If all the preceding items are correct, then suspect the
target indicated in the log. The least likely cause would be a problem in the
ServeRAID controller.
What is PFA and do all drives have this capability?: PFA stands for Predictive Failure
Analysis. Most server class hard disk drives have the capability of monitoring internal
parameters in the drive that could predict a future failure in the drive. The algorithms
and data monitored are very complex and in most cases proprietary.
If a drive determines that a failure is likely, then it notifies the ServeRAID controller of
a possible future failure. This notification is included as a Device Event Table entry.
Any drive with a PFA entry in the Device Event Table should be replaced as soon as
possible.
How should PFA events be handled?: For RAID-1 or RAID-5, replace the drive
immediately and rebuild the array. For RAID-0 back up all data to tape immediately,
replace the drive and then restore the data to the RAID-0 array.
207
Installing and configuring ServeRAID controllers