You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

188 lines
9.1 KiB

Demonstrations of biotop, the Linux eBPF/bcc version.
Short for block device I/O top, biotop summarizes which processes are
performing disk I/O. It's top for disks. Sample output:
# ./biotop
Tracing... Output every 1 secs. Hit Ctrl-C to end
08:04:11 loadavg: 1.48 0.87 0.45 1/287 14547
PID COMM D MAJ MIN DISK I/O Kbytes AVGms
14501 cksum R 202 1 xvda1 361 28832 3.39
6961 dd R 202 1 xvda1 1628 13024 0.59
13855 dd R 202 1 xvda1 1627 13016 0.59
326 jbd2/xvda1-8 W 202 1 xvda1 3 168 3.00
1880 supervise W 202 1 xvda1 2 8 6.71
1873 supervise W 202 1 xvda1 2 8 2.51
1871 supervise W 202 1 xvda1 2 8 1.57
1876 supervise W 202 1 xvda1 2 8 1.22
1892 supervise W 202 1 xvda1 2 8 0.62
1878 supervise W 202 1 xvda1 2 8 0.78
1886 supervise W 202 1 xvda1 2 8 1.30
1894 supervise W 202 1 xvda1 2 8 3.46
1869 supervise W 202 1 xvda1 2 8 0.73
1888 supervise W 202 1 xvda1 2 8 1.48
By default the screen refreshes every 1 second, and shows the top 20 disk
consumers, sorted on total Kbytes. The first line printed is the header,
which has the time and then the contents of /proc/loadavg.
For the interval summarized by the output above, the "cksum" command performed
361 disk reads to the "xvda1" device, for a total of 28832 Kbytes, with an
average I/O time of 3.39 ms. Two "dd" processes were also reading from the
same disk, which a higher I/O rate and lower latency. While the average I/O
size is not printed, it can be determined by dividing the Kbytes column by
the I/O column.
The columns through to Kbytes show the workload applied. The final column,
AVGms, shows resulting performance. Other bcc tools can be used to get more
details when needed: biolatency and biosnoop.
Many years ago I created the original "iotop", and later regretted not calling
it diskiotop or blockiotop, as "io" alone is ambiguous. This time it is biotop.
The -C option can be used to prevent the screen from clearing (my preference).
Here's using it with a 5 second interval:
# ./biotop -C 5
Tracing... Output every 5 secs. Hit Ctrl-C to end
08:09:44 loadavg: 0.42 0.44 0.39 2/282 22115
PID COMM D MAJ MIN DISK I/O Kbytes AVGms
22069 dd R 202 1 xvda1 5993 47976 0.33
326 jbd2/xvda1-8 W 202 1 xvda1 3 168 2.67
1866 svscan R 202 1 xvda1 33 132 1.24
1880 supervise W 202 1 xvda1 10 40 0.56
1873 supervise W 202 1 xvda1 10 40 0.79
1871 supervise W 202 1 xvda1 10 40 0.78
1876 supervise W 202 1 xvda1 10 40 0.68
1892 supervise W 202 1 xvda1 10 40 0.71
1878 supervise W 202 1 xvda1 10 40 0.65
1886 supervise W 202 1 xvda1 10 40 0.78
1894 supervise W 202 1 xvda1 10 40 0.80
1869 supervise W 202 1 xvda1 10 40 0.91
1888 supervise W 202 1 xvda1 10 40 0.63
22069 bash R 202 1 xvda1 1 16 19.94
9251 kworker/u16:2 W 202 16 xvdb 2 8 0.13
08:09:49 loadavg: 0.47 0.44 0.39 1/282 22231
PID COMM D MAJ MIN DISK I/O Kbytes AVGms
22069 dd R 202 1 xvda1 13450 107600 0.35
22199 cksum R 202 1 xvda1 941 45548 4.63
326 jbd2/xvda1-8 W 202 1 xvda1 3 168 2.93
24467 kworker/0:2 W 202 16 xvdb 1 64 0.28
1880 supervise W 202 1 xvda1 10 40 0.81
1873 supervise W 202 1 xvda1 10 40 0.81
1871 supervise W 202 1 xvda1 10 40 1.03
1876 supervise W 202 1 xvda1 10 40 0.76
1892 supervise W 202 1 xvda1 10 40 0.74
1878 supervise W 202 1 xvda1 10 40 0.94
1886 supervise W 202 1 xvda1 10 40 0.76
1894 supervise W 202 1 xvda1 10 40 0.69
1869 supervise W 202 1 xvda1 10 40 0.72
1888 supervise W 202 1 xvda1 10 40 1.70
22199 bash R 202 1 xvda1 2 20 0.35
482 xfsaild/md0 W 202 16 xvdb 5 13 0.27
482 xfsaild/md0 W 202 32 xvdc 2 8 0.33
31331 pickup R 202 1 xvda1 1 4 0.31
08:09:54 loadavg: 0.51 0.45 0.39 2/282 22346
PID COMM D MAJ MIN DISK I/O Kbytes AVGms
22069 dd R 202 1 xvda1 14689 117512 0.32
326 jbd2/xvda1-8 W 202 1 xvda1 3 168 2.33
1880 supervise W 202 1 xvda1 10 40 0.65
1873 supervise W 202 1 xvda1 10 40 1.08
1871 supervise W 202 1 xvda1 10 40 0.66
1876 supervise W 202 1 xvda1 10 40 0.79
1892 supervise W 202 1 xvda1 10 40 0.67
1878 supervise W 202 1 xvda1 10 40 0.66
1886 supervise W 202 1 xvda1 10 40 1.02
1894 supervise W 202 1 xvda1 10 40 0.88
1869 supervise W 202 1 xvda1 10 40 0.89
1888 supervise W 202 1 xvda1 10 40 1.25
08:09:59 loadavg: 0.55 0.46 0.40 2/282 22461
PID COMM D MAJ MIN DISK I/O Kbytes AVGms
22069 dd R 202 1 xvda1 14442 115536 0.33
326 jbd2/xvda1-8 W 202 1 xvda1 3 168 3.46
1880 supervise W 202 1 xvda1 10 40 0.87
1873 supervise W 202 1 xvda1 10 40 0.87
1871 supervise W 202 1 xvda1 10 40 0.78
1876 supervise W 202 1 xvda1 10 40 0.86
1892 supervise W 202 1 xvda1 10 40 0.89
1878 supervise W 202 1 xvda1 10 40 0.87
1886 supervise W 202 1 xvda1 10 40 0.86
1894 supervise W 202 1 xvda1 10 40 1.06
1869 supervise W 202 1 xvda1 10 40 1.12
1888 supervise W 202 1 xvda1 10 40 0.98
08:10:04 loadavg: 0.59 0.47 0.40 3/282 22576
PID COMM D MAJ MIN DISK I/O Kbytes AVGms
22069 dd R 202 1 xvda1 14179 113432 0.34
326 jbd2/xvda1-8 W 202 1 xvda1 3 168 2.39
1880 supervise W 202 1 xvda1 10 40 0.81
1873 supervise W 202 1 xvda1 10 40 1.02
1871 supervise W 202 1 xvda1 10 40 1.15
1876 supervise W 202 1 xvda1 10 40 1.10
1892 supervise W 202 1 xvda1 10 40 0.77
1878 supervise W 202 1 xvda1 10 40 0.72
1886 supervise W 202 1 xvda1 10 40 0.81
1894 supervise W 202 1 xvda1 10 40 0.86
1869 supervise W 202 1 xvda1 10 40 0.83
1888 supervise W 202 1 xvda1 10 40 0.79
24467 kworker/0:2 R 202 32 xvdc 3 12 0.26
1056 cron R 202 1 xvda1 2 8 0.30
24467 kworker/0:2 R 202 16 xvdb 1 4 0.23
08:10:09 loadavg: 0.54 0.46 0.40 2/281 22668
PID COMM D MAJ MIN DISK I/O Kbytes AVGms
22069 dd R 202 1 xvda1 250 2000 0.34
326 jbd2/xvda1-8 W 202 1 xvda1 3 168 2.40
1880 supervise W 202 1 xvda1 8 32 0.93
1873 supervise W 202 1 xvda1 8 32 0.76
1871 supervise W 202 1 xvda1 8 32 0.60
1876 supervise W 202 1 xvda1 8 32 0.61
1892 supervise W 202 1 xvda1 8 32 0.68
1878 supervise W 202 1 xvda1 8 32 0.90
1886 supervise W 202 1 xvda1 8 32 0.57
1894 supervise W 202 1 xvda1 8 32 0.97
1869 supervise W 202 1 xvda1 8 32 0.69
1888 supervise W 202 1 xvda1 8 32 0.67
This shows another "dd" command reading from xvda1. On this system, various
"supervise" processes do 8 disk writes per second, every second (they are
creating and updating "status" files).
USAGE message:
# ./biotop.py -h
usage: biotop.py [-h] [-C] [-r MAXROWS] [interval] [count]
Block device (disk) I/O by process
positional arguments:
interval output interval, in seconds
count number of outputs
optional arguments:
-h, --help show this help message and exit
-C, --noclear don't clear the screen
-r MAXROWS, --maxrows MAXROWS
maximum rows to print, default 20
examples:
./biotop # block device I/O top, 1 second refresh
./biotop -C # don't clear the screen
./biotop 5 # 5 second summaries
./biotop 5 10 # 5 second summaries, 10 times only