diff options
Diffstat (limited to 'Documentation/nvme-wdc-vs-smart-add-log.txt')
| -rw-r--r-- | Documentation/nvme-wdc-vs-smart-add-log.txt | 268 |
1 files changed, 268 insertions, 0 deletions
diff --git a/Documentation/nvme-wdc-vs-smart-add-log.txt b/Documentation/nvme-wdc-vs-smart-add-log.txt new file mode 100644 index 0000000..a9c4ead --- /dev/null +++ b/Documentation/nvme-wdc-vs-smart-add-log.txt @@ -0,0 +1,268 @@ +nvme-wdc-vs-smart-add-log(1) +============================ + +NAME +---- +nvme-wdc-vs-smart-add-log - Send NVMe WDC vs-smart-add-log Vendor Unique Command, return result + +SYNOPSIS +-------- +[verse] +'nvme wdc vs-smart-add-log' <device> [--interval=<NUM>, -i <NUM>] [--output-format=<normal|json> -o <normal|json>] + +DESCRIPTION +----------- +For the NVMe device given, send a Vendor Unique WDC vs-smart-add-log command and +provide the additional smart log. The --interval option will return performance +statistics from the specified reporting interval. + +The <device> parameter is mandatory and may be either the NVMe character +device (ex: /dev/nvme0). + +This will only work on WDC devices supporting this feature. +Results for any other device are undefined. + +On success it returns 0, error code otherwise. + +OPTIONS +------- +-i <NUM>:: +--interval=<NUM>:: + Return the statistics from specific interval, defaults to 14 + +-o <format>:: +--output-format=<format>:: + Set the reporting format to 'normal', or + 'json'. Only one output format can be used at a time. + Default is normal. + +Valid Interval values and description :- + +[cols="2*", frame="topbot", align="center", options="header"] +|=== +|Value |Description + +|*1* +|Most recent five (5) minute accumulated set. + +|*2-12* +|Previous five (5) minute accumulated sets. + +|*13* +|The accumulated total of sets 1 through 12 that contain the previous hour of +accumulated statistics. + +|*14* +|The statistical set accumulated since power-up. + +|*15* +|The statistical set accumulated during the entire lifetime of the device. +|=== + +CA Log Page Data Output Explanation +----------------------------------- +[cols="2*", frame="topbot", align="center", options="header"] +|=== +|Field |Description + +|*Physical NAND bytes written.* +|The number of bytes written to NAND. 16 bytes - hi/lo + +|*Physical NAND bytes read* +|The number of bytes read from NAND. 16 bytes - hi/lo + +|*Bad NAND Block Count* +|Raw and normalized count of the number of NAND blocks that have been +retired after the drives manufacturing tests (i.e. grown back blocks). +2 bytes normalized, 6 bytes raw count + +|*Uncorrectable Read Error Count* +|Total count of NAND reads that were not correctable by read retries, all +levels of ECC, or XOR (as applicable). 8 bytes + +|*Soft ECC Error Count* +|Total count of NAND reads that were not correctable by read retries, or +first-level ECC. 8 bytes + +|*SSD End to End Detection Count* +|A count of the detected errors by the SSD end to end error correction which +includes DRAM, SRAM, or other storage element ECC/CRC protection mechanism (not +NAND ECC). 4 bytes + +|*SSD End to End Correction Count* +|A count of the corrected errors by the SSD end to end error correction which +includes DRAM, SRAM, or other storage element ECC/CRC protection mechanism (not +NAND ECC). 4 bytes + +|*System Data % Used* +|A normalized cumulative count of the number of erase cycles per block since +leaving the factory for the system (FW and metadata) area. Starts at 0 and +increments. 100 indicates that the estimated endurance has been consumed. + +|*User Data Max Erase Count* +|The maximum erase count across all NAND blocks in the drive. 4 bytes + +|*User Data Min Erase Count* +|The minimum erase count across all NAND blocks in the drive. 4 bytes + +|*Refresh Count* +|A count of the number of blocks that have been re-allocated due to +background operations only. 8 bytes + +|*Program Fail Count* +|Raw and normalized count of total program failures. Normalized count +starts at 100 and shows the percent of remaining allowable failures. +2 bytes normalized, 6 bytes raw count + +|*User Data Erase Fail Count* +|Raw and normalized count of total erase failures in the user area. +Normalized count starts at 100 and shows the percent of remaining +allowable failures. 2 bytes normalized, 6 bytes raw count + +|*System Area Erase Fail Count* +|Raw and normalized count of total erase failures in the system area. +Normalized count starts at 100 and shows the percent of remaining +allowable failures. 2 bytes normalized, 6 bytes raw count + +|*Thermal Throttling Status* +|The current status of thermal throttling (enabled or disabled). +2 bytes + +|*Thermal Throttling Count* +|A count of the number of thermal throttling events. 2 bytes + +|*PCIe Correctable Error Count* +|Summation counter of all PCIe correctable errors (Bad TLP, Bad +DLLP, Receiver error, Replay timeouts, Replay rollovers). 8 bytes +|=== + + +C1 Log Page Data Output Explanation +----------------------------------- +[cols="2*", frame="topbot", align="center", options="header"] +|=== +|Field |Description + +|*Host Read Commands* +|Number of host read commands received during the reporting period. + +|*Host Read Blocks* +|Number of 512-byte blocks requested during the reporting period. + +|*Average Read Size* +|Average Read size is calculated using (Host Read Blocks/Host Read Commands). + +|*Host Read Cache Hit Commands* +|Number of host read commands that serviced entirely from the on-board read +cache during the reporting period. No access to the NAND flash memory was required. +This count is only updated if the entire command was serviced from the cache memory. + +|*Host Read Cache Hit Percentage* +|Percentage of host read commands satisfied from the cache. + +|*Host Read Cache Hit Blocks* +|Number of 512-byte blocks of data that have been returned for Host Read Cache Hit +Commands during the reporting period. This count is only updated with the blocks +returned for host read commands that were serviced entirely from cache memory. + +|*Average Read Cache Hit Size* +|Average size of read commands satisfied from the cache. + +|*Host Read Commands Stalled* +|Number of host read commands that were stalled due to a lack of resources within +the SSD during the reporting period (NAND flash command queue full, low cache page count, +cache page contention, etc.). Commands are not considered stalled if the only reason for +the delay was waiting for the data to be physically read from the NAND flash. It is normal +to expect this count to equal zero on heavily utilized systems. + +|*Host Read Commands Stalled Percentage* +|Percentage of read commands that were stalled. If the figure is consistently high, +then consideration should be given to spreading the data across multiple SSDs. + +|*Host Write Commands* +|Number of host write commands received during the reporting period. + +|*Host Write Blocks* +|Number of 512-byte blocks written during the reporting period. + +|*Average Write Size* +|Average Write size calculated using (Host Write Blocks/Host Write Commands). + +|*Host Write Odd Start Commands* +|Number of host write commands that started on a non-aligned boundary during +the reporting period. The size of the boundary alignment is normally 4K; therefore +this returns the number of commands that started on a non-4K aligned boundary. +The SSD requires slightly more time to process non-aligned write commands than it +does to process aligned write commands. + +|*Host Write Odd Start Commands Percentage* +|Percentage of host write commands that started on a non-aligned boundary. If this +figure is equal to or near 100%, and the NAND Read Before Write value is also high, +then the user should investigate the possibility of offsetting the file system. For +Microsoft Windows systems, the user can use Diskpart. For Unix-based operating systems, +there is normally a method whereby file system partitions can be placed where required. + +|*Host Write Odd End Commands* +|Number of host write commands that ended on a non-aligned boundary during the +reporting period. The size of the boundary alignment is normally 4K; therefore this +returns the number of commands that ended on a non-4K aligned boundary. + +|*Host Write Odd End Commands Percentage* +|Percentage of host write commands that ended on a non-aligned boundary. + +|*Host Write Commands Stalled* +|Number of host write commands that were stalled due to a lack of resources within the +SSD during the reporting period. The most likely cause is that the write data was being +received faster than it could be saved to the NAND flash memory. If there was a large +volume of read commands being processed simultaneously, then other causes might include +the NAND flash command queue being full, low cache page count, or cache page contention, etc. +It is normal to expect this count to be non-zero on heavily utilized systems. + +|*Host Write Commands Stalled Percentage* +|Percentage of write commands that were stalled. If the figure is consistently high, then +consideration should be given to spreading the data across multiple SSDs. + +|*NAND Read Commands* +|Number of read commands issued to the NAND devices during the reporting period. +This figure will normally be much higher than the host read commands figure, as the data +needed to satisfy a single host read command may be spread across several NAND flash devices. + +|*NAND Read Blocks* +|Number of 512-byte blocks requested from NAND flash devices during the reporting period. +This figure would normally be about the same as the host read blocks figure + +|*Average NAND Read Size* +|Average size of NAND read commands. + +|*NAND Write Commands* +|Number of write commands issued to the NAND devices during the reporting period. +There is no real correlation between the number of host write commands issued and the +number of NAND Write Commands. + +|*NAND Write Blocks* +|Number of 512-byte blocks written to the NAND flash devices during the reporting period. +This figure would normally be about the same as the host write blocks figure. + +|*Average NAND Write Size* +|Average size of NAND write commands. This figure should never be greater than 128K, as +this is the maximum size write that is ever issued to a NAND device. + +|*NAND Read Before Write* +|This is the number of read before write operations that were required to process +non-aligned host write commands during the reporting period. See Host Write Odd Start +Commands and Host Write Odd End Commands. NAND Read Before Write operations have +a detrimental effect on the overall performance of the device. +|=== + + +EXAMPLES +-------- +* Has the program issue WDC vs-smart-add-log Vendor Unique Command with default interval (14) : ++ +------------ +# nvme wdc vs-smart-add-log /dev/nvme0 +------------ + +NVME +---- +Part of the nvme-user suite. |