1runqlen(8) System Manager's Manual runqlen(8)
2
3
4
6 runqlen - Scheduler run queue length as a histogram.
7
9 runqlen [-h] [-T] [-O] [-C] [interval] [count]
10
12 This program summarizes scheduler queue length as a histogram, and can
13 also show run queue occupancy. It works by sampling the run queue
14 length on all CPUs at 99 Hertz.
15
16 This tool can be used to identify imbalances, eg, when processes are
17 bound to CPUs causing queueing, or interrupt mappings causing the same.
18
19 Since this uses BPF, only the root user can use this tool.
20
22 CONFIG_BPF and bcc.
23
25 -h Print usage message.
26
27 -T Include timestamps on output.
28
29 -O Report run queue occupancy.
30
31 -C Report for each CPU.
32
33 interval
34 Output interval, in seconds.
35
36 count Number of outputs.
37
39 Summarize run queue length as a histogram:
40 # runqlen
41
42 Print 1 second summaries, 10 times:
43 # runqlen 1 10
44
45 Print output every second, with timestamps, and show each CPU sepa‐
46 rately:
47 # runqlen -CT 1
48
49 Print run queue occupancy every second:
50 # runqlen -O 1
51
52 Print run queue occupancy, with timetamps, for each CPU:
53 # runqlen -COT 1
54
56 runqlen
57 Scheduler run queue length: the number of threads (tasks) wait‐
58 ing to run, (excluding including the currently running task).
59
60 count Number of samples at this queue length.
61
62 distribution
63 An ASCII bar chart to visualize the distribution (count column)
64
66 This uses sampling at 99 Hertz (on all CPUs), and in-kernel summaries,
67 which should make overhead negligible. This does not trace scheduler
68 events, like runqlen does, which comes at a much higher overhead cost.
69
71 This is from bcc.
72
73 https://github.com/iovisor/bcc
74
75 Also look in the bcc distribution for a companion _examples.txt file
76 containing example usage, output, and commentary for this tool.
77
79 Linux
80
82 Unstable - in development.
83
85 Brendan Gregg
86
88 runqlat(8), runqslower(8), pidstat(1)
89
90
91
92USER COMMANDS 2016-12-12 runqlen(8)