blkparse - produce formatted output of event streams of block devices blkparse [ options ]

The blkparse utility will attempt to combine streams of events for various devices on various CPUs, and produce a formatted output of the event information. Specifically, it will take the (machine-readable) output of the blktrace utility and convert it to a nicely formatted and human-readable form. As with blktrace, some details concerning blkparse will help in understanding the command line options presented below.

2 - By default, blkparse expects to run in a post-processing mode; one where the trace events have been saved by a previous run of blktrace, and blkparse is combining event streams and dumping formatted data. blkparse may be run in a live manner concurrently with blktrace by specifying -i - to blkparse, and combining it with the live option for blktrace. An example would be: % blktrace -d /dev/sda -o - | blkparse -i -

2 - You can set how many blkparse batches event reads via the -b option, the default is to handle events in batches of 512.

2 - If you have saved event traces in blktrace with different output names (via the -o option to blktrace), you must specify the same input name via the -i option.

2 - The format of the output data can be controlled via the -f or -F options -- see OUTPUT DESCRIPTION AND FORMATTING for details.

By default, blkparse sends formatted data to standard output. This may be changed via the -o option, or text output can be disabled via the -O option. A merged binary stream can be produced using the -d option.

-b batch

--batch={batch}

-i file

--input=file

-F typ,fmt

--format=typ,fmt

-f fmt

--format-spec=fmt

-M

--no-msgs

-h

--hash-by-name

-o file

--output=file

-O

--no-text-output

-d file

--dump-binary=file

-q

--quiet

-s

--per-program-stats

-t

--track-ios

-w span

--stopwatch=span

-v

--verbose

-V

--version

The following trace actions are recognised:

4 C -- complete A previously issued request has been completed. The output will detail the sector and size of that request, as well as the success or failure of it.

4 D -- issued A request that previously resided on the block layer queue or in the i/o scheduler has been sent to the driver.

4 I -- inserted A request is being sent to the i/o scheduler for addition to the internal queue and later service by the driver. The request is fully formed at this time.

4 Q -- queued This notes intent to queue i/o at the given location. No real requests exists yet.

4 B -- bounced The data pages attached to this bio are not reachable by the hardware and must be bounced to a lower memory location. This causes a big slowdown in i/o performance, since the data must be copied to/from kernel buffers. Usually this can be fixed with using better hardware -- either a better i/o controller, or a platform with an IOMMU.

4 M -- back merge A previously inserted request exists that ends on the boundary of where this i/o begins, so the i/o scheduler can merge them together.

4 F -- front merge Same as the back merge, except this i/o ends where a previously inserted requests starts.

4 M --front or back merge One of the above

4 M -- front or back merge One of the above.

4 G -- get request To send any type of request to a block device, a struct request container must be allocated first.

4 S -- sleep No available request structures were available, so the issuer has to wait for one to be freed.

4 P -- plug When i/o is queued to a previously empty block device queue, Linux will plug the queue in anticipation of future ios being added before this data is needed.

4 U -- unplug Some request data already queued in the device, start sending requests to the driver. This may happen automatically if a timeout period has passed (see next entry) or if a number of requests have been added to the queue.

4 T -- unplug due to timer If nobody requests the i/o that was queued after plugging the queue, Linux will automatically unplug it after a defined period has passed.

4 X -- split On raid or device mapper setups, an incoming i/o may straddle a device or internal zone and needs to be chopped up into smaller pieces for service. This may indicate a performance problem due to a bad setup of that raid/dm device, but may also just be part of normal boundary conditions. dm is notably bad at this and will clone lots of i/o.

4 A -- remap For stacked devices, incoming i/o is remapped to device below it in the i/o stack. The remap action details what exactly is being remapped to what.

The output from blkparse can be tailored for specific use -- in particular, to ease parsing of output, and/or limit output fields to those the user wants to see. The data for fields which can be output include: Action, a (small) string (1 or 2 characters) -- see table below for more details CPU id Command RWBS field, a (small) string (1-3 characters) -- see section below for more details 7-character string containing the major and minor numbers of the event's device (separated by a comma). Error value Minor number of event's device. Major number of event's device. Number of blocks Number of bytes Process ID Display packet data -- series of hexadecimal values Sequence numbers Sector number Time stamp (nanoseconds) Time stamp (seconds) Elapsed value in microseconds (-t command line option) Payload unsigned integer

Note that the user can optionally specify field display width, and optionally a left-aligned specifier. These precede field specifiers, with a '%' character, followed by the optional left-alignment specifier (-) followed by the width (a decimal number) and then the field. Thus, to specify the command in a 12-character field that is left aligned: -f "%-12C"

The following table shows the various actions which may be output: IO was remapped to a different device IO bounced IO completion IO issued to driver IO front merged with request on queue Get request IO inserted onto request queue IO back merged with request on queue Plug request IO handled by request queue code Sleep request Unplug due to timeout Unplug request Split This is a small string containing at least one character ('R' for read, 'W' for write, or 'D' for block discard operation), and optionally either a 'B' (for barrier operations) or 'S' (for synchronous operations). The standard header (or initial fields displayed) include: "%D %2c %8s %5T.%9t %5p %2a %3d" Breaking this down: Displays the event's device major/minor as: %3d,%-3d. CPU ID (2-character field). Sequence number 5-character field for the seconds portion of the time stamp and a 9-character field for the nanoseconds in the time stamp. 5-character field for the process ID. 2-character field for one of the actions. 3-character field for the RWBS data. Seeing this in action: 8,0 3 1 0.000000000 697 G W 223490 + 8 [kjournald] The header is the data in this line up to the 223490 (starting block). The default output for all event types includes this header. C -- complete B -- bounced

D -- issued

I -- inserted

Q -- queued

F -- front merge

G -- get request

M -- back merge

S -- sleep

P -- plug U -- unplug

T -- unplug due to timer

X -- split A -- remap To trace the i/o on the device /dev/hda and parse the output to human readable form, use the following command: % blktrace -d /dev/sda -o - | blkparse -i - (see blktrace (8) for more information). This same behaviour can be achieve with the convenience script btrace. The command % btrace /dev/sda has exactly the same effect as the previous command. See btrace (8) for more information. To trace the i/o on a device and save the output for later processing with blkparse, use blktrace like this: % blktrace /dev/sda /dev/sdb This will trace i/o on the devices /dev/sda and /dev/sdb and save the recorded information in the files sda and sdb in the current directory, for the two different devices, respectively. This trace i