On Oracle x86 servers using MegaRAID disk controllers, Serial Attached SCSI (SAS) data path errors can occur. To triage and isolate a data path problem on the SAS disk controller, disk backplane (DBP), SAS cable, SAS expander, or hard disk drive (HDD), gather and review the events in the disk controller event log. Classify and analyze all failure events reported by the disk controller based on the server SAS topology.
To classify a MegaRAID disk controller event:
Gather and parse the MegaRAID disk controller event logs either by running the automated sundiag utility or manually using the MegaCLI or StorCLI commands.
For Oracle Exadata Database Machine database or storage cell servers, run the sundiag utility.
For the Sun Server X3-2, use the MegaCLI command.
For example, manually gather and parse the controller event log by using the MegaCLI command. At the root prompt, type:
root# ./MegaCli64 adpeventlog getevents –f event.log aall Success in AdpEventLogExit Code: 0x00
The following SCSI sense key errors found in the event log in SAS data path failures indicate a SAS data path fault:
B/4B/05 :SERIOUS: DATA OFFSET ERROR B/4B/03 :SERIOUS: ACK/NAK TIMEOUT B/47/01 :SERIOUS: DATA PHASE CRC ERROR DETECTED B/4B/00 :SERIOUS: DATA PHASE ERROR
A communication fault between the disk and the host bus adapter causes these errors. The presence of these errors, even on a single disk, means there is a data path issue. The RAID controller, SAS cables, SAS expander, or disk backplane might be causing the interruption to the communication in the path between the RAID controller and the disks.
Oracle Service personnel can find more information about the diagnosis and triage of hard disk and SAS data path failures on x86 servers at the My Oracle Support web site: https://support.oracle.com. Refer to the Knowledge Article Doc ID 2161195.1. If there are multiple, simultaneous disk problems on an Exadata server, Oracle Service personnel can refer to Knowledge Article Doc ID 1370640.1.