IOSTAT
NAME:Iostat reports CPU statistics and I/O statistics.
Syntax:Iostat [-c] [-d] [-N] [-n] [-h] [-k |-m] [-t] [-V] [-x] [-y] [-z] [-j {ID | LABEL | PATH | UUID | ...} [device [...] | ALL] [device [...] | ALL] [-p [device [,...] | ALL] [interval [count]
Option:-C only displays the CPU usage report-d only displays the device usage report-N displays the registered device mapped name. -N display Network File System (NFS) Report-h makes it easy for NFS reporting users to read (try adding without-h) -j {ID | LABEL | PATH | UUID | ...} [device [...] | ALL] specify the persistent device names type, you can specify all devices in the latter-k to display the write read volume per second in KB-m to display the write read volume per second in MB-p use the-p option to enter the device name, report the partitions of devices and devices. ALL indicates that the system defines ALL devices and partitions. If-j is used before this parameter, the device uses the specified persistent name type (If option-j is defined before this option, devices entered on the command line can be specified with the chosen persistent name type .). -T print the time when each report is generated-V print tool version-y ignore the first report, statistical interval data-z ignore the device that is not active during sampling
Note:Interval is used to specify the interval, and count is used to specify the number of occurrences within the interval. If no count is specified, it is considered as countless times. By default, the first report of iostat is an average value since it was started. Unless-y is specified, the first report is discarded. The next report is the count of the time interval from the previous report to the present.
Report:
3 Reports: CPU, device, NFS
CPU: The CPU report is a total value, and multiple CPUs only return one. % User-level usage % nice cpu usage at nice limited level % system-level usage % iowait CPU idle, while the system requests the device time % steal when the hypervisor provides services for another virtual processor, it unconsciously waits for the ratio of the virtual CPU or CPU % idel CPU idle ratio
Device: Reports are based on physical devices or partitions. If device or partition is not specified, ALL the resulting partitions will be reported. If the keyword ALL is written, ALL defined tps will be reported: Transmission per second, transmission is an IO request to the device. Multiple logical requests can be combined into one transmission. The transmission size is not clear. Blk_read/s (rsec/s ): the block read per second (the block size after 2.4 is the same as that of the fan area). The slice size is 512BBlk_wrtn/s (wsec/s ): the number of blocks written per second (the size of the block after 2.4 is the same as that of the Sector). The slice size is 512BBlk_read: Total number of reads Blk_wrtn: Total number of writes kB_read/s (mB_read/s, rkb/s, RMB/s): number of bytes read per second kB_wrtn/s (mB_wrtn/s, wkb, wmb/s): number of bytes written per second kB_read (mB_read ): total read bytes kB_wrtn (mB_wrtn): Total write bytes rrqm/s: Read queue length per second wrqm/s: Write queue length per second r/s: number of reads requested to the device per second w/s: Number of writes requested to the device per second avgrq-sz: Average number of request sectors avgqu-sz: average length of Request queue await: from the time the request was sent to the processed time, the time contained in the queue Svctm: Man said this value is unreliable and will not explain % util: cpu usage when the request is sent to the device, if the device is saturated, this value is close to 100% NFS: This is not the focus of this time.