blkparse

Parse and display blktrace output for a device

$ blkparse [sda]

$ blkparse [sda] [sdb]

$ blktrace -d [/dev/sda] -o - | blkparse -i -

$ blkparse [sda] -o [output.txt]

$ blkparse -f "%D %a %S %n\n" [sda]

$ blkparse -q [sda]

blkparse [options] [device...]

blkparse produces formatted output from event streams recorded by blktrace. It combines event streams from multiple CPUs and devices, converting machine-readable trace data into human-readable format.
The tool is essential for analyzing block I/O patterns, debugging storage performance issues, and understanding how the Linux I/O scheduler processes requests.

-i file

Input file (use "-" for stdin)

Output file (default: stdout)

Input directory for trace files

Custom output format string

Quiet mode; only print summary at end

Show per-program statistics

Track individual I/O completions

Time span to process (in seconds)

More verbose output

Q - Queued: Intent to queue I/O noted
I - Inserted: Request sent to I/O scheduler
D - Issued: Request sent to driver
C - Complete: Request completed
M - Back merge: Request merged with existing
F - Front merge: Request merged at front
P - Plug: Queue plugged
U - Unplug: Queue unplugged

Requires trace data from blktrace to function. Trace files can become very large for busy systems. Running blktrace requires root privileges. Output can be overwhelming; use format options to limit fields.

blkparse was written by Jens Axboe, Alan D. Brunelle, and Nathan Scott as part of the blktrace toolkit. It was developed to provide detailed I/O tracing capabilities for Linux storage subsystem analysis and debugging.

blktrace(8), btrace(8), btt(1), iostat(1)