1 .TH "LTTNG" "1" "February 9, 2012" "" ""
4 lttng \(em LTTng 2.0 tracer control command line tool
10 lttng [OPTIONS] <COMMAND>
15 The LTTng project aims at providing highly efficient tracing tools for Linux.
16 It's tracers help tracking down performance issues and debugging problems
17 involving multiple concurrent processes and threads. Tracing across multiple
18 systems is also possible.
20 The \fBlttng\fP command line tool from lttng-tools package is used to control
21 both kernel and user-space tracing. Every interactions with the tracer should
22 be done by this tool or by the liblttng-ctl provided with the lttng-tools
25 LTTng uses a session daemon (lttng-sessiond(8)), acting as a tracing registry,
26 which permits you to interact with multiple tracers (kernel and user-space)
27 inside the same container, a tracing session. Traces can be gathered from the
28 kernel and/or instrumented applications (lttng-ust(3)). Aggregating and reading
29 those traces is done using the babeltrace(1) text viewer.
31 In order to trace the kernel, the session daemon needs to be running as root.
32 LTTng provides the use of a \fBtracing group\fP (default: tracing). Whomever is
33 in that group can interact with the root session daemon and thus trace the
34 kernel. Session daemons can co-exist meaning that you can have a session daemon
35 running as Alice that can be use to trace her applications along side with a
36 root daemon or even a Bob daemon. We highly recommand to start the session
37 daemon at boot time for stable and long term tracing.
39 Every user-space applications instrumented with lttng-ust(3), will
40 automatically register to the session daemon. This feature gives you the
41 ability to list available traceable applications and tracepoints on a per user
42 basis. (See \fBlist\fP command).
46 This program follow the usual GNU command line syntax with long options starting with
47 two dashes. Below is a summary of the available options.
52 Show summary of possible options and commands.
54 .BR "\-v, \-\-verbose"
56 FIXME : details (\-v : sessiond verbose, \-vv : consumerd verbose, etc) ?
59 Suppress all messages (even errors).
61 .BR "\-g, \-\-group NAME"
62 Set unix tracing group name. (default: tracing)
64 .BR "\-n, \-\-no-sessiond"
65 Don't automatically spawn a session daemon.
67 .BR "\-\-sessiond\-path"
68 Set session daemon full binary path.
70 .BR "\-\-list\-options"
71 Simple listing of lttng options.
73 .BR "\-\-list\-commands"
74 Simple listing of lttng commands.
80 Add context to event(s) and/or channel(s).
82 A context is basically extra information appended to a channel or event. For
83 instance, you could ask the tracer to add the PID information within the
84 "sched_switch" kernel event. You can also add performance monitoring unit
85 counters (perf PMU) using the perf kernel API).
87 For example, this command will add the context information 'prio' and two perf
88 counters (hardware branch misses and cache misses), to all events in the trace
91 # lttng add-context \-k \-t prio \-t perf:branch-misses \-t perf:cache-misses
93 Please take a look at the help (\-h/\-\-help) for a detailed list of available
96 If no channel and no event is given (\-c/\-e), the context is added to all
97 channels (which applies automatically to all events in that channel). Otherwise
98 the context will be added only to the channel (\-c) and/or event (\-e) indicated.
100 If \fB\-s, \-\-session\fP is omitted, the session name is taken from the .lttngrc
108 Show summary of possible options and commands.
109 \-s, \-\-session NAME
110 Apply on session name.
111 \-c, \-\-channel NAME
112 Apply on channel name.
116 Apply for the kernel tracer
118 Apply for the user-space tracer
120 Context type. You can repeat this option on the command line. Please
121 use "lttng add-context \-h" to list all available types.
126 .IP "\fBcalibrate\fP"
128 Quantify LTTng overhead
130 The LTTng calibrate command can be used to find out the combined average
131 overhead of the LTTng tracer and the instrumentation mechanisms used. This
132 overhead can be calibrated in terms of time or using any of the PMU performance
133 counter available on the system.
135 For now, the only calibration implemented is that of the kernel function
136 instrumentation (kretprobes).
138 * Calibrate kernel function instrumentation
140 Let's use an example to show this calibration. We use an i7 processor with 4
141 general-purpose PMU registers. This information is available by issuing dmesg,
142 looking for "generic registers".
144 This sequence of commands will gather a trace executing a kretprobe hooked on
145 an empty function, gathering PMU counters LLC (Last Level Cache) misses
146 information (see lttng add-context \-\-help to see the list of available PMU
149 # lttng create calibrate-function
150 # lttng enable-event calibrate \-\-kernel \-\-function lttng_calibrate_kretprobe
151 # lttng add-context \-\-kernel \-t perf:LLC-load-misses \-t perf:LLC-store-misses \\
152 \-t perf:LLC-prefetch-misses
154 # for a in $(seq 1 10); do \\
155 lttng calibrate \-\-kernel \-\-function;
158 # babeltrace $(ls \-1drt ~/lttng-traces/calibrate-function-* | tail \-n 1)
160 The output from babeltrace can be saved to a text file and opened in a
161 spreadsheet (e.g. oocalc) to focus on the per-PMU counter delta between
162 consecutive "calibrate_entry" and "calibrate_return" events. Note that these
163 counters are per-CPU, so scheduling events would need to be present to account
164 for migration between CPU. Therefore, for calibration purposes, only events
165 staying on the same CPU must be considered.
167 The average result, for the i7, on 10 samples:
170 perf_LLC_load_misses: 5.0 0.577
171 perf_LLC_store_misses: 1.6 0.516
172 perf_LLC_prefetch_misses: 9.0 14.742
174 As we can notice, the load and store misses are relatively stable across runs
175 (their standard deviation is relatively low) compared to the prefetch misses.
176 We can conclude from this information that LLC load and store misses can be
177 accounted for quite precisely, but prefetches within a function seems to behave
178 too erratically (not much causality link between the code executed and the CPU
179 prefetch activity) to be accounted for.
186 Show summary of possible options and commands.
188 Apply for the kernel tracer
190 Apply for the user-space tracer
192 Dynamic function entry/return probe (default)
197 .IP "\fBcreate\fP [OPTIONS] [NAME]
199 Create tracing session.
201 A tracing session contains channel(s) which contains event(s). It is domain
202 agnostic meaning that you can enable channels and events for either the
203 user-space tracer and/or the kernel tracer. It acts like a container
204 aggregating multiple tracing sources.
206 On creation, a \fB.lttngrc\fP file is created in your $HOME directory
207 containing the current session name. If NAME is omitted, a session name is
208 automatically created having this form: 'auto-yyyymmdd-hhmms'.
210 If no \fB\-o, \-\-output\fP is specified, the traces will be written in
218 Show summary of possible options and commands.
220 Simple listing of options
222 Specify output path for traces
227 .IP "\fBdestroy\fP [OPTIONS] [NAME]"
229 Teardown tracing session
231 Free memory on the session daemon and tracer side. It's gone!
233 If NAME is omitted, the session name is taken from the .lttngrc file.
240 Show summary of possible options and commands.
242 Simple listing of options
247 .IP "\fBenable-channel\fP NAME[,NAME2,...] [-k|-u] [OPTIONS]"
249 Enable tracing channel
251 If \fB\-s, \-\-session\fP is omitted, the session name is taken from the .lttngrc
261 Simple listing of options
263 Apply on session name
265 Apply to the kernel tracer
267 Apply to the user-space tracer
270 Discard event when subbuffers are full (default)
272 Flight recorder mode : overwrites events when subbuffers are full
274 Subbuffer size in bytes (default: 4096, kernel default: 262144)
276 Number of subbufers (default: 4)
277 Needs to be a power of 2 for kernel and ust tracers
279 Switch subbuffer timer interval in usec (default: 0)
280 Needs to be a power of 2 for kernel and ust tracers
282 Read timer interval in usec (default: 200)
287 .IP "\fBenable-event\fP NAME[,NAME2,...] [-k|-u] [OPTIONS]"
291 A tracing event is always assigned to a channel. If \fB\-c, \-\-channel\fP is
292 omitted, a default channel named '\fBchannel0\fP' is created and the event is
293 added to it. For the user-space tracer, using \fB\-a, \-\-all\fP is the same as
294 using the wildcard "*".
296 If \fB\-s, \-\-session\fP is omitted, the session name is taken from the .lttngrc
304 Show summary of possible options and commands.
306 Simple listing of options
308 Apply on session name
310 Apply on channel name
312 Enable all tracepoints
314 Apply for the kernel tracer
316 Apply for the user-space tracer
319 Tracepoint event (default)
320 - userspace tracer supports wildcards at end of string. Don't forget to
321 quote to deal with bash expansion.
327 \-\-probe [addr | symbol | symbol+offset]
328 Dynamic probe. Addr and offset can be octal (0NNN...), decimal (NNN...)
329 or hexadecimal (0xNNN...)
330 \-\-function [addr | symbol | symbol+offset]
331 Dynamic function entry/return probe. Addr and offset can be octal
332 (0NNN...), decimal (NNN...) or hexadecimal (0xNNN...)
335 Enabling syscalls tracing (kernel tracer), you will not be able to disable them
336 with disable-event. This is a known limitation. You can disable the entire
337 channel to do the trick.
340 .IP "\fBdisable-channel\fP NAME[,NAME2,...] [\-k|\-u] [OPTIONS]"
342 Disable tracing channel
344 Disabling a channel makes all event(s) in that channel to stop tracing. You can
345 enable it back by calling \fBlttng enable-channel NAME\fP again.
347 If \fB\-s, \-\-session\fP is omitted, the session name is taken from the .lttngrc
355 Show summary of possible options and commands.
357 Simple listing of options
359 Apply on session name
361 Apply for the kernel tracer
363 Apply for the user-space tracer
366 .IP "\fBdisable-event\fP NAME[,NAME2,...] [\-k|\-u] [OPTIONS]"
368 Disable tracing event
370 The event, once disabled, can be re-enabled by calling \fBlttng enable-event
373 If \fB\-s, \-\-session\fP is omitted, the session name is taken from the .lttngrc
381 Show summary of possible options and commands.
383 Simple listing of options
385 Apply on session name
387 Apply for the kernel tracer
389 Apply for the user-space tracer
392 .IP "\fBlist\fP [\-k|\-u] [SESSION [SESSION_OPTIONS]]"
394 List tracing session information.
396 With no arguments, it will list available tracing session(s).
398 With \-k alone, it will list all available kernel events (except the system
400 With \-u alone, it will list all available user-space events from registered
401 applications. Here is an example of 'lttng list \-u':
403 PID: 7448 - Name: /tmp/lttng-ust/tests/hello/.libs/lt-hello
404 ust_tests_hello:tptest_sighandler (type: tracepoint)
405 ust_tests_hello:tptest (type: tracepoint)
407 You can now enable any event listed by using the name :
408 \fBust_tests_hello:tptest\fP.
415 Show summary of possible options and commands.
417 Simple listing of options
419 Select kernel domain (FIXME : apparition de la notion de "domain" ici)
421 Select user-space domain.
424 \-c, \-\-channel NAME
425 List details of a channel
427 List available domain(s)
430 .IP "\fBset-session\fP NAME"
432 Set current session name
434 Will change the session name in the .lttngrc file.
441 Show summary of possible options and commands.
443 Simple listing of options
448 .IP "\fBstart\fP [OPTIONS] [NAME]"
452 It will start tracing for all tracers for a specific tracing session.
454 If NAME is omitted, the session name is taken from the .lttngrc file.
461 Show summary of possible options and commands.
463 Simple listing of options
468 .IP "\fBstop\fP [OPTIONS] [NAME]"
472 It will stop tracing for all tracers for a specific tracing session.
474 If NAME is omitted, the session name is taken from the .lttngrc file.
481 Show summary of possible options and commands.
483 Simple listing of options
490 Show version information
497 Show summary of possible options and commands.
499 Simple listing of options
504 .IP "\fBview\fP [SESSION_NAME] [OPTIONS]"
506 View traces of a tracing session
508 By default, the babeltrace viewer will be used for text viewing.
510 The SESSION_NAME is an optional session name. If not specified, lttng will get
511 it from the configuration file (.lttngrc).
520 Simple listing of options
521 \-t, \-\-trace-path PATH
522 Trace directory path for the viewer
524 Specify viewer and/or options to use
525 This will completely override the default viewers so
526 please make sure to specify the full command. The trace
527 directory path of the session will be appended at the end
531 .SH "ENVIRONMENT VARIABLES"
534 Note that all command line options override environment variables.
538 .IP "LTTNG_SESSIOND_PATH_ENV"
539 Allows one to specify the full session daemon binary path to lttng command line
540 tool. You can also use \-\-sessiond-path option having the same effect.
544 babeltrace(1), lttng-ust(3), lttng-sessiond(8)
549 No show stopper bugs known yet at this stable version.
551 If you encounter any issues or usability problem, please report it on our
552 mailing list <lttng-dev@lists.lttng.org> to help improve this project.
556 lttng is distributed under the GNU General Public License version 2. See the file
559 A Web site is available at http://lttng.org for more information on the LTTng
562 You can also find our git tree at http://git.lttng.org.
564 Mailing lists for support and development: <lttng-dev@lists.lttng.org>.
566 You can find us on IRC server irc.oftc.net (OFTC) in #lttng.
571 Thanks to Yannick Brosseau without whom this project would never have been so
572 lean and mean! Also thanks to the Ericsson teams working on tracing which
573 helped us greatly with detailled bug reports and unusual test cases.
575 Thanks to our beloved packager Alexandre Montplaisir-Goncalves (Ubuntu and PPA
576 maintainer) and Jon Bernard for our Debian packages.
578 Special thanks to Michel Dagenais and the DORSAL laboratory at Polytechnique de
579 Montreal for the LTTng journey.
584 lttng-tools was originally written by Mathieu Desnoyers, Julien Desfossez and
585 David Goulet. More people have since contributed to it. It is currently
586 maintained by David Goulet <dgoulet@efficios.com>.