6.3.4.3 Monitoring Exadata Smart Flash Log Using Exadata Metrics
Exadata metrics that are related to Exadata Smart Flash Log provide information about flash log utilization, such as the number of
megabytes written per second. Flash Log metrics are identified in the
METRICCURRENT
, METRICDEFINITION
, and
METRICHISTORY
objects as having
objectType=FLASHLOG
.
Example 6-4 Displaying Flash Log Metric Definitions
This example shows how to display the flash log metric definitions that are available in the Oracle Exadata System Software.
CellCLI> LIST METRICDEFINITION ATTRIBUTES NAME,DESCRIPTION WHERE OBJECTTYPE = FLASHLOG
FL_ACTUAL_OUTLIERS "The number of times redo writes to flash and disk both exceeded the outlier threshold"
FL_BY_KEEP "The amount of redo data saved on flash due to disk I/O errors"
FL_DISK_FIRST "The number of times redo writes first completed to disk"
FL_DISK_IO_ERRS "The number of disk I/O errors encountered by Smart Flash Logging"
FL_EFFICIENCY_PERCENTAGE "The efficiency of Smart Flash Logging expressed as a percentage"
FL_EFFICIENCY_PERCENTAGE_HOUR "The efficiency of Smart Flash Logging over the last hour expressed as a percentage"
FL_FLASH_FIRST "The number of times redo writes first completed to flash"
FL_FLASH_IO_ERRS "The number of flash I/O errors encountered by Smart Flash Logging"
FL_FLASH_ONLY_OUTLIERS "The number of times redo writes to flash exceeded the outlier threshold"
FL_IO_DB_BY_W "The number of MB written to hard disk by Smart Flash Logging"
FL_IO_DB_BY_W_SEC "The rate which is the number of MB per second written to hard disk by Smart Flash Logging"
FL_IO_FL_BY_W "The number of MB written to flash by Smart Flash Logging"
FL_IO_FL_BY_W_SEC "The rate which is the number of MB per second written to flash by Smart Flash Logging"
FL_IO_TM_W "Cumulative latency of all redo log writes"
FL_IO_TM_W_RQ "Average latency of all redo log writes"
FL_IO_W "The number of writes serviced by Smart Flash Logging"
FL_IO_W_SKIP_BUSY "The number of redo writes that could not be serviced by Smart Flash Logging because too much data had not yet been written to disk"
FL_IO_W_SKIP_BUSY_MIN "The number of redo writes during the last minute that could not be serviced by Smart Flash Logging because too much data had not yet been written to disk"
FL_IO_W_SKIP_DISABLED_GD "The number of redo writes that could not be serviced by Smart Flash Logging because it was disabled for the redo log's grid disk"
FL_IO_W_SKIP_IORM_LIMIT "The number of redo writes that could not be serviced by Smart Flash Logging because the IORM limit had been reached for the redo log's grid disk"
FL_IO_W_SKIP_IORM_PLAN "The number of redo writes that could not be serviced by Smart Flash Logging because it was disabled by the IORM plan"
FL_IO_W_SKIP_LARGE "The number of large redo writes that could not be serviced by Smart Flash Logging because the size of the data was larger than the amount of available space on any flash disk"
FL_IO_W_SKIP_LOG_ON_FAST_DEV "The number of redo writes that bypassed Smart Flash Logging because the redo log resides on a fast device"
FL_IO_W_SKIP_NO_BUFFER "The number of redo writes that could not be serviced by Smart Flash Logging because of lack of available buffers"
FL_IO_W_SKIP_NO_FL_DISKS "The number of redo writes that could not be serviced by Smart Flash Logging because there were no available Flash Log disks"
FL_PREVENTED_OUTLIERS "The number of times redo writes to disk exceeded the outlier threshold; these would have been outliers had it not been for Smart Flash Logging"
FL_RQ_TM_W "Cumulative latency of all redo log write requests (includes network and other processing overhead)"
FL_RQ_TM_W_RQ "Average latency of all redo log write requests"
FL_RQ_W "The number of redo log write requests serviced (includes requests which were not handled by Smart Flash Logging)"
FL_SKIP_OUTLIERS "The number of times redo writes to disk exceeded the outlier threshold when Smart Flash Logging was not used"
Note the following additional details:
-
FL_RQ_TM_W
tracks the cumulative redo log write request latency, which includes networking and other overhead. To determine the overhead component, get the latency overhead due to factors such as network and processing, you can use subtractFL_IO_TM_W
fromFL_RQ_TM_W
. -
FL_SKIP_OUTLIERS
tracks the number of outliers when a redo log write skips using the flash log. An outlier is a redo log write that exceeds 0.5 seconds.Additionally, metrics starting with
FL_IO_W_SKIP
track other situations where the flash log is not used. - For simultaneous writes to disk and flash storage, the disk
controller write cache can absorb some writes quicker than flash. Consequently,
it is normal for a significant proportion of redo log write operations to
complete to disk before flash. In some cases, the value of
FL_DISK_FIRST
can exceed the value ofFL_FLASH_FIRST
. However, this does not mean that the Exadata Smart Flash Logging feature is ineffective or unnecessary.
Parent topic: Monitoring Exadata Smart Flash Log