fluent-plugin-perf-tools 0.1.0

Sign up to get free protection for your applications and to get access to all the features.
Files changed (98) hide show
  1. checksums.yaml +7 -0
  2. data/.gitignore +15 -0
  3. data/.rubocop.yml +26 -0
  4. data/.ruby-version +1 -0
  5. data/CHANGELOG.md +5 -0
  6. data/CODE_OF_CONDUCT.md +84 -0
  7. data/Gemfile +5 -0
  8. data/LICENSE.txt +21 -0
  9. data/README.md +43 -0
  10. data/Rakefile +17 -0
  11. data/bin/console +15 -0
  12. data/bin/setup +8 -0
  13. data/fluent-plugin-perf-tools.gemspec +48 -0
  14. data/lib/fluent/plugin/in_perf_tools.rb +42 -0
  15. data/lib/fluent/plugin/perf_tools/cachestat.rb +65 -0
  16. data/lib/fluent/plugin/perf_tools/command.rb +30 -0
  17. data/lib/fluent/plugin/perf_tools/version.rb +9 -0
  18. data/lib/fluent/plugin/perf_tools.rb +11 -0
  19. data/perf-tools/LICENSE +339 -0
  20. data/perf-tools/README.md +205 -0
  21. data/perf-tools/bin/bitesize +1 -0
  22. data/perf-tools/bin/cachestat +1 -0
  23. data/perf-tools/bin/execsnoop +1 -0
  24. data/perf-tools/bin/funccount +1 -0
  25. data/perf-tools/bin/funcgraph +1 -0
  26. data/perf-tools/bin/funcslower +1 -0
  27. data/perf-tools/bin/functrace +1 -0
  28. data/perf-tools/bin/iolatency +1 -0
  29. data/perf-tools/bin/iosnoop +1 -0
  30. data/perf-tools/bin/killsnoop +1 -0
  31. data/perf-tools/bin/kprobe +1 -0
  32. data/perf-tools/bin/opensnoop +1 -0
  33. data/perf-tools/bin/perf-stat-hist +1 -0
  34. data/perf-tools/bin/reset-ftrace +1 -0
  35. data/perf-tools/bin/syscount +1 -0
  36. data/perf-tools/bin/tcpretrans +1 -0
  37. data/perf-tools/bin/tpoint +1 -0
  38. data/perf-tools/bin/uprobe +1 -0
  39. data/perf-tools/deprecated/README.md +1 -0
  40. data/perf-tools/deprecated/execsnoop-proc +150 -0
  41. data/perf-tools/deprecated/execsnoop-proc.8 +80 -0
  42. data/perf-tools/deprecated/execsnoop-proc_example.txt +46 -0
  43. data/perf-tools/disk/bitesize +175 -0
  44. data/perf-tools/examples/bitesize_example.txt +63 -0
  45. data/perf-tools/examples/cachestat_example.txt +58 -0
  46. data/perf-tools/examples/execsnoop_example.txt +153 -0
  47. data/perf-tools/examples/funccount_example.txt +126 -0
  48. data/perf-tools/examples/funcgraph_example.txt +2178 -0
  49. data/perf-tools/examples/funcslower_example.txt +110 -0
  50. data/perf-tools/examples/functrace_example.txt +341 -0
  51. data/perf-tools/examples/iolatency_example.txt +350 -0
  52. data/perf-tools/examples/iosnoop_example.txt +302 -0
  53. data/perf-tools/examples/killsnoop_example.txt +62 -0
  54. data/perf-tools/examples/kprobe_example.txt +379 -0
  55. data/perf-tools/examples/opensnoop_example.txt +47 -0
  56. data/perf-tools/examples/perf-stat-hist_example.txt +149 -0
  57. data/perf-tools/examples/reset-ftrace_example.txt +88 -0
  58. data/perf-tools/examples/syscount_example.txt +297 -0
  59. data/perf-tools/examples/tcpretrans_example.txt +93 -0
  60. data/perf-tools/examples/tpoint_example.txt +210 -0
  61. data/perf-tools/examples/uprobe_example.txt +321 -0
  62. data/perf-tools/execsnoop +292 -0
  63. data/perf-tools/fs/cachestat +167 -0
  64. data/perf-tools/images/perf-tools_2016.png +0 -0
  65. data/perf-tools/iolatency +296 -0
  66. data/perf-tools/iosnoop +296 -0
  67. data/perf-tools/kernel/funccount +146 -0
  68. data/perf-tools/kernel/funcgraph +259 -0
  69. data/perf-tools/kernel/funcslower +248 -0
  70. data/perf-tools/kernel/functrace +192 -0
  71. data/perf-tools/kernel/kprobe +270 -0
  72. data/perf-tools/killsnoop +263 -0
  73. data/perf-tools/man/man8/bitesize.8 +70 -0
  74. data/perf-tools/man/man8/cachestat.8 +111 -0
  75. data/perf-tools/man/man8/execsnoop.8 +104 -0
  76. data/perf-tools/man/man8/funccount.8 +76 -0
  77. data/perf-tools/man/man8/funcgraph.8 +166 -0
  78. data/perf-tools/man/man8/funcslower.8 +129 -0
  79. data/perf-tools/man/man8/functrace.8 +123 -0
  80. data/perf-tools/man/man8/iolatency.8 +116 -0
  81. data/perf-tools/man/man8/iosnoop.8 +169 -0
  82. data/perf-tools/man/man8/killsnoop.8 +100 -0
  83. data/perf-tools/man/man8/kprobe.8 +162 -0
  84. data/perf-tools/man/man8/opensnoop.8 +113 -0
  85. data/perf-tools/man/man8/perf-stat-hist.8 +111 -0
  86. data/perf-tools/man/man8/reset-ftrace.8 +49 -0
  87. data/perf-tools/man/man8/syscount.8 +96 -0
  88. data/perf-tools/man/man8/tcpretrans.8 +93 -0
  89. data/perf-tools/man/man8/tpoint.8 +140 -0
  90. data/perf-tools/man/man8/uprobe.8 +168 -0
  91. data/perf-tools/misc/perf-stat-hist +223 -0
  92. data/perf-tools/net/tcpretrans +311 -0
  93. data/perf-tools/opensnoop +280 -0
  94. data/perf-tools/syscount +192 -0
  95. data/perf-tools/system/tpoint +232 -0
  96. data/perf-tools/tools/reset-ftrace +123 -0
  97. data/perf-tools/user/uprobe +390 -0
  98. metadata +349 -0
@@ -0,0 +1,150 @@
1
+ #!/usr/bin/perl
2
+ #
3
+ # execsnoop - trace process exec() with arguments. /proc version.
4
+ # Written using Linux ftrace.
5
+ #
6
+ # This shows the execution of new processes, especially short-lived ones that
7
+ # can be missed by sampling tools such as top(1).
8
+ #
9
+ # USAGE: ./execsnoop [-h] [-n name]
10
+ #
11
+ # REQUIREMENTS: FTRACE CONFIG, sched:sched_process_exec tracepoint (you may
12
+ # already have these on recent kernels), and Perl.
13
+ #
14
+ # This traces exec() from the fork()->exec() sequence, which means it won't
15
+ # catch new processes that only fork(), and, it will catch processes that
16
+ # re-exec. This instruments sched:sched_process_exec without buffering, and then
17
+ # in user-space (this program) reads PPID and process arguments asynchronously
18
+ # from /proc.
19
+ #
20
+ # If the process traced is very short-lived, this program may miss reading
21
+ # arguments and PPID details. In that case, "<?>" and "?" will be printed
22
+ # respectively. This program is best-effort, and should be improved in the
23
+ # future when other kernel capabilities are made available. If you need a
24
+ # more reliable tool now, then consider other tracing alternatives (eg,
25
+ # SystemTap). This tool is really a proof of concept to see what ftrace can
26
+ # currently do.
27
+ #
28
+ # From perf-tools: https://github.com/brendangregg/perf-tools
29
+ #
30
+ # See the execsnoop(8) man page (in perf-tools) for more info.
31
+ #
32
+ # COPYRIGHT: Copyright (c) 2014 Brendan Gregg.
33
+ #
34
+ # This program is free software; you can redistribute it and/or
35
+ # modify it under the terms of the GNU General Public License
36
+ # as published by the Free Software Foundation; either version 2
37
+ # of the License, or (at your option) any later version.
38
+ #
39
+ # This program is distributed in the hope that it will be useful,
40
+ # but WITHOUT ANY WARRANTY; without even the implied warranty of
41
+ # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
42
+ # GNU General Public License for more details.
43
+ #
44
+ # You should have received a copy of the GNU General Public License
45
+ # along with this program; if not, write to the Free Software Foundation,
46
+ # Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
47
+ #
48
+ # (http://www.gnu.org/copyleft/gpl.html)
49
+ #
50
+ # 07-Jul-2014 Brendan Gregg Created this.
51
+
52
+ use strict;
53
+ use warnings;
54
+ use POSIX qw(strftime);
55
+ use Getopt::Long;
56
+ my $tracing = "/sys/kernel/debug/tracing";
57
+ my $flock = "/var/tmp/.ftrace-lock";
58
+ my $tpdir = "sched/sched_process_exec";
59
+ my $tptext = $tpdir; $tptext =~ s/\//:/;
60
+ local $SIG{INT} = \&cleanup;
61
+ local $SIG{QUIT} = \&cleanup;
62
+ local $SIG{TERM} = \&cleanup;
63
+ local $SIG{PIPE} = \&cleanup;
64
+ local $SIG{HUP} = \&cleanup;
65
+ $| = 1;
66
+
67
+ ### options
68
+ my ($name, $help);
69
+ GetOptions("name=s" => \$name,
70
+ "help" => \$help)
71
+ or usage();
72
+ usage() if $help;
73
+ sub usage {
74
+ print STDERR "USAGE: execsnoop [-h] [-n name]\n";
75
+ print STDERR " eg,\n";
76
+ print STDERR " execsnoop -n ls # show \"ls\" cmds only.\n";
77
+ exit;
78
+ }
79
+
80
+ sub ldie {
81
+ unlink $flock;
82
+ die @_;
83
+ }
84
+
85
+ sub writeto {
86
+ my ($string, $file) = @_;
87
+ open FILE, ">$file" or return 0;
88
+ print FILE $string or return 0;
89
+ close FILE or return 0;
90
+ }
91
+
92
+ ### check permissions
93
+ chdir "$tracing" or ldie "ERROR: accessing tracing. Root? Kernel has FTRACE?" .
94
+ "\ndebugfs mounted? (mount -t debugfs debugfs /sys/kernel/debug)";
95
+
96
+ ### ftrace lock
97
+ if (-e $flock) {
98
+ open FLOCK, $flock; my $fpid = <FLOCK>; chomp $fpid; close FLOCK;
99
+ die "ERROR: ftrace may be in use by PID $fpid ($flock)";
100
+ }
101
+ writeto "$$", $flock or die "ERROR: unable to write $flock.";
102
+
103
+ ### setup and begin tracing
104
+ writeto "nop", "current_tracer" or ldie "ERROR: disabling current_tracer.";
105
+ writeto "1", "events/$tpdir/enable" or ldie "ERROR: enabling tracepoint " .
106
+ "\"$tptext\" (tracepoint missing in this kernel version?)";
107
+ open TPIPE, "trace_pipe" or warn "ERROR: opening trace_pipe.";
108
+ printf "%-8s %6s %6s %s\n", "TIME", "PID", "PPID", "ARGS";
109
+
110
+ while (<TPIPE>) {
111
+ my ($taskpid, $rest) = split;
112
+ my ($task, $pid) = $taskpid =~ /(.*)-(\d+)/;
113
+
114
+ next if (defined $name and $name ne $task);
115
+
116
+ my $args = "$task <?>";
117
+ if (open CMDLINE, "/proc/$pid/cmdline") {
118
+ my $arglist = <CMDLINE>;
119
+ if (defined $arglist) {
120
+ $arglist =~ s/\000/ /g;
121
+ $args = $arglist;
122
+ }
123
+ close CMDLINE;
124
+ }
125
+
126
+ my $ppid = "?";
127
+ if (open STAT, "/proc/$pid/stat") {
128
+ my $fields = <STAT>;
129
+ if (defined $fields) {
130
+ $ppid = (split ' ', $fields)[3];
131
+ }
132
+ close STAT;
133
+ }
134
+
135
+ my $now = strftime "%H:%M:%S", localtime;
136
+ printf "%-8s %6s %6s %s\n", $now, $pid, $ppid, $args;
137
+ }
138
+
139
+ ### end tracing
140
+ cleanup();
141
+
142
+ sub cleanup {
143
+ print "\nEnding tracing...\n";
144
+ close TPIPE;
145
+ writeto "0", "events/$tpdir/enable" or
146
+ ldie "ERROR: disabling \"$tptext\"";
147
+ writeto "", "trace";
148
+ unlink $flock;
149
+ exit;
150
+ }
@@ -0,0 +1,80 @@
1
+ .TH execsnoop\-proc 8 "2014-07-07" "USER COMMANDS"
2
+ .SH NAME
3
+ execsnoop\-proc \- trace process exec() with arguments. Uses Linux ftrace. /proc version.
4
+ .SH SYNOPSIS
5
+ .B execsnoop\-proc
6
+ [\-h] [\-n name]
7
+ .SH DESCRIPTION
8
+ execsnoop\-proc traces process execution, showing PID, PPID, and argument details
9
+ if possible.
10
+
11
+ This traces exec() from the fork()->exec() sequence, which means it won't
12
+ catch new processes that only fork(), and, it will catch processes that
13
+ re-exec. This instruments sched:sched_process_exec without buffering, and then
14
+ in user-space (this program) reads PPID and process arguments asynchronously
15
+ from /proc.
16
+
17
+ If the process traced is very short-lived, this program may miss reading
18
+ arguments and PPID details. In that case, "<?>" and "?" will be printed
19
+ respectively.
20
+
21
+ This program is best-effort (a hack), and should be improved in the future when
22
+ other kernel capabilities are made available. It may be useful in the meantime.
23
+ If you need a more reliable tool now, consider other tracing alternates (eg,
24
+ SystemTap). This tool is really a proof of concept to see what ftrace can
25
+ currently do.
26
+
27
+ See execsnoop(8) for another version that reads arguments from registers
28
+ instead of /proc.
29
+
30
+ Since this uses ftrace, only the root user can use this tool.
31
+ .SH REQUIREMENTS
32
+ FTRACE CONFIG and the sched:sched_process_exec tracepoint, which you may already
33
+ have enabled and available on recent kernels, and Perl.
34
+ .SH OPTIONS
35
+ \-n name
36
+ Only show processes that match this name. This is filtered in user space.
37
+ .TP
38
+ \-h
39
+ Print usage message.
40
+ .SH EXAMPLES
41
+ .TP
42
+ Trace all new processes and arguments (if possible):
43
+ .B execsnoop\-proc
44
+ .TP
45
+ Trace all new processes with process name "sed":
46
+ .B execsnoop\-proc -n sed
47
+ .SH FIELDS
48
+ .TP
49
+ TIME
50
+ Time of process exec(): HH:MM:SS.
51
+ .TP
52
+ PID
53
+ Process ID.
54
+ .TP
55
+ PPID
56
+ Parent process ID, if this was able to be read (may be missed for short-lived
57
+ processes). If it is unable to be read, "?" is printed.
58
+ .TP
59
+ ARGS
60
+ Command line arguments, if these were able to be read in time (may be missed
61
+ for short-lived processes). If they are unable to be read, "<?>" is printed.
62
+ .SH OVERHEAD
63
+ This reads and processes exec() events in user space as they occur. Since the
64
+ rate of exec() is expected to be low (< 500/s), the overhead is expected to
65
+ be small or negligible.
66
+ .SH SOURCE
67
+ This is from the perf-tools collection.
68
+ .IP
69
+ https://github.com/brendangregg/perf-tools
70
+ .PP
71
+ Also look under the examples directory for a text file containing example
72
+ usage, output, and commentary for this tool.
73
+ .SH OS
74
+ Linux
75
+ .SH STABILITY
76
+ Unstable - in development.
77
+ .SH AUTHOR
78
+ Brendan Gregg
79
+ .SH SEE ALSO
80
+ execsnoop(8), top(1)
@@ -0,0 +1,46 @@
1
+ Demonstrations of execsnoop-proc, the Linux ftrace version.
2
+
3
+ Here's execsnoop showing what's really executed by "man ls":
4
+
5
+ # ./execsnoop
6
+ TIME PID PPID ARGS
7
+ 17:52:37 22406 25781 man ls
8
+ 17:52:37 22413 22406 preconv -e UTF-8
9
+ 17:52:37 22416 22406 pager -s
10
+ 17:52:37 22415 22406 /bin/sh /usr/bin/nroff -mandoc -rLL=162n -rLT=162n -Tutf8
11
+ 17:52:37 22414 22406 tbl
12
+ 17:52:37 22419 22418 locale charmap
13
+ 17:52:37 22420 22415 groff -mtty-char -Tutf8 -mandoc -rLL=162n -rLT=162n
14
+ 17:52:37 22421 22420 troff -mtty-char -mandoc -rLL=162n -rLT=162n -Tutf8
15
+ 17:52:37 22422 22420 grotty
16
+
17
+
18
+ These are short-lived processes, where the argument and PPID details are often
19
+ missed by execsnoop:
20
+
21
+ # ./execsnoop
22
+ TIME PID PPID ARGS
23
+ 18:00:33 26750 1961 multilog <?>
24
+ 18:00:33 26749 1972 multilog <?>
25
+ 18:00:33 26749 1972 multilog <?>
26
+ 18:00:33 26751 ? mkdir <?>
27
+ 18:00:33 26749 1972 multilog <?>
28
+ 18:00:33 26752 ? chown <?>
29
+ 18:00:33 26750 1961 multilog <?>
30
+ 18:00:33 26750 1961 multilog <?>
31
+ 18:00:34 26753 1961 multilog <?>
32
+ 18:00:34 26754 1972 multilog <?>
33
+ [...]
34
+
35
+ This will be fixed in a later version, but likely requires some kernel or
36
+ tracer changes first (fetching cmdline as the probe fires).
37
+
38
+
39
+ The previous examples were on Linux 3.14 and 3.16 kernels. Here's a 3.2 system
40
+ I'm running:
41
+
42
+ # ./execsnoop
43
+ ERROR: enabling tracepoint "sched:sched_process_exec" (tracepoint missing in this kernel version?) at ./execsnoop line 78.
44
+
45
+ This kernel version is missing the sched_process_exec probe, which is pretty
46
+ annoying.
@@ -0,0 +1,175 @@
1
+ #!/bin/bash
2
+ #
3
+ # bitesize - show disk I/O size as a histogram.
4
+ # Written using Linux perf_events (aka "perf").
5
+ #
6
+ # This can be used to characterize the distribution of block device I/O
7
+ # sizes. To study I/O in more detail, see iosnoop(8).
8
+ #
9
+ # USAGE: bitesize [-h] [-b buckets] [seconds]
10
+ # eg,
11
+ # ./bitesize 10
12
+ #
13
+ # Run "bitesize -h" for full usage.
14
+ #
15
+ # REQUIREMENTS: perf_events and block:block_rq_issue tracepoint, which you may
16
+ # already have on recent kernels.
17
+ #
18
+ # This uses multiple counting tracepoints with different filters, one for each
19
+ # histogram bucket. While this is summarized in-kernel, the use of multiple
20
+ # tracepoints does add addiitonal overhead, which is more evident if you add
21
+ # more buckets. In the future this functionality will be available in an
22
+ # efficient way in the kernel, and this tool can be rewritten.
23
+ #
24
+ # From perf-tools: https://github.com/brendangregg/perf-tools
25
+ #
26
+ # COPYRIGHT: Copyright (c) 2014 Brendan Gregg.
27
+ #
28
+ # This program is free software; you can redistribute it and/or
29
+ # modify it under the terms of the GNU General Public License
30
+ # as published by the Free Software Foundation; either version 2
31
+ # of the License, or (at your option) any later version.
32
+ #
33
+ # This program is distributed in the hope that it will be useful,
34
+ # but WITHOUT ANY WARRANTY; without even the implied warranty of
35
+ # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
36
+ # GNU General Public License for more details.
37
+ #
38
+ # You should have received a copy of the GNU General Public License
39
+ # along with this program; if not, write to the Free Software Foundation,
40
+ # Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
41
+ #
42
+ # (http://www.gnu.org/copyleft/gpl.html)
43
+ #
44
+ # 22-Jul-2014 Brendan Gregg Created this.
45
+
46
+ duration=0
47
+ buckets=(1 8 64 128)
48
+ secsz=512
49
+ trap ':' INT QUIT TERM PIPE HUP
50
+
51
+ function usage {
52
+ cat <<-END >&2
53
+ USAGE: bitesize [-h] [-b buckets] [seconds]
54
+ -b buckets # specify histogram buckets (Kbytes)
55
+ -h # this usage message
56
+ eg,
57
+ bitesize # trace I/O size until Ctrl-C
58
+ bitesize 10 # trace I/O size for 10 seconds
59
+ bitesize -b "8 16 32" # specify custom bucket points
60
+ END
61
+ exit
62
+ }
63
+
64
+ function die {
65
+ echo >&2 "$@"
66
+ exit 1
67
+ }
68
+
69
+ ### process options
70
+ while getopts b:h opt
71
+ do
72
+ case $opt in
73
+ b) buckets=($OPTARG) ;;
74
+ h|?) usage ;;
75
+ esac
76
+ done
77
+ shift $(( $OPTIND - 1 ))
78
+ tpoint=block:block_rq_issue
79
+ var=nr_sector
80
+ duration=$1
81
+
82
+ ### convert buckets (Kbytes) to disk sectors
83
+ i=0
84
+ sectors=(${buckets[*]})
85
+ ((max_i = ${#buckets[*]} - 1))
86
+ while (( i <= max_i )); do
87
+ (( sectors[$i] = ${sectors[$i]} * 1024 / $secsz ))
88
+ # avoid negative array index errors for old version bash
89
+ if (( i > 0 ));then
90
+ if (( ${sectors[$i]} <= ${sectors[$i - 1]} )); then
91
+ die "ERROR: bucket list must increase in size."
92
+ fi
93
+ fi
94
+ (( i++ ))
95
+ done
96
+
97
+ ### build list of tracepoints and filters for each histogram bucket
98
+ max_b=${buckets[$max_i]}
99
+ max_s=${sectors[$max_i]}
100
+ tpoints="-e $tpoint --filter \"$var < ${sectors[0]}\""
101
+ awkarray=
102
+ i=0
103
+ while (( i < max_i )); do
104
+ tpoints="$tpoints -e $tpoint --filter \"$var >= ${sectors[$i]} && "
105
+ tpoints="$tpoints $var < ${sectors[$i + 1]}\""
106
+ awkarray="$awkarray buckets[$i]=${buckets[$i]};"
107
+ (( i++ ))
108
+ done
109
+ awkarray="$awkarray buckets[$max_i]=${buckets[$max_i]};"
110
+ tpoints="$tpoints -e $tpoint --filter \"$var >= ${sectors[$max_i]}\""
111
+
112
+ ### prepare to run
113
+ if (( duration )); then
114
+ etext="for $duration seconds"
115
+ cmd="sleep $duration"
116
+ else
117
+ etext="until Ctrl-C"
118
+ cmd="sleep 999999"
119
+ fi
120
+ echo "Tracing block I/O size (bytes), $etext..."
121
+
122
+ ### run perf
123
+ out="-o /dev/stdout" # a workaround needed in linux 3.2; not by 3.4.15
124
+ stat=$(eval perf stat $tpoints -a $out $cmd 2>&1)
125
+ if (( $? != 0 )); then
126
+ echo >&2 "ERROR running perf:"
127
+ echo >&2 "$stat"
128
+ exit
129
+ fi
130
+
131
+ ### find max value for ASCII histogram
132
+ most=$(echo "$stat" | awk -v tpoint=$tpoint '
133
+ $2 == tpoint { gsub(/,/, ""); if ($1 > m) { m = $1 } }
134
+ END { print m }'
135
+ )
136
+
137
+ ### process output
138
+ echo
139
+ echo "$stat" | awk -v tpoint=$tpoint -v max_i=$max_i -v most=$most '
140
+ function star(sval, smax, swidth) {
141
+ stars = ""
142
+ # using int could avoid error on gawk
143
+ if (int(smax) == 0) return ""
144
+ for (si = 0; si < (swidth * sval / smax); si++) {
145
+ stars = stars "#"
146
+ }
147
+ return stars
148
+ }
149
+ BEGIN {
150
+ '"$awkarray"'
151
+ printf(" %-15s: %-8s %s\n", "Kbytes", "I/O",
152
+ "Distribution")
153
+ }
154
+ /Performance counter stats/ { i = -1 }
155
+ # reverse order of rule set is important
156
+ { ok = 0 }
157
+ $2 == tpoint { num = $1; gsub(/,/, "", num); ok = 1 }
158
+ ok && i >= max_i {
159
+ printf(" %10.1f -> %-10s: %-8s |%-38s|\n",
160
+ buckets[i], "", num, star(num, most, 38))
161
+ next
162
+ }
163
+ ok && i >= 0 && i < max_i {
164
+ printf(" %10.1f -> %-10.1f: %-8s |%-38s|\n",
165
+ buckets[i], buckets[i+1] - 0.1, num,
166
+ star(num, most, 38))
167
+ i++
168
+ next
169
+ }
170
+ ok && i == -1 {
171
+ printf(" %10s -> %-10.1f: %-8s |%-38s|\n", "",
172
+ buckets[0] - 0.1, num, star(num, most, 38))
173
+ i++
174
+ }
175
+ '
@@ -0,0 +1,63 @@
1
+ Demonstrations of bitesize, the Linux perf_events version.
2
+
3
+
4
+ bitesize traces block I/O issued, and reports a histogram of I/O size. By
5
+ default five buckets are used to gather statistics on common I/O sizes:
6
+
7
+ # ./bitesize
8
+ Tracing block I/O size (bytes), until Ctrl-C...
9
+ ^C
10
+ Kbytes : I/O Distribution
11
+ -> 0.9 : 0 | |
12
+ 1.0 -> 7.9 : 38 |# |
13
+ 8.0 -> 63.9 : 10108 |######################################|
14
+ 64.0 -> 127.9 : 13 |# |
15
+ 128.0 -> : 1 |# |
16
+
17
+ In this case, most of the I/O was between 8 and 63.9 Kbytes. The "63.9"
18
+ really means "less than 64".
19
+
20
+
21
+ Specifying custom buckets to examine the I/O size in more detail:
22
+
23
+ # ./bitesize -b "8 16 24 32"
24
+ Tracing block I/O size (bytes), until Ctrl-C...
25
+ ^C
26
+ Kbytes : I/O Distribution
27
+ -> 7.9 : 89 |# |
28
+ 8.0 -> 15.9 : 14665 |######################################|
29
+ 16.0 -> 23.9 : 657 |## |
30
+ 24.0 -> 31.9 : 661 |## |
31
+ 32.0 -> : 376 |# |
32
+
33
+ The I/O is mostly between 8 and 15.9 Kbytes
34
+
35
+ It's probably 8 Kbytes. Checking:
36
+
37
+ # ./bitesize -b "8 9"
38
+ Tracing block I/O size (bytes), until Ctrl-C...
39
+ ^C
40
+ Kbytes : I/O Distribution
41
+ -> 7.9 : 62 |# |
42
+ 8.0 -> 8.9 : 11719 |######################################|
43
+ 9.0 -> : 1358 |##### |
44
+
45
+ It is.
46
+
47
+ The overhead of this tool is relative to the number of buckets used, hence only
48
+ using what is necessary.
49
+
50
+ To study this I/O in more detail, I can use iosnoop(8) and capture it to a file
51
+ for post-processing.
52
+
53
+
54
+ Use -h to print the USAGE message:
55
+
56
+ # ./bitesize -h
57
+ USAGE: bitesize [-h] [-b buckets] [seconds]
58
+ -b buckets # specify histogram buckets (Kbytes)
59
+ -h # this usage message
60
+ eg,
61
+ bitesize # trace I/O size until Ctrl-C
62
+ bitesize 10 # trace I/O size for 10 seconds
63
+ bitesize -b "8 16 32" # specify custom bucket points
@@ -0,0 +1,58 @@
1
+ Demonstrations of cachestat, the Linux ftrace version.
2
+
3
+
4
+ Here is some sample output showing file system cache statistics, followed by
5
+ the workload that caused it:
6
+
7
+ # ./cachestat -t
8
+ Counting cache functions... Output every 1 seconds.
9
+ TIME HITS MISSES DIRTIES RATIO BUFFERS_MB CACHE_MB
10
+ 08:28:57 415 0 0 100.0% 1 191
11
+ 08:28:58 411 0 0 100.0% 1 191
12
+ 08:28:59 362 97 0 78.9% 0 8
13
+ 08:29:00 411 0 0 100.0% 0 9
14
+ 08:29:01 775 20489 0 3.6% 0 89
15
+ 08:29:02 411 0 0 100.0% 0 89
16
+ 08:29:03 6069 0 0 100.0% 0 89
17
+ 08:29:04 15249 0 0 100.0% 0 89
18
+ 08:29:05 411 0 0 100.0% 0 89
19
+ 08:29:06 411 0 0 100.0% 0 89
20
+ 08:29:07 411 0 3 100.0% 0 89
21
+ [...]
22
+
23
+ I used the -t option to include the TIME column, to make describing the output
24
+ easier.
25
+
26
+ The workload was:
27
+
28
+ # echo 1 > /proc/sys/vm/drop_caches; sleep 2; cksum 80m; sleep 2; cksum 80m
29
+
30
+ At 8:28:58, the page cache was dropped by the first command, which can be seen
31
+ by the drop in size for "CACHE_MB" (page cache size) from 191 Mbytes to 8.
32
+ After a 2 second sleep, a cksum command was issued at 8:29:01, for an 80 Mbyte
33
+ file (called "80m"), which caused a total of ~20,400 misses ("MISSES" column),
34
+ and the page cache size to grow by 80 Mbytes. The hit ratio during this dropped
35
+ to 3.6%. Finally, after another 2 second sleep, at 8:29:03 the cksum command
36
+ was run a second time, this time hitting entirely from cache.
37
+
38
+ Instrumenting all file system cache accesses does cost some overhead, and this
39
+ tool might slow your target system by 2% or so. Test before use if this is a
40
+ concern.
41
+
42
+ This tool also uses dynamic tracing, and is tied to Linux kernel implementation
43
+ details. If it doesn't work for you, it probably needs fixing.
44
+
45
+
46
+ Use -h to print the USAGE message:
47
+
48
+ # ./cachestat -h
49
+ USAGE: cachestat [-Dht] [interval]
50
+ -D # print debug counters
51
+ -h # this usage message
52
+ -t # include timestamp
53
+ interval # output interval in secs (default 1)
54
+ eg,
55
+ cachestat # show stats every second
56
+ cachestat 5 # show stats every 5 seconds
57
+
58
+ See the man page and example file for more info.