Bug ID: 27759886 Fixed in Firmware Release RF30
Oracle 6.4 TB NVMe SSD v1 may report uncorrectable errors or assert after not being powered for three or more months. For best practice, secure erase Oracle 6.4 TB NVMe SSD v1s before use (especially if use is reading from the drive as a test) and especially if the drive has been unpowered for more than three months. If the NAND media is not refreshed for approximately three months, the drive may experience media errors.
Over time, the drive firmware policy refreshes the media in the background while it remains powered-on. If the drive has been powered on long enough for the background refresh policy to be applied to all bits, the drive is not at risk for this issue. The time required to refresh all the bits is approximately 14 days and varies by product.
If the number of bits experiencing this issue exceeds the error-correction code (ECC) capability, it may result in an uncorrectable read error. If the uncorrectable read errors occur during normal drive operation, the drive will report an increased number of SMART media errors to the host. If the uncorrectable read errors occur during drive power-on, the drive will report either an ASSERT or BAD_CONTEXT error code to the host.
The following screen shows an ASSERT or BAD_CONTEXT event at power-on after the media has not been refreshed for a time.
Firmware QDV1RD28: ASSERT_100452A0, BAD_CONTEXT_1042, or BAD_CONTEXT_1043
Workaround:
Select one of the following methods before use of the drive for operation or test. An off-line server can be used.
Download and use RF30 mitigation firmware to secure erase the drive, using the nvmeadmin utility. See Secure Erase Oracle 6.4 TB NVMe SSD v1.
Wait two weeks for a media refresh while the drive is powered-on before using the drive.
Caution - All device data will be destroyed after an erase. |
If immediate refresh of all bits is desired, secure erase the drive, using the Oracle Hardware Management Pack NVMe admin utility.
For more information on CLI commands, refer to Oracle Hardware Management Pack 2.4 Server CLI Tools User's Guide: https://www.oracle.com/goto/ohmp/docs. See Server Management Tools.
Stop all IO to the NVMe device before attempting this action. Manually quiesce I/O and device usage.
List all server devices. Type: # fwupdate list controller. For example:
# fwupdate list controller ================================================== CONTROLLER ================================================== ID Type Manufacturer Model Product Name FW Version --------------------------------------------------------------------------- c0 SAS LSI Logic 0x0097 ORACLE-T7 05.00.00.00 c1 SAS LSI Logic 0x0097 SAS9300-8e 13.00.00.00 c2 SAS LSI Logic 0x0097 ORACLE-T7 05.00.00.00 c3 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c4 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c5 NVMe Intel 0x0a54 7335940:ICDPC2DD2ORA6.4T 8DV1RF35 c6 NVMe Intel 0x0a54 7335940:ICDPC2DD2ORA6.4T 8DV1RF35 c7 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c8 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c9 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c10 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c11 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c12 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c13 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35 c14 NVMe Intel 0x0a54 7335943:ICDPC5ED2ORA6.4T 8DV1RF35
Back up drive user data, if necessary.
Download the RF30 mitigation firmware image, if necessary.
See Download the Device Software Package at Accessing Software Updates and Firmware Downloads.
Update the affected drive with the RF30 mitigation firmware, if necessary. See Update the NVMe Storage Drive Firmware.
Type: # fwupdate update nvme-controller-firmware -f <FW file> -n <controller ID> . For example:
# fwupdate update nvme-controller-firmware -f ICDPC2DD2.RF30.fw -n c3 The following components will be upgraded as shown: ========================================================== ID Priority Action Status Old Firmware Ver. Proposed Ver. New Firmware Ver. System Reboot ------------------------------------------------------------------------------------------------------------------------ c3 1 Check FW Success QDV1RF30 Not Provided N/A N/A Do you wish to process all of the above component upgrades? [y/n]? y Updating c3: Success Verifying all priority 1 updates No metadata provided, so version verification can not be completed Execution Summary ========================================================== ID Priority Action Status Old Firmware Ver. Proposed Ver. New Firmware Ver. System Reboot ------------------------------------------------------------------------------------------------------------------------ c3 1 Validate Pending QDV1RF30 Not Provided 8DV1RF35 N/A
Power down the host using Oracle ILOM. Type: # fwupdate reset controller -n <controller ID>. Alternately, hotplug disable and remove the devices and re-insert them.
Reset the affected drive. Type: # fwupdate reset controller -n <controller ID>.
Check NVMe device status to identify NVMe controllers and current firmware versions. Type #nvmeadm list -v. For example:
# nvmeadm list -v
SUNW-NVME-1
PCI Vendor ID: 0x8086
Serial Number: PHLE713400T56P4BGN-1
Model Number: 7335943:ICDPC5ED2ORA6.4T
Firmware Revision: 8DV1RF35
Number of Namespaces: 1
SUNW-NVME-2
PCI Vendor ID: 0x8086
Serial Number: PHLE713400T56P4BGN-2
Model Number: 7335943:ICDPC5ED2ORA6.4T
Firmware Revision: 8DV1RF35
Number of Namespaces: 1
SUNW-NVME-3
PCI Vendor ID: 0x8086
Serial Number: PHLE7464009N6P4OGN
Model Number: 7335940:ICDPC2DD2ORA6.4T
Firmware Revision: 8DV1RF35
Number of Namespaces: 1
SUNW-NVME-4
PCI Vendor ID: 0x8086
Serial Number: PHLE7464009L6P4OGN
Model Number: 7335940:ICDPC2DD2ORA6.4T
Firmware Revision: 8DV1RF35
Number of Namespaces: 1
Take the drive offline. Type: # nvmeadm offline -n <namespace> <controller name> namespacecontroller_name. The namespace is 1 in the following example:
# nvmeadm offline -n 1 SUNW-NVME-3
Securely erase the affected drive three times.
To securely erase all namespaces (SES=1), type: # nvmeadm erase -a controller_name. For example:
# nvmeadm erase -a SUNW-NVME-3 SUNW-NVME-3 Erase data on all namespaces (Y/N)?y Erase device SUNW-NVME-3 successfully.
Repeat this command two more times to securely erase all namespaces. For example:
# nvmeadm erase -a SUNW-NVME-3 SUNW-NVME-3 Erase data on all namespaces (Y/N)?y Erase device SUNW-NVME-3 successfully. # nvmeadm erase -a SUNW-NVME-3 SUNW-NVME-3 Erase data on all namespaces (Y/N)?y Erase device SUNW-NVME-3 successfully.
Place the drive online. Type: # nvmeadm online -n <namespace> <controller name> namespacecontroller_name. The namespace is 1 in the following example:
# nvmeadm online -n 1 SUNW-NVME-3
Identify the affected drive block device name of SUNW-NVME-3 using the nvmeadm namespace command. The Block Device Name is /dev/rdsk/c28t1d0s2 in the following example:
# nvmeadm namespace -v SUNW-NVME-3
SUNW-NVME-3
Namespace: 1
Block Size: 512
Capacity: 6401252745216
Metadata Size: 0
Block Device Name: /dev/rdsk/c28t1d0s2
Status: online
Verify that the device status is online.
Execute more than 490 GB of sequential write to the affected drive. For example:
# dd if=/dev/zero of=/dev/rdsk/c28t1d0p0 bs=131072 count=3814698 3814698+0 records in 3814698+0 records out
The drive Block Device Name is /dev/rdsk/c28t1d0s2. Use the Block Device Name that represents the entire drive to utilize the raw device as shown in the above example.
Verify drive health. See Verify Oracle 6.4 TB NVMe SSD v1 Operation.
Restore user data, if necessary.