1 The LTTng Documentation
2 =======================
3 Philippe Proulx <pproulx@efficios.com>
7 include::../common/copyright.txt[]
10 include::../common/welcome.txt[]
13 include::../common/audience.txt[]
17 === What's in this documentation?
19 The LTTng Documentation is divided into the following sections:
21 * **<<nuts-and-bolts,Nuts and bolts>>** explains the
22 rudiments of software tracing and the rationale behind the
25 You can skip this section if you’re familiar with software tracing and
26 with the LTTng project.
28 * **<<installing-lttng,Installation>>** describes the steps to
29 install the LTTng packages on common Linux distributions and from
32 You can skip this section if you already properly installed LTTng on
35 * **<<getting-started,Quick start>>** is a concise guide to
36 getting started quickly with LTTng kernel and user space tracing.
38 We recommend this section if you're new to LTTng or to software tracing
41 You can skip this section if you're not new to LTTng.
43 * **<<core-concepts,Core concepts>>** explains the concepts at
46 It's a good idea to become familiar with the core concepts
47 before attempting to use the toolkit.
49 * **<<plumbing,Components of LTTng>>** describes the various components
50 of the LTTng machinery, like the daemons, the libraries, and the
51 command-line interface.
52 * **<<instrumenting,Instrumentation>>** shows different ways to
53 instrument user applications and the Linux kernel.
55 Instrumenting source code is essential to provide a meaningful
58 You can skip this section if you do not have a programming background.
60 * **<<controlling-tracing,Tracing control>>** is divided into topics
61 which demonstrate how to use the vast array of features that
62 LTTng{nbsp}{revision} offers.
63 * **<<reference,Reference>>** contains reference tables.
64 * **<<glossary,Glossary>>** is a specialized dictionary of terms related
65 to LTTng or to the field of software tracing.
68 include::../common/convention.txt[]
71 include::../common/acknowledgements.txt[]
75 == What's new in LTTng {revision}?
77 LTTng{nbsp}{revision} bears the name _Isseki Nicho_. The result of a
78 collaboration between http://www.dieuduciel.com/[Dieu du Ciel!] and
79 Nagano-based Shiga Kogen,
80 https://www.beeradvocate.com/beer/profile/1141/53111/[_**Isseki
81 Nicho**_] is a strong Imperial Dark Saison offering a rich roasted malt
82 flavor combined with a complex fruity finish typical of Saison yeasts.
84 New features and changes in LTTng{nbsp}{revision}:
86 * **Tracing control**:
87 ** You can attach <<java-application-context,Java application-specific
88 context fields>> to a <<channel,channel>> with the
89 man:lttng-add-context(1) command:
94 $ lttng add-context --jul --type='$app.retriever:cur_msg_id'
98 Here, `$app` is the prefix of all application-specific context fields,
99 `retriever` names a _context information retriever_ defined at the
100 application level, and `cur_msg_id` names a context field read from this
103 Both the `java.util.logging` and Apache log4j <<domain,tracing domains>>
106 ** You can use Java application-specific <<adding-context,context>>
107 fields in the <<enabling-disabling-events,filter expression>> of an
108 <<event,event rule>>:
113 $ lttng enable-event --log4j my_logger \
114 --filter='$app.retriever:cur_msg_id == 23'
118 ** New `lttng status` command which is the equivalent of +lttng list
119 __CUR__+, where +__CUR__+ is the name of the current
120 <<tracing-session,tracing session>>.
122 See man:lttng-status(1).
124 ** New `lttng metadata regenerate` command to
125 <<metadata-regenerate,regenerate the metadata file of an LTTng
126 trace>> at any moment. This command is meant to be used to resample
127 the wall time following a major
128 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
129 so that a system which boots with an incorrect wall time can be
130 traced before its wall time is NTP-corrected.
132 See man:lttng-metadata(1).
134 ** New command-line interface warnings when <<event,event records>> or
135 whole sub-buffers are
136 <<channel-overwrite-mode-vs-discard-mode,lost>>. The warning messages
137 are printed when a <<tracing-session,tracing session>> is
138 <<basic-tracing-session-control,stopped>> (man:lttng-stop(1)
141 * **User space tracing**:
142 ** Shared object base address dump in order to map <<event,event
143 records>> to original source location (file and line number).
145 If you attach the `ip` and `vpid` <<adding-context,context fields>> to a
146 user space <<channel,channel>> and if you use the
147 <<liblttng-ust-dl,path:{liblttng-ust-dl.so} helper>>, you can retrieve
148 the source location where a given event record was generated.
150 The http://diamon.org/babeltrace/[Babeltrace] trace viewer supports this
151 state dump and those context fields since version 1.4 to print the
152 source location of a given event record. http://tracecompass.org/[Trace
153 Compass] also supports this since version 2.0.
155 ** A <<java-application,Java application>> which uses
156 `java.util.logging` now adds an LTTng-UST log handler to the desired
159 The previous workflow was to initialize the LTTng-UST Java agent
160 by calling `LTTngAgent.getLTTngAgent()`. This had the effect of adding
161 an LTTng-UST log handler to the root loggers.
163 ** A <<java-application,Java application>> which uses Apache log4j now
164 adds an LTTng-UST log appender to the desired log4j loggers.
166 The previous workflow was to initialize the LTTng-UST Java agent
167 by calling `LTTngAgent.getLTTngAgent()`. This had the effect of adding
168 an LTTng-UST appender to the root loggers.
170 ** Any <<java-application,Java application>> can provide
171 <<java-application-context,dynamic context fields>> while running
172 thanks to a new API provided by the <<lttng-ust-agents,LTTng-UST Java
173 agent>>. You can require LTTng to record specific context fields in
174 event records, and you can use them in the filter expression of
175 <<event,event rules>>.
177 * **Linux kernel tracing**:
178 ** The LTTng kernel modules can now be built into a Linux kernel image,
179 that is, not as loadable modules.
182 https://github.com/lttng/lttng-modules/blob/stable-{revision}/README.md#kernel-built-in-support[`README.md`]
185 ** New instrumentation:
186 *** ARM64 architecture support.
188 *** x86 `irq_vectors`.
189 ** New <<adding-context,context fields>>:
192 *** `need_reschedule`
193 *** `migratable` (specific to RT-Preempt)
194 ** Clock source plugin support for advanced cases where a custom source
195 of time is needed to timestamp LTTng event records.
197 See https://github.com/lttng/lttng-modules/blob/stable-{revision}/lttng-clock.h[`lttng-clock.h`]
198 for an overview of the small API.
201 ** The link:/man[man pages] of the man:lttng(1) command-line tool are
202 split into one man page per command (à la Git), for example:
207 $ man lttng-enable-event
211 You can also use the `--help` option of any man:lttng(1) command to
214 The content and formatting of all the LTTng man pages has improved
221 What is LTTng? As its name suggests, the _Linux Trace Toolkit: next
222 generation_ is a modern toolkit for tracing Linux systems and
223 applications. So your first question might be:
230 As the history of software engineering progressed and led to what
231 we now take for granted--complex, numerous and
232 interdependent software applications running in parallel on
233 sophisticated operating systems like Linux--the authors of such
234 components, software developers, began feeling a natural
235 urge to have tools that would ensure the robustness and good performance
236 of their masterpieces.
238 One major achievement in this field is, inarguably, the
239 https://www.gnu.org/software/gdb/[GNU debugger (GDB)],
240 an essential tool for developers to find and fix bugs. But even the best
241 debugger won't help make your software run faster, and nowadays, faster
242 software means either more work done by the same hardware, or cheaper
243 hardware for the same work.
245 A _profiler_ is often the tool of choice to identify performance
246 bottlenecks. Profiling is suitable to identify _where_ performance is
247 lost in a given software. The profiler outputs a profile, a statistical
248 summary of observed events, which you may use to discover which
249 functions took the most time to execute. However, a profiler won't
250 report _why_ some identified functions are the bottleneck. Bottlenecks
251 might only occur when specific conditions are met, conditions that are
252 sometimes impossible to capture by a statistical profiler, or impossible
253 to reproduce with an application altered by the overhead of an
254 event-based profiler. For a thorough investigation of software
255 performance issues, a history of execution is essential, with the
256 recorded values of variables and context fields you choose, and
257 with as little influence as possible on the instrumented software. This
258 is where tracing comes in handy.
260 _Tracing_ is a technique used to understand what goes on in a running
261 software system. The software used for tracing is called a _tracer_,
262 which is conceptually similar to a tape recorder. When recording,
263 specific instrumentation points placed in the software source code
264 generate events that are saved on a giant tape: a _trace_ file. You
265 can trace user applications and the operating system at the same time,
266 opening the possibility of resolving a wide range of problems that would
267 otherwise be extremely challenging.
269 Tracing is often compared to _logging_. However, tracers and loggers are
270 two different tools, serving two different purposes. Tracers are
271 designed to record much lower-level events that occur much more
272 frequently than log messages, often in the range of thousands per
273 second, with very little execution overhead. Logging is more appropriate
274 for a very high-level analysis of less frequent events: user accesses,
275 exceptional conditions (errors and warnings, for example), database
276 transactions, instant messaging communications, and such. Simply put,
277 logging is one of the many use cases that can be satisfied with tracing.
279 The list of recorded events inside a trace file can be read manually
280 like a log file for the maximum level of detail, but it is generally
281 much more interesting to perform application-specific analyses to
282 produce reduced statistics and graphs that are useful to resolve a
283 given problem. Trace viewers and analyzers are specialized tools
286 In the end, this is what LTTng is: a powerful, open source set of
287 tools to trace the Linux kernel and user applications at the same time.
288 LTTng is composed of several components actively maintained and
289 developed by its link:/community/#where[community].
292 [[lttng-alternatives]]
293 === Alternatives to noch:{LTTng}
295 Excluding proprietary solutions, a few competing software tracers
298 * https://github.com/dtrace4linux/linux[dtrace4linux] is a port of
299 Sun Microsystems's DTrace to Linux. The cmd:dtrace tool interprets
300 user scripts and is responsible for loading code into the
301 Linux kernel for further execution and collecting the outputted data.
302 * https://en.wikipedia.org/wiki/Berkeley_Packet_Filter[eBPF] is a
303 subsystem in the Linux kernel in which a virtual machine can execute
304 programs passed from the user space to the kernel. You can attach
305 such programs to tracepoints and KProbes thanks to a system call, and
306 they can output data to the user space when executed thanks to
307 different mechanisms (pipe, VM register values, and eBPF maps, to name
309 * https://www.kernel.org/doc/Documentation/trace/ftrace.txt[ftrace]
310 is the de facto function tracer of the Linux kernel. Its user
311 interface is a set of special files in sysfs.
312 * https://perf.wiki.kernel.org/[perf] is
313 a performance analyzing tool for Linux which supports hardware
314 performance counters, tracepoints, as well as other counters and
315 types of probes. perf's controlling utility is the cmd:perf command
317 * http://linux.die.net/man/1/strace[strace]
318 is a command-line utility which records system calls made by a
319 user process, as well as signal deliveries and changes of process
320 state. strace makes use of https://en.wikipedia.org/wiki/Ptrace[ptrace]
321 to fulfill its function.
322 * http://www.sysdig.org/[sysdig], like SystemTap, uses scripts to
323 analyze Linux kernel events. You write scripts, or _chisels_ in
324 sysdig's jargon, in Lua and sysdig executes them while the system is
325 being traced or afterwards. sysdig's interface is the cmd:sysdig
326 command-line tool as well as the curses-based cmd:csysdig tool.
327 * https://sourceware.org/systemtap/[SystemTap] is a Linux kernel and
328 user space tracer which uses custom user scripts to produce plain text
329 traces. SystemTap converts the scripts to the C language, and then
330 compiles them as Linux kernel modules which are loaded to produce
331 trace data. SystemTap's primary user interface is the cmd:stap
334 The main distinctive features of LTTng is that it produces correlated
335 kernel and user space traces, as well as doing so with the lowest
336 overhead amongst other solutions. It produces trace files in the
337 http://diamon.org/ctf[CTF] format, a file format optimized
338 for the production and analyses of multi-gigabyte data.
340 LTTng is the result of more than 10 years of active open source
341 development by a community of passionate developers.
342 LTTng{nbsp}{revision} is currently available on major desktop and server
345 The main interface for tracing control is a single command-line tool
346 named cmd:lttng. The latter can create several tracing sessions, enable
347 and disable events on the fly, filter events efficiently with custom
348 user expressions, start and stop tracing, and much more. LTTng can
349 record the traces on the file system or send them over the network, and
350 keep them totally or partially. You can view the traces once tracing
351 becomes inactive or in real-time.
353 <<installing-lttng,Install LTTng now>> and
354 <<getting-started,start tracing>>!
360 **LTTng** is a set of software <<plumbing,components>> which interact to
361 <<instrumenting,instrument>> the Linux kernel and user applications, and
362 to <<controlling-tracing,control tracing>> (start and stop
363 tracing, enable and disable event rules, and the rest). Those
364 components are bundled into the following packages:
366 * **LTTng-tools**: Libraries and command-line interface to
368 * **LTTng-modules**: Linux kernel modules to instrument and
370 * **LTTng-UST**: Libraries and Java/Python packages to instrument and
371 trace user applications.
373 Most distributions mark the LTTng-modules and LTTng-UST packages as
374 optional when installing LTTng-tools (which is always required). In the
375 following sections, we always provide the steps to install all three,
378 * You only need to install LTTng-modules if you intend to trace the
380 * You only need to install LTTng-UST if you intend to trace user
384 .Availability of LTTng{nbsp}{revision} for major Linux distributions as of 14 March 2017.
386 |Distribution |Available in releases |Alternatives
388 |https://www.ubuntu.com/[Ubuntu]
389 |<<ubuntu,Ubuntu{nbsp}16.10 _Yakkety Yak_>>.
390 |LTTng{nbsp}{revision} for Ubuntu{nbsp}14.04 _Trusty Tahr_
391 and Ubuntu{nbsp}16.04 _Xenial Xerus_:
392 <<ubuntu-ppa,use the LTTng Stable{nbsp}{revision} PPA>>.
394 LTTng{nbsp}2.9 for Ubuntu{nbsp}14.04 _Trusty Tahr_
395 and Ubuntu{nbsp}16.04 _Xenial Xerus_:
396 link:/docs/v2.9#doc-ubuntu-ppa[use the LTTng Stable{nbsp}2.9 PPA].
398 LTTng{nbsp}2.9 for Ubuntu{nbsp}17.04 _Zesty Zapus_ (not released
401 <<building-from-source,Build LTTng{nbsp}{revision} from source>> for
402 other Ubuntu releases.
404 |https://getfedora.org/[Fedora]
405 |<<fedora,Fedora{nbsp}25>>.
406 |<<building-from-source,Build LTTng{nbsp}{revision} from source>> for
407 other Fedora releases.
409 |https://www.debian.org/[Debian]
411 |link:/docs/v2.9#doc-debian[LTTng{nbsp}2.9 for Debian "stretch" (testing)
412 and Debian "sid" (unstable)].
414 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
416 |https://www.opensuse.org/[openSUSE]
418 |<<building-from-source,Build LTTng{nbsp}{revision} from source>>.
420 |https://www.archlinux.org/[Arch Linux]
422 |link:/docs/v2.9#doc-arch-linux[LTTng{nbsp}2.9 from the AUR].
424 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
426 |https://alpinelinux.org/[Alpine Linux]
428 |link:/docs/v2.9#doc-alpine-linux[LTTng{nbsp}2.9 for Alpine Linux "edge"].
430 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
432 |https://www.redhat.com/[RHEL] and https://www.suse.com/[SLES]
433 |See http://packages.efficios.com/[EfficiOS Enterprise Packages].
436 |https://buildroot.org/[Buildroot]
437 |<<buildroot,Buildroot 2016.11>>.
438 |link:/docs/v2.9#doc-buildroot[LTTng{nbsp}2.9 for Buildroot 2017.02].
440 <<building-from-source,Build LTTng{nbsp}{revision} from source>> for
441 other Buildroot releases.
443 |http://www.openembedded.org/wiki/Main_Page[OpenEmbedded] and
444 https://www.yoctoproject.org/[Yocto]
445 |<<oe-yocto,Yocto Project{nbsp}2.2 _Morty_>> (`openembedded-core` layer).
446 |LTTng{nbsp}2.9 for Yocto Project{nbsp}2.3 _Pyro_
447 (`openembedded-core` layer) (not released yet).
449 <<building-from-source,Build LTTng{nbsp}{revision} from source>> for
450 other OpenEmbedded releases.
455 === [[ubuntu-official-repositories]]Ubuntu
457 LTTng{nbsp}{revision} is available on Ubuntu{nbsp}16.10 _Yakkety Yak_.
458 For previous releases of Ubuntu, <<ubuntu-ppa,use the LTTng
459 Stable{nbsp}{revision} PPA>>.
461 To install LTTng{nbsp}{revision} on Ubuntu{nbsp}16.10 _Yakkety Yak_:
463 . Install the main LTTng{nbsp}{revision} packages:
468 # apt-get install lttng-tools
469 # apt-get install lttng-modules-dkms
470 # apt-get install liblttng-ust-dev
474 . **If you need to instrument and trace
475 <<java-application,Java applications>>**, install the LTTng-UST
481 # apt-get install liblttng-ust-agent-java
485 . **If you need to instrument and trace
486 <<python-application,Python{nbsp}3 applications>>**, install the
487 LTTng-UST Python agent:
492 # apt-get install python3-lttngust
498 ==== noch:{LTTng} Stable {revision} PPA
500 The https://launchpad.net/~lttng/+archive/ubuntu/stable-{revision}[LTTng
501 Stable{nbsp}{revision} PPA] offers the latest stable
502 LTTng{nbsp}{revision} packages for:
504 * Ubuntu{nbsp}14.04 _Trusty Tahr_
505 * Ubuntu{nbsp}16.04 _Xenial Xerus_
507 To install LTTng{nbsp}{revision} from the LTTng Stable{nbsp}{revision} PPA:
509 . Add the LTTng Stable{nbsp}{revision} PPA repository and update the
515 # apt-add-repository ppa:lttng/stable-2.8
520 . Install the main LTTng{nbsp}{revision} packages:
525 # apt-get install lttng-tools
526 # apt-get install lttng-modules-dkms
527 # apt-get install liblttng-ust-dev
531 . **If you need to instrument and trace
532 <<java-application,Java applications>>**, install the LTTng-UST
538 # apt-get install liblttng-ust-agent-java
542 . **If you need to instrument and trace
543 <<python-application,Python{nbsp}3 applications>>**, install the
544 LTTng-UST Python agent:
549 # apt-get install python3-lttngust
557 To install LTTng{nbsp}{revision} on Fedora{nbsp}25:
559 . Install the LTTng-tools{nbsp}{revision} and LTTng-UST{nbsp}{revision}
565 # yum install lttng-tools
566 # yum install lttng-ust
570 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
576 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.8.tar.bz2 &&
577 tar -xf lttng-modules-latest-2.8.tar.bz2 &&
578 cd lttng-modules-2.8.* &&
580 sudo make modules_install &&
586 .Java and Python application instrumentation and tracing
588 If you need to instrument and trace <<java-application,Java
589 applications>> on openSUSE, you need to build and install
590 LTTng-UST{nbsp}{revision} <<building-from-source,from source>> and pass
591 the `--enable-java-agent-jul`, `--enable-java-agent-log4j`, or
592 `--enable-java-agent-all` options to the `configure` script, depending
593 on which Java logging framework you use.
595 If you need to instrument and trace <<python-application,Python
596 applications>> on openSUSE, you need to build and install
597 LTTng-UST{nbsp}{revision} from source and pass the
598 `--enable-python-agent` option to the `configure` script.
602 [[enterprise-distributions]]
603 === RHEL, SUSE, and other enterprise distributions
605 To install LTTng on enterprise Linux distributions, such as Red Hat
606 Enterprise Linux (RHEL) and SUSE Linux Enterprise Server (SUSE), please
607 see http://packages.efficios.com/[EfficiOS Enterprise Packages].
613 To install LTTng{nbsp}{revision} on Buildroot{nbsp}2016.11:
615 . Launch the Buildroot configuration tool:
624 . In **Kernel**, check **Linux kernel**.
625 . In **Toolchain**, check **Enable WCHAR support**.
626 . In **Target packages**{nbsp}→ **Debugging, profiling and benchmark**,
627 check **lttng-modules** and **lttng-tools**.
628 . In **Target packages**{nbsp}→ **Libraries**{nbsp}→
629 **Other**, check **lttng-libust**.
633 === OpenEmbedded and Yocto
635 LTTng{nbsp}{revision} recipes are available in the
636 http://layers.openembedded.org/layerindex/branch/master/layer/openembedded-core/[`openembedded-core`]
637 layer for Yocto Project{nbsp}2.2 _Morty_ under the following names:
643 With BitBake, the simplest way to include LTTng recipes in your target
644 image is to add them to `IMAGE_INSTALL_append` in path:{conf/local.conf}:
647 IMAGE_INSTALL_append = " lttng-tools lttng-modules lttng-ust"
652 . Select a machine and an image recipe.
653 . Click **Edit image recipe**.
654 . Under the **All recipes** tab, search for **lttng**.
655 . Check the desired LTTng recipes.
658 .Java and Python application instrumentation and tracing
660 If you need to instrument and trace <<java-application,Java
661 applications>> on openSUSE, you need to build and install
662 LTTng-UST{nbsp}{revision} <<building-from-source,from source>> and pass
663 the `--enable-java-agent-jul`, `--enable-java-agent-log4j`, or
664 `--enable-java-agent-all` options to the `configure` script, depending
665 on which Java logging framework you use.
667 If you need to instrument and trace <<python-application,Python
668 applications>> on openSUSE, you need to build and install
669 LTTng-UST{nbsp}{revision} from source and pass the
670 `--enable-python-agent` option to the `configure` script.
674 [[building-from-source]]
675 === Build from source
677 To build and install LTTng{nbsp}{revision} from source:
679 . Using your distribution's package manager, or from source, install
680 the following dependencies of LTTng-tools and LTTng-UST:
683 * https://sourceforge.net/projects/libuuid/[libuuid]
684 * http://directory.fsf.org/wiki/Popt[popt]
685 * http://liburcu.org/[Userspace RCU]
686 * http://www.xmlsoft.org/[libxml2]
689 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
695 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.8.tar.bz2 &&
696 tar -xf lttng-modules-latest-2.8.tar.bz2 &&
697 cd lttng-modules-2.8.* &&
699 sudo make modules_install &&
704 . Download, build, and install the latest LTTng-UST{nbsp}{revision}:
710 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.8.tar.bz2 &&
711 tar -xf lttng-ust-latest-2.8.tar.bz2 &&
712 cd lttng-ust-2.8.* &&
722 .Java and Python application tracing
724 If you need to instrument and trace <<java-application,Java
725 applications>>, pass the `--enable-java-agent-jul`,
726 `--enable-java-agent-log4j`, or `--enable-java-agent-all` options to the
727 `configure` script, depending on which Java logging framework you use.
729 If you need to instrument and trace <<python-application,Python
730 applications>>, pass the `--enable-python-agent` option to the
731 `configure` script. You can set the `PYTHON` environment variable to the
732 path to the Python interpreter for which to install the LTTng-UST Python
740 By default, LTTng-UST libraries are installed to
741 dir:{/usr/local/lib}, which is the de facto directory in which to
742 keep self-compiled and third-party libraries.
744 When <<building-tracepoint-providers-and-user-application,linking an
745 instrumented user application with `liblttng-ust`>>:
747 * Append `/usr/local/lib` to the env:LD_LIBRARY_PATH environment
749 * Pass the `-L/usr/local/lib` and `-Wl,-rpath,/usr/local/lib` options to
750 man:gcc(1), man:g++(1), or man:clang(1).
754 . Download, build, and install the latest LTTng-tools{nbsp}{revision}:
760 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.8.tar.bz2 &&
761 tar -xf lttng-tools-latest-2.8.tar.bz2 &&
762 cd lttng-tools-2.8.* &&
770 TIP: The https://github.com/eepp/vlttng[vlttng tool] can do all the
771 previous steps automatically for a given version of LTTng and confine
772 the installed files in a specific directory. This can be useful to test
773 LTTng without installing it on your system.
779 This is a short guide to get started quickly with LTTng kernel and user
782 Before you follow this guide, make sure to <<installing-lttng,install>>
785 This tutorial walks you through the steps to:
787 . <<tracing-the-linux-kernel,Trace the Linux kernel>>.
788 . <<tracing-your-own-user-application,Trace a user application>> written
790 . <<viewing-and-analyzing-your-traces,View and analyze the
794 [[tracing-the-linux-kernel]]
795 === Trace the Linux kernel
797 The following command lines start with the `#` prompt because you need
798 root privileges to trace the Linux kernel. You can also trace the kernel
799 as a regular user if your Unix user is a member of the
800 <<tracing-group,tracing group>>.
802 . Create a <<tracing-session,tracing session>> which writes its traces
803 to dir:{/tmp/my-kernel-trace}:
808 # lttng create my-kernel-session --output=/tmp/my-kernel-trace
812 . List the available kernel tracepoints and system calls:
817 # lttng list --kernel
818 # lttng list --kernel --syscall
822 . Create <<event,event rules>> which match the desired instrumentation
823 point names, for example the `sched_switch` and `sched_process_fork`
824 tracepoints, and the man:open(2) and man:close(2) system calls:
829 # lttng enable-event --kernel sched_switch,sched_process_fork
830 # lttng enable-event --kernel --syscall open,close
834 You can also create an event rule which matches _all_ the Linux kernel
835 tracepoints (this will generate a lot of data when tracing):
840 # lttng enable-event --kernel --all
844 . <<basic-tracing-session-control,Start tracing>>:
853 . Do some operation on your system for a few seconds. For example,
854 load a website, or list the files of a directory.
855 . <<basic-tracing-session-control,Stop tracing>> and destroy the
866 The man:lttng-destroy(1) command does not destroy the trace data; it
867 only destroys the state of the tracing session.
869 . For the sake of this example, make the recorded trace accessible to
875 sudo chown -R $(whoami) /tmp/my-kernel-trace
879 See <<viewing-and-analyzing-your-traces,View and analyze the
880 recorded events>> to view the recorded events.
883 [[tracing-your-own-user-application]]
884 === Trace a user application
886 This section steps you through a simple example to trace a
887 _Hello world_ program written in C.
889 To create the traceable user application:
891 . Create the tracepoint provider header file, which defines the
892 tracepoints and the events they can generate:
898 #undef TRACEPOINT_PROVIDER
899 #define TRACEPOINT_PROVIDER hello_world
901 #undef TRACEPOINT_INCLUDE
902 #define TRACEPOINT_INCLUDE "./hello-tp.h"
904 #if !defined(_HELLO_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
907 #include <lttng/tracepoint.h>
917 ctf_string(my_string_field, my_string_arg)
918 ctf_integer(int, my_integer_field, my_integer_arg)
922 #endif /* _HELLO_TP_H */
924 #include <lttng/tracepoint-event.h>
928 . Create the tracepoint provider package source file:
934 #define TRACEPOINT_CREATE_PROBES
935 #define TRACEPOINT_DEFINE
937 #include "hello-tp.h"
941 . Build the tracepoint provider package:
946 $ gcc -c -I. hello-tp.c
950 . Create the _Hello World_ application source file:
957 #include "hello-tp.h"
959 int main(int argc, char *argv[])
963 puts("Hello, World!\nPress Enter to continue...");
966 * The following getchar() call is only placed here for the purpose
967 * of this demonstration, to pause the application in order for
968 * you to have time to list its tracepoints. It is not
974 * A tracepoint() call.
976 * Arguments, as defined in hello-tp.h:
978 * 1. Tracepoint provider name (required)
979 * 2. Tracepoint name (required)
980 * 3. my_integer_arg (first user-defined argument)
981 * 4. my_string_arg (second user-defined argument)
983 * Notice the tracepoint provider and tracepoint names are
984 * NOT strings: they are in fact parts of variables that the
985 * macros in hello-tp.h create.
987 tracepoint(hello_world, my_first_tracepoint, 23, "hi there!");
989 for (x = 0; x < argc; ++x) {
990 tracepoint(hello_world, my_first_tracepoint, x, argv[x]);
993 puts("Quitting now!");
994 tracepoint(hello_world, my_first_tracepoint, x * x, "x^2");
1001 . Build the application:
1010 . Link the application with the tracepoint provider package,
1011 `liblttng-ust`, and `libdl`:
1016 $ gcc -o hello hello.o hello-tp.o -llttng-ust -ldl
1020 Here's the whole build process:
1023 .User space tracing tutorial's build steps.
1024 image::ust-flow.png[]
1026 To trace the user application:
1028 . Run the application with a few arguments:
1033 $ ./hello world and beyond
1042 Press Enter to continue...
1046 . Start an LTTng <<lttng-sessiond,session daemon>>:
1051 $ lttng-sessiond --daemonize
1055 Note that a session daemon might already be running, for example as
1056 a service that the distribution's service manager started.
1058 . List the available user space tracepoints:
1063 $ lttng list --userspace
1067 You see the `hello_world:my_first_tracepoint` tracepoint listed
1068 under the `./hello` process.
1070 . Create a <<tracing-session,tracing session>>:
1075 $ lttng create my-user-space-session
1079 . Create an <<event,event rule>> which matches the
1080 `hello_world:my_first_tracepoint` event name:
1085 $ lttng enable-event --userspace hello_world:my_first_tracepoint
1089 . <<basic-tracing-session-control,Start tracing>>:
1098 . Go back to the running `hello` application and press Enter. The
1099 program executes all `tracepoint()` instrumentation points and exits.
1100 . <<basic-tracing-session-control,Stop tracing>> and destroy the
1111 The man:lttng-destroy(1) command does not destroy the trace data; it
1112 only destroys the state of the tracing session.
1114 By default, LTTng saves the traces in
1115 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
1116 where +__name__+ is the tracing session name. The
1117 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
1119 See <<viewing-and-analyzing-your-traces,View and analyze the
1120 recorded events>> to view the recorded events.
1123 [[viewing-and-analyzing-your-traces]]
1124 === View and analyze the recorded events
1126 Once you have completed the <<tracing-the-linux-kernel,Trace the Linux
1127 kernel>> and <<tracing-your-own-user-application,Trace a user
1128 application>> tutorials, you can inspect the recorded events.
1130 Many tools are available to read LTTng traces:
1132 * **cmd:babeltrace** is a command-line utility which converts trace
1133 formats; it supports the format that LTTng produces, CTF, as well as a
1134 basic text output which can be ++grep++ed. The cmd:babeltrace command
1135 is part of the http://diamon.org/babeltrace[Babeltrace] project.
1136 * Babeltrace also includes
1137 **https://www.python.org/[Python] bindings** so
1138 that you can easily open and read an LTTng trace with your own script,
1139 benefiting from the power of Python.
1140 * http://tracecompass.org/[**Trace Compass**]
1141 is a graphical user interface for viewing and analyzing any type of
1142 logs or traces, including LTTng's.
1143 * https://github.com/lttng/lttng-analyses[**LTTng analyses**] is a
1144 project which includes many high-level analyses of LTTng kernel
1145 traces, like scheduling statistics, interrupt frequency distribution,
1146 top CPU usage, and more.
1148 NOTE: This section assumes that the traces recorded during the previous
1149 tutorials were saved to their default location, in the
1150 dir:{$LTTNG_HOME/lttng-traces} directory. The env:LTTNG_HOME
1151 environment variable defaults to `$HOME` if not set.
1154 [[viewing-and-analyzing-your-traces-bt]]
1155 ==== Use the cmd:babeltrace command-line tool
1157 The simplest way to list all the recorded events of a trace is to pass
1158 its path to cmd:babeltrace with no options:
1162 $ babeltrace ~/lttng-traces/my-user-space-session*
1165 cmd:babeltrace finds all traces recursively within the given path and
1166 prints all their events, merging them in chronological order.
1168 You can pipe the output of cmd:babeltrace into a tool like man:grep(1) for
1173 $ babeltrace /tmp/my-kernel-trace | grep _switch
1176 You can pipe the output of cmd:babeltrace into a tool like man:wc(1) to
1177 count the recorded events:
1181 $ babeltrace /tmp/my-kernel-trace | grep _open | wc --lines
1185 [[viewing-and-analyzing-your-traces-bt-python]]
1186 ==== Use the Babeltrace Python bindings
1188 The <<viewing-and-analyzing-your-traces-bt,text output of cmd:babeltrace>>
1189 is useful to isolate events by simple matching using man:grep(1) and
1190 similar utilities. However, more elaborate filters, such as keeping only
1191 event records with a field value falling within a specific range, are
1192 not trivial to write using a shell. Moreover, reductions and even the
1193 most basic computations involving multiple event records are virtually
1194 impossible to implement.
1196 Fortunately, Babeltrace ships with Python 3 bindings which makes it easy
1197 to read the event records of an LTTng trace sequentially and compute the
1198 desired information.
1200 The following script accepts an LTTng Linux kernel trace path as its
1201 first argument and prints the short names of the top 5 running processes
1202 on CPU 0 during the whole trace:
1207 from collections import Counter
1213 if len(sys.argv) != 2:
1214 msg = 'Usage: python3 {} TRACEPATH'.format(sys.argv[0])
1215 print(msg, file=sys.stderr)
1218 # A trace collection contains one or more traces
1219 col = babeltrace.TraceCollection()
1221 # Add the trace provided by the user (LTTng traces always have
1223 if col.add_trace(sys.argv[1], 'ctf') is None:
1224 raise RuntimeError('Cannot add trace')
1226 # This counter dict contains execution times:
1228 # task command name -> total execution time (ns)
1229 exec_times = Counter()
1231 # This contains the last `sched_switch` timestamp
1235 for event in col.events:
1236 # Keep only `sched_switch` events
1237 if event.name != 'sched_switch':
1240 # Keep only events which happened on CPU 0
1241 if event['cpu_id'] != 0:
1245 cur_ts = event.timestamp
1251 # Previous task command (short) name
1252 prev_comm = event['prev_comm']
1254 # Initialize entry in our dict if not yet done
1255 if prev_comm not in exec_times:
1256 exec_times[prev_comm] = 0
1258 # Compute previous command execution time
1259 diff = cur_ts - last_ts
1261 # Update execution time of this command
1262 exec_times[prev_comm] += diff
1264 # Update last timestamp
1268 for name, ns in exec_times.most_common(5):
1270 print('{:20}{} s'.format(name, s))
1275 if __name__ == '__main__':
1276 sys.exit(0 if top5proc() else 1)
1283 $ python3 top5proc.py /tmp/my-kernel-trace/kernel
1289 swapper/0 48.607245889 s
1290 chromium 7.192738188 s
1291 pavucontrol 0.709894415 s
1292 Compositor 0.660867933 s
1293 Xorg.bin 0.616753786 s
1296 Note that `swapper/0` is the "idle" process of CPU 0 on Linux; since we
1297 weren't using the CPU that much when tracing, its first position in the
1302 == [[understanding-lttng]]Core concepts
1304 From a user's perspective, the LTTng system is built on a few concepts,
1305 or objects, on which the <<lttng-cli,cmd:lttng command-line tool>>
1306 operates by sending commands to the <<lttng-sessiond,session daemon>>.
1307 Understanding how those objects relate to eachother is key in mastering
1310 The core concepts are:
1312 * <<tracing-session,Tracing session>>
1313 * <<domain,Tracing domain>>
1314 * <<channel,Channel and ring buffer>>
1315 * <<"event","Instrumentation point, event rule, event, and event record">>
1321 A _tracing session_ is a stateful dialogue between you and
1322 a <<lttng-sessiond,session daemon>>. You can
1323 <<creating-destroying-tracing-sessions,create a new tracing
1324 session>> with the `lttng create` command.
1326 Anything that you do when you control LTTng tracers happens within a
1327 tracing session. In particular, a tracing session:
1330 * Has its own set of trace files.
1331 * Has its own state of activity (started or stopped).
1332 * Has its own <<tracing-session-mode,mode>> (local, network streaming,
1334 * Has its own <<channel,channels>> which have their own
1335 <<event,event rules>>.
1338 .A _tracing session_ contains <<channel,channels>> that are members of <<domain,tracing domains>> and contain <<event,event rules>>.
1339 image::concepts.png[]
1341 Those attributes and objects are completely isolated between different
1344 A tracing session is analogous to a cash machine session:
1345 the operations you do on the banking system through the cash machine do
1346 not alter the data of other users of the same system. In the case of
1347 the cash machine, a session lasts as long as your bank card is inside.
1348 In the case of LTTng, a tracing session lasts from the `lttng create`
1349 command to the `lttng destroy` command.
1352 .Each Unix user has its own set of tracing sessions.
1353 image::many-sessions.png[]
1356 [[tracing-session-mode]]
1357 ==== Tracing session mode
1359 LTTng can send the generated trace data to different locations. The
1360 _tracing session mode_ dictates where to send it. The following modes
1361 are available in LTTng{nbsp}{revision}:
1364 LTTng writes the traces to the file system of the machine being traced
1367 Network streaming mode::
1368 LTTng sends the traces over the network to a
1369 <<lttng-relayd,relay daemon>> running on a remote system.
1372 LTTng does not write the traces by default. Instead, you can request
1373 LTTng to <<taking-a-snapshot,take a snapshot>>, that is, a copy of the
1374 current tracing buffers, and to write it to the target's file system
1375 or to send it over the network to a <<lttng-relayd,relay daemon>>
1376 running on a remote system.
1379 This mode is similar to the network streaming mode, but a live
1380 trace viewer can connect to the distant relay daemon to
1381 <<lttng-live,view event records as LTTng generates them>> by
1388 A _tracing domain_ is a namespace for event sources. A tracing domain
1389 has its own properties and features.
1391 There are currently five available tracing domains:
1395 * `java.util.logging` (JUL)
1399 You must specify a tracing domain when using some commands to avoid
1400 ambiguity. For example, since all the domains support named tracepoints
1401 as event sources (instrumentation points that you manually insert in the
1402 source code), you need to specify a tracing domain when
1403 <<enabling-disabling-events,creating an event rule>> because all the
1404 tracing domains could have tracepoints with the same names.
1406 Some features are reserved to specific tracing domains. Dynamic function
1407 entry and return instrumentation points, for example, are currently only
1408 supported in the Linux kernel tracing domain, but support for other
1409 tracing domains could be added in the future.
1411 You can create <<channel,channels>> in the Linux kernel and user space
1412 tracing domains. The other tracing domains have a single default
1417 === Channel and ring buffer
1419 A _channel_ is an object which is responsible for a set of ring buffers.
1420 Each ring buffer is divided into multiple sub-buffers. When an LTTng
1421 tracer emits an event, it can record it to one or more
1422 sub-buffers. The attributes of a channel determine what to do when
1423 there's no space left for a new event record because all sub-buffers
1424 are full, where to send a full sub-buffer, and other behaviours.
1426 A channel is always associated to a <<domain,tracing domain>>. The
1427 `java.util.logging` (JUL), log4j, and Python tracing domains each have
1428 a default channel which you cannot configure.
1430 A channel also owns <<event,event rules>>. When an LTTng tracer emits
1431 an event, it records it to the sub-buffers of all
1432 the enabled channels with a satisfied event rule, as long as those
1433 channels are part of active <<tracing-session,tracing sessions>>.
1436 [[channel-buffering-schemes]]
1437 ==== Per-user vs. per-process buffering schemes
1439 A channel has at least one ring buffer _per CPU_. LTTng always
1440 records an event to the ring buffer associated to the CPU on which it
1443 Two _buffering schemes_ are available when you
1444 <<enabling-disabling-channels,create a channel>> in the
1445 user space <<domain,tracing domain>>:
1447 Per-user buffering::
1448 Allocate one set of ring buffers--one per CPU--shared by all the
1449 instrumented processes of each Unix user.
1453 .Per-user buffering scheme.
1454 image::per-user-buffering.png[]
1457 Per-process buffering::
1458 Allocate one set of ring buffers--one per CPU--for each
1459 instrumented process.
1463 .Per-process buffering scheme.
1464 image::per-process-buffering.png[]
1467 The per-process buffering scheme tends to consume more memory than the
1468 per-user option because systems generally have more instrumented
1469 processes than Unix users running instrumented processes. However, the
1470 per-process buffering scheme ensures that one process having a high
1471 event throughput won't fill all the shared sub-buffers of the same
1474 The Linux kernel tracing domain has only one available buffering scheme
1475 which is to allocate a single set of ring buffers for the whole system.
1476 This scheme is similar to the per-user option, but with a single, global
1477 user "running" the kernel.
1480 [[channel-overwrite-mode-vs-discard-mode]]
1481 ==== Overwrite vs. discard event loss modes
1483 When an event occurs, LTTng records it to a specific sub-buffer (yellow
1484 arc in the following animation) of a specific channel's ring buffer.
1485 When there's no space left in a sub-buffer, the tracer marks it as
1486 consumable (red) and another, empty sub-buffer starts receiving the
1487 following event records. A <<lttng-consumerd,consumer daemon>>
1488 eventually consumes the marked sub-buffer (returns to white).
1491 [role="docsvg-channel-subbuf-anim"]
1496 In an ideal world, sub-buffers are consumed faster than they are filled,
1497 as is the case in the previous animation. In the real world,
1498 however, all sub-buffers can be full at some point, leaving no space to
1499 record the following events.
1501 By design, LTTng is a _non-blocking_ tracer: when no empty sub-buffer is
1502 available, it is acceptable to lose event records when the alternative
1503 would be to cause substantial delays in the instrumented application's
1504 execution. LTTng privileges performance over integrity; it aims at
1505 perturbing the traced system as little as possible in order to make
1506 tracing of subtle race conditions and rare interrupt cascades possible.
1508 When it comes to losing event records because no empty sub-buffer is
1509 available, the channel's _event loss mode_ determines what to do. The
1510 available event loss modes are:
1513 Drop the newest event records until a the tracer
1514 releases a sub-buffer.
1517 Clear the sub-buffer containing the oldest event records and start
1518 writing the newest event records there.
1520 This mode is sometimes called _flight recorder mode_ because it's
1522 https://en.wikipedia.org/wiki/Flight_recorder[flight recorder]:
1523 always keep a fixed amount of the latest data.
1525 Which mechanism you should choose depends on your context: prioritize
1526 the newest or the oldest event records in the ring buffer?
1528 Beware that, in overwrite mode, the tracer abandons a whole sub-buffer
1529 as soon as a there's no space left for a new event record, whereas in
1530 discard mode, the tracer only discards the event record that doesn't
1533 In discard mode, LTTng increments a count of lost event records when
1534 an event record is lost and saves this count to the trace. In
1535 overwrite mode, LTTng keeps no information when it overwrites a
1536 sub-buffer before consuming it.
1538 There are a few ways to decrease your probability of losing event
1540 <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>> shows
1541 how you can fine-une the sub-buffer count and size of a channel to
1542 virtually stop losing event records, though at the cost of greater
1546 [[channel-subbuf-size-vs-subbuf-count]]
1547 ==== Sub-buffer count and size
1549 When you <<enabling-disabling-channels,create a channel>>, you can
1550 set its number of sub-buffers and their size.
1552 Note that there is noticeable CPU overhead introduced when
1553 switching sub-buffers (marking a full one as consumable and switching
1554 to an empty one for the following events to be recorded). Knowing this,
1555 the following list presents a few practical situations along with how
1556 to configure the sub-buffer count and size for them:
1558 * **High event throughput**: In general, prefer bigger sub-buffers to
1559 lower the risk of losing event records.
1561 Having bigger sub-buffers also ensures a lower sub-buffer switching
1564 The number of sub-buffers is only meaningful if you create the channel
1565 in overwrite mode: in this case, if a sub-buffer overwrite happens, the
1566 other sub-buffers are left unaltered.
1568 * **Low event throughput**: In general, prefer smaller sub-buffers
1569 since the risk of losing event records is low.
1571 Because events occur less frequently, the sub-buffer switching frequency
1572 should remain low and thus the tracer's overhead should not be a
1575 * **Low memory system**: If your target system has a low memory
1576 limit, prefer fewer first, then smaller sub-buffers.
1578 Even if the system is limited in memory, you want to keep the
1579 sub-buffers as big as possible to avoid a high sub-buffer switching
1582 Note that LTTng uses http://diamon.org/ctf/[CTF] as its trace format,
1583 which means event data is very compact. For example, the average
1584 LTTng kernel event record weights about 32{nbsp}bytes. Thus, a
1585 sub-buffer size of 1{nbsp}MiB is considered big.
1587 The previous situations highlight the major trade-off between a few big
1588 sub-buffers and more, smaller sub-buffers: sub-buffer switching
1589 frequency vs. how much data is lost in overwrite mode. Assuming a
1590 constant event throughput and using the overwrite mode, the two
1591 following configurations have the same ring buffer total size:
1594 [role="docsvg-channel-subbuf-size-vs-count-anim"]
1599 * **2 sub-buffers of 4{nbsp}MiB each**: Expect a very low sub-buffer
1600 switching frequency, but if a sub-buffer overwrite happens, half of
1601 the event records so far (4{nbsp}MiB) are definitely lost.
1602 * **8 sub-buffers of 1{nbsp}MiB each**: Expect 4{nbsp}times the tracer's
1603 overhead as the previous configuration, but if a sub-buffer
1604 overwrite happens, only the eighth of event records so far are
1607 In discard mode, the sub-buffers count parameter is pointless: use two
1608 sub-buffers and set their size according to the requirements of your
1612 [[channel-switch-timer]]
1613 ==== Switch timer period
1615 The _switch timer period_ is an important configurable attribute of
1616 a channel to ensure periodic sub-buffer flushing.
1618 When the _switch timer_ expires, a sub-buffer switch happens. You can
1619 set the switch timer period attribute when you
1620 <<enabling-disabling-channels,create a channel>> to ensure that event
1621 data is consumed and committed to trace files or to a distant relay
1622 daemon periodically in case of a low event throughput.
1625 [role="docsvg-channel-switch-timer"]
1630 This attribute is also convenient when you use big sub-buffers to cope
1631 with a sporadic high event throughput, even if the throughput is
1635 [[channel-read-timer]]
1636 ==== Read timer period
1638 By default, the LTTng tracers use a notification mechanism to signal a
1639 full sub-buffer so that a consumer daemon can consume it. When such
1640 notifications must be avoided, for example in real-time applications,
1641 you can use the channel's _read timer_ instead. When the read timer
1642 fires, the <<lttng-consumerd,consumer daemon>> checks for full,
1643 consumable sub-buffers.
1646 [[tracefile-rotation]]
1647 ==== Trace file count and size
1649 By default, trace files can grow as large as needed. You can set the
1650 maximum size of each trace file that a channel writes when you
1651 <<enabling-disabling-channels,create a channel>>. When the size of
1652 a trace file reaches the channel's fixed maximum size, LTTng creates
1653 another file to contain the next event records. LTTng appends a file
1654 count to each trace file name in this case.
1656 If you set the trace file size attribute when you create a channel, the
1657 maximum number of trace files that LTTng creates is _unlimited_ by
1658 default. To limit them, you can also set a maximum number of trace
1659 files. When the number of trace files reaches the channel's fixed
1660 maximum count, the oldest trace file is overwritten. This mechanism is
1661 called _trace file rotation_.
1665 === Instrumentation point, event rule, event, and event record
1667 An _event rule_ is a set of conditions which must be **all** satisfied
1668 for LTTng to record an occuring event.
1670 You set the conditions when you <<enabling-disabling-events,create
1673 You always attach an event rule to <<channel,channel>> when you create
1676 When an event passes the conditions of an event rule, LTTng records it
1677 in one of the attached channel's sub-buffers.
1679 The available conditions, as of LTTng{nbsp}{revision}, are:
1681 * The event rule _is enabled_.
1682 * The instrumentation point's type _is{nbsp}T_.
1683 * The instrumentation point's name (sometimes called _event name_)
1684 _matches{nbsp}N_, but _is not{nbsp}E_.
1685 * The instrumentation point's log level _is as severe as{nbsp}L_, or
1686 _is exactly{nbsp}L_.
1687 * The fields of the event's payload _satisfy_ a filter
1688 expression{nbsp}__F__.
1690 As you can see, all the conditions but the dynamic filter are related to
1691 the event rule's status or to the instrumentation point, not to the
1692 occurring events. This is why, without a filter, checking if an event
1693 passes an event rule is not a dynamic task: when you create or modify an
1694 event rule, all the tracers of its tracing domain enable or disable the
1695 instrumentation points themselves once. This is possible because the
1696 attributes of an instrumentation point (type, name, and log level) are
1697 defined statically. In other words, without a dynamic filter, the tracer
1698 _does not evaluate_ the arguments of an instrumentation point unless it
1699 matches an enabled event rule.
1701 Note that, for LTTng to record an event, the <<channel,channel>> to
1702 which a matching event rule is attached must also be enabled, and the
1703 tracing session owning this channel must be active.
1706 .Logical path from an instrumentation point to an event record.
1707 image::event-rule.png[]
1709 .Event, event record, or event rule?
1711 With so many similar terms, it's easy to get confused.
1713 An **event** is the consequence of the execution of an _instrumentation
1714 point_, like a tracepoint that you manually place in some source code,
1715 or a Linux kernel KProbe. An event is said to _occur_ at a specific
1716 time. Different actions can be taken upon the occurance of an event,
1717 like record the event's payload to a buffer.
1719 An **event record** is the representation of an event in a sub-buffer. A
1720 tracer is responsible for capturing the payload of an event, current
1721 context variables, the event's ID, and the event's timestamp. LTTng
1722 can append this sub-buffer to a trace file.
1724 An **event rule** is a set of conditions which must all be satisfied for
1725 LTTng to record an occuring event. Events still occur without
1726 satisfying event rules, but LTTng does not record them.
1731 == Components of noch:{LTTng}
1733 The second _T_ in _LTTng_ stands for _toolkit_: it would be wrong
1734 to call LTTng a simple _tool_ since it is composed of multiple
1735 interacting components. This section describes those components,
1736 explains their respective roles, and shows how they connect together to
1737 form the LTTng ecosystem.
1739 The following diagram shows how the most important components of LTTng
1740 interact with user applications, the Linux kernel, and you:
1743 .Control and trace data paths between LTTng components.
1744 image::plumbing.png[]
1746 The LTTng project incorporates:
1748 * **LTTng-tools**: Libraries and command-line interface to
1749 control tracing sessions.
1750 ** <<lttng-sessiond,Session daemon>> (man:lttng-sessiond(8)).
1751 ** <<lttng-consumerd,Consumer daemon>> (man:lttng-consumerd(8)).
1752 ** <<lttng-relayd,Relay daemon>> (man:lttng-relayd(8)).
1753 ** <<liblttng-ctl-lttng,Tracing control library>> (`liblttng-ctl`).
1754 ** <<lttng-cli,Tracing control command-line tool>> (man:lttng(1)).
1755 * **LTTng-UST**: Libraries and Java/Python packages to trace user
1757 ** <<lttng-ust,User space tracing library>> (`liblttng-ust`) and its
1758 headers to instrument and trace any native user application.
1759 ** <<prebuilt-ust-helpers,Preloadable user space tracing helpers>>:
1760 *** `liblttng-ust-libc-wrapper`
1761 *** `liblttng-ust-pthread-wrapper`
1762 *** `liblttng-ust-cyg-profile`
1763 *** `liblttng-ust-cyg-profile-fast`
1764 *** `liblttng-ust-dl`
1765 ** User space tracepoint provider source files generator command-line
1766 tool (man:lttng-gen-tp(1)).
1767 ** <<lttng-ust-agents,LTTng-UST Java agent>> to instrument and trace
1768 Java applications using `java.util.logging` or
1769 Apache log4j 1.2 logging.
1770 ** <<lttng-ust-agents,LTTng-UST Python agent>> to instrument
1771 Python applications using the standard `logging` package.
1772 * **LTTng-modules**: <<lttng-modules,Linux kernel modules>> to trace
1774 ** LTTng kernel tracer module.
1775 ** Tracing ring buffer kernel modules.
1776 ** Probe kernel modules.
1777 ** LTTng logger kernel module.
1781 === Tracing control command-line interface
1784 .The tracing control command-line interface.
1785 image::plumbing-lttng-cli.png[]
1787 The _man:lttng(1) command-line tool_ is the standard user interface to
1788 control LTTng <<tracing-session,tracing sessions>>. The cmd:lttng tool
1789 is part of LTTng-tools.
1791 The cmd:lttng tool is linked with
1792 <<liblttng-ctl-lttng,`liblttng-ctl`>> to communicate with
1793 one or more <<lttng-sessiond,session daemons>> behind the scenes.
1795 The cmd:lttng tool has a Git-like interface:
1799 $ lttng <GENERAL OPTIONS> <COMMAND> <COMMAND OPTIONS>
1802 The <<controlling-tracing,Tracing control>> section explores the
1803 available features of LTTng using the cmd:lttng tool.
1806 [[liblttng-ctl-lttng]]
1807 === Tracing control library
1810 .The tracing control library.
1811 image::plumbing-liblttng-ctl.png[]
1813 The _LTTng control library_, `liblttng-ctl`, is used to communicate
1814 with a <<lttng-sessiond,session daemon>> using a C API that hides the
1815 underlying protocol's details. `liblttng-ctl` is part of LTTng-tools.
1817 The <<lttng-cli,cmd:lttng command-line tool>>
1818 is linked with `liblttng-ctl`.
1820 You can use `liblttng-ctl` in C or $$C++$$ source code by including its
1825 #include <lttng/lttng.h>
1828 Some objects are referenced by name (C string), such as tracing
1829 sessions, but most of them require to create a handle first using
1830 `lttng_create_handle()`.
1832 The best available developer documentation for `liblttng-ctl` is, as of
1833 LTTng{nbsp}{revision}, its installed header files. Every function and
1834 structure is thoroughly documented.
1838 === User space tracing library
1841 .The user space tracing library.
1842 image::plumbing-liblttng-ust.png[]
1844 The _user space tracing library_, `liblttng-ust` (see man:lttng-ust(3)),
1845 is the LTTng user space tracer. It receives commands from a
1846 <<lttng-sessiond,session daemon>>, for example to
1847 enable and disable specific instrumentation points, and writes event
1848 records to ring buffers shared with a
1849 <<lttng-consumerd,consumer daemon>>.
1850 `liblttng-ust` is part of LTTng-UST.
1852 Public C header files are installed beside `liblttng-ust` to
1853 instrument any <<c-application,C or $$C++$$ application>>.
1855 <<lttng-ust-agents,LTTng-UST agents>>, which are regular Java and Python
1856 packages, use their own library providing tracepoints which is
1857 linked with `liblttng-ust`.
1859 An application or library does not have to initialize `liblttng-ust`
1860 manually: its constructor does the necessary tasks to properly register
1861 to a session daemon. The initialization phase also enables the
1862 instrumentation points matching the <<event,event rules>> that you
1866 [[lttng-ust-agents]]
1867 === User space tracing agents
1870 .The user space tracing agents.
1871 image::plumbing-lttng-ust-agents.png[]
1873 The _LTTng-UST Java and Python agents_ are regular Java and Python
1874 packages which add LTTng tracing capabilities to the
1875 native logging frameworks. The LTTng-UST agents are part of LTTng-UST.
1877 In the case of Java, the
1878 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[`java.util.logging`
1879 core logging facilities] and
1880 https://logging.apache.org/log4j/1.2/[Apache log4j 1.2] are supported.
1881 Note that Apache Log4{nbsp}2 is not supported.
1883 In the case of Python, the standard
1884 https://docs.python.org/3/library/logging.html[`logging`] package
1885 is supported. Both Python 2 and Python 3 modules can import the
1886 LTTng-UST Python agent package.
1888 The applications using the LTTng-UST agents are in the
1889 `java.util.logging` (JUL),
1890 log4j, and Python <<domain,tracing domains>>.
1892 Both agents use the same mechanism to trace the log statements. When an
1893 agent is initialized, it creates a log handler that attaches to the root
1894 logger. The agent also registers to a <<lttng-sessiond,session daemon>>.
1895 When the application executes a log statement, it is passed to the
1896 agent's log handler by the root logger. The agent's log handler calls a
1897 native function in a tracepoint provider package shared library linked
1898 with <<lttng-ust,`liblttng-ust`>>, passing the formatted log message and
1899 other fields, like its logger name and its log level. This native
1900 function contains a user space instrumentation point, hence tracing the
1903 The log level condition of an
1904 <<event,event rule>> is considered when tracing
1905 a Java or a Python application, and it's compatible with the standard
1906 JUL, log4j, and Python log levels.
1910 === LTTng kernel modules
1913 .The LTTng kernel modules.
1914 image::plumbing-lttng-modules.png[]
1916 The _LTTng kernel modules_ are a set of Linux kernel modules
1917 which implement the kernel tracer of the LTTng project. The LTTng
1918 kernel modules are part of LTTng-modules.
1920 The LTTng kernel modules include:
1922 * A set of _probe_ modules.
1924 Each module attaches to a specific subsystem
1925 of the Linux kernel using its tracepoint instrument points. There are
1926 also modules to attach to the entry and return points of the Linux
1927 system call functions.
1929 * _Ring buffer_ modules.
1931 A ring buffer implementation is provided as kernel modules. The LTTng
1932 kernel tracer writes to the ring buffer; a
1933 <<lttng-consumerd,consumer daemon>> reads from the ring buffer.
1935 * The _LTTng kernel tracer_ module.
1936 * The _LTTng logger_ module.
1938 The LTTng logger module implements the special path:{/proc/lttng-logger}
1939 file so that any executable can generate LTTng events by opening and
1940 writing to this file.
1942 See <<proc-lttng-logger-abi,LTTng logger>>.
1944 Generally, you do not have to load the LTTng kernel modules manually
1945 (using man:modprobe(8), for example): a root <<lttng-sessiond,session
1946 daemon>> loads the necessary modules when starting. If you have extra
1947 probe modules, you can specify to load them to the session daemon on
1950 The LTTng kernel modules are installed in
1951 +/usr/lib/modules/__release__/extra+ by default, where +__release__+ is
1952 the kernel release (see `uname --kernel-release`).
1959 .The session daemon.
1960 image::plumbing-sessiond.png[]
1962 The _session daemon_, man:lttng-sessiond(8), is a daemon responsible for
1963 managing tracing sessions and for controlling the various components of
1964 LTTng. The session daemon is part of LTTng-tools.
1966 The session daemon sends control requests to and receives control
1969 * The <<lttng-ust,user space tracing library>>.
1971 Any instance of the user space tracing library first registers to
1972 a session daemon. Then, the session daemon can send requests to
1973 this instance, such as:
1976 ** Get the list of tracepoints.
1977 ** Share an <<event,event rule>> so that the user space tracing library
1978 can enable or disable tracepoints. Amongst the possible conditions
1979 of an event rule is a filter expression which `liblttng-ust` evalutes
1980 when an event occurs.
1981 ** Share <<channel,channel>> attributes and ring buffer locations.
1984 The session daemon and the user space tracing library use a Unix
1985 domain socket for their communication.
1987 * The <<lttng-ust-agents,user space tracing agents>>.
1989 Any instance of a user space tracing agent first registers to
1990 a session daemon. Then, the session daemon can send requests to
1991 this instance, such as:
1994 ** Get the list of loggers.
1995 ** Enable or disable a specific logger.
1998 The session daemon and the user space tracing agent use a TCP connection
1999 for their communication.
2001 * The <<lttng-modules,LTTng kernel tracer>>.
2002 * The <<lttng-consumerd,consumer daemon>>.
2004 The session daemon sends requests to the consumer daemon to instruct
2005 it where to send the trace data streams, amongst other information.
2007 * The <<lttng-relayd,relay daemon>>.
2009 The session daemon receives commands from the
2010 <<liblttng-ctl-lttng,tracing control library>>.
2012 The root session daemon loads the appropriate
2013 <<lttng-modules,LTTng kernel modules>> on startup. It also spawns
2014 a <<lttng-consumerd,consumer daemon>> as soon as you create
2015 an <<event,event rule>>.
2017 The session daemon does not send and receive trace data: this is the
2018 role of the <<lttng-consumerd,consumer daemon>> and
2019 <<lttng-relayd,relay daemon>>. It does, however, generate the
2020 http://diamon.org/ctf/[CTF] metadata stream.
2022 Each Unix user can have its own session daemon instance. The
2023 tracing sessions managed by different session daemons are completely
2026 The root user's session daemon is the only one which is
2027 allowed to control the LTTng kernel tracer, and its spawned consumer
2028 daemon is the only one which is allowed to consume trace data from the
2029 LTTng kernel tracer. Note, however, that any Unix user which is a member
2030 of the <<tracing-group,tracing group>> is allowed
2031 to create <<channel,channels>> in the
2032 Linux kernel <<domain,tracing domain>>, and thus to trace the Linux
2035 The <<lttng-cli,cmd:lttng command-line tool>> automatically starts a
2036 session daemon when using its `create` command if none is currently
2037 running. You can also start the session daemon manually.
2044 .The consumer daemon.
2045 image::plumbing-consumerd.png[]
2047 The _consumer daemon_, man:lttng-consumerd(8), is a daemon which shares
2048 ring buffers with user applications or with the LTTng kernel modules to
2049 collect trace data and send it to some location (on disk or to a
2050 <<lttng-relayd,relay daemon>> over the network). The consumer daemon
2051 is part of LTTng-tools.
2053 You do not start a consumer daemon manually: a consumer daemon is always
2054 spawned by a <<lttng-sessiond,session daemon>> as soon as you create an
2055 <<event,event rule>>, that is, before you start tracing. When you kill
2056 its owner session daemon, the consumer daemon also exits because it is
2057 the session daemon's child process. Command-line options of
2058 man:lttng-sessiond(8) target the consumer daemon process.
2060 There are up to two running consumer daemons per Unix user, whereas only
2061 one session daemon can run per user. This is because each process can be
2062 either 32-bit or 64-bit: if the target system runs a mixture of 32-bit
2063 and 64-bit processes, it is more efficient to have separate
2064 corresponding 32-bit and 64-bit consumer daemons. The root user is an
2065 exception: it can have up to _three_ running consumer daemons: 32-bit
2066 and 64-bit instances for its user applications, and one more
2067 reserved for collecting kernel trace data.
2075 image::plumbing-relayd.png[]
2077 The _relay daemon_, man:lttng-relayd(8), is a daemon acting as a bridge
2078 between remote session and consumer daemons, local trace files, and a
2079 remote live trace viewer. The relay daemon is part of LTTng-tools.
2081 The main purpose of the relay daemon is to implement a receiver of
2082 <<sending-trace-data-over-the-network,trace data over the network>>.
2083 This is useful when the target system does not have much file system
2084 space to record trace files locally.
2086 The relay daemon is also a server to which a
2087 <<lttng-live,live trace viewer>> can
2088 connect. The live trace viewer sends requests to the relay daemon to
2089 receive trace data as the target system emits events. The
2090 communication protocol is named _LTTng live_; it is used over TCP
2093 Note that you can start the relay daemon on the target system directly.
2094 This is the setup of choice when the use case is to view events as
2095 the target system emits them without the need of a remote system.
2099 == [[using-lttng]]Instrumentation
2101 There are many examples of tracing and monitoring in our everyday life:
2103 * You have access to real-time and historical weather reports and
2104 forecasts thanks to weather stations installed around the country.
2105 * You know your heart is safe thanks to an electrocardiogram.
2106 * You make sure not to drive your car too fast and to have enough fuel
2107 to reach your destination thanks to gauges visible on your dashboard.
2109 All the previous examples have something in common: they rely on
2110 **instruments**. Without the electrodes attached to the surface of your
2111 body's skin, cardiac monitoring is futile.
2113 LTTng, as a tracer, is no different from those real life examples. If
2114 you're about to trace a software system or, in other words, record its
2115 history of execution, you better have **instrumentation points** in the
2116 subject you're tracing, that is, the actual software.
2118 Various ways were developed to instrument a piece of software for LTTng
2119 tracing. The most straightforward one is to manually place
2120 instrumentation points, called _tracepoints_, in the software's source
2121 code. It is also possible to add instrumentation points dynamically in
2122 the Linux kernel <<domain,tracing domain>>.
2124 If you're only interested in tracing the Linux kernel, your
2125 instrumentation needs are probably already covered by LTTng's built-in
2126 <<lttng-modules,Linux kernel tracepoints>>. You may also wish to trace a
2127 user application which is already instrumented for LTTng tracing.
2128 In such cases, you can skip this whole section and read the topics of
2129 the <<controlling-tracing,Tracing control>> section.
2131 Many methods are available to instrument a piece of software for LTTng
2134 * <<c-application,User space instrumentation for C and $$C++$$
2136 * <<prebuilt-ust-helpers,Prebuilt user space tracing helpers>>.
2137 * <<java-application,User space Java agent>>.
2138 * <<python-application,User space Python agent>>.
2139 * <<proc-lttng-logger-abi,LTTng logger>>.
2140 * <<instrumenting-linux-kernel,LTTng kernel tracepoints>>.
2144 === [[cxx-application]]User space instrumentation for C and $$C++$$ applications
2146 The procedure to instrument a C or $$C++$$ user application with
2147 the <<lttng-ust,LTTng user space tracing library>>, `liblttng-ust`, is:
2149 . <<tracepoint-provider,Create the source files of a tracepoint provider
2151 . <<probing-the-application-source-code,Add tracepoints to
2152 the application's source code>>.
2153 . <<building-tracepoint-providers-and-user-application,Build and link
2154 a tracepoint provider package and the user application>>.
2156 If you need quick, man:printf(3)-like instrumentation, you can skip
2157 those steps and use <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>>
2160 IMPORTANT: You need to <<installing-lttng,install>> LTTng-UST to
2161 instrument a user application with `liblttng-ust`.
2164 [[tracepoint-provider]]
2165 ==== Create the source files of a tracepoint provider package
2167 A _tracepoint provider_ is a set of compiled functions which provide
2168 **tracepoints** to an application, the type of instrumentation point
2169 supported by LTTng-UST. Those functions can emit events with
2170 user-defined fields and serialize those events as event records to one
2171 or more LTTng-UST <<channel,channel>> sub-buffers. The `tracepoint()`
2172 macro, which you <<probing-the-application-source-code,insert in a user
2173 application's source code>>, calls those functions.
2175 A _tracepoint provider package_ is an object file (`.o`) or a shared
2176 library (`.so`) which contains one or more tracepoint providers.
2177 Its source files are:
2179 * One or more <<tpp-header,tracepoint provider header>> (`.h`).
2180 * A <<tpp-source,tracepoint provider package source>> (`.c`).
2182 A tracepoint provider package is dynamically linked with `liblttng-ust`,
2183 the LTTng user space tracer, at run time.
2186 .User application linked with `liblttng-ust` and containing a tracepoint provider.
2187 image::ust-app.png[]
2189 NOTE: If you need quick, man:printf(3)-like instrumentation, you can
2190 skip creating and using a tracepoint provider and use
2191 <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>> instead.
2195 ===== Create a tracepoint provider header file template
2197 A _tracepoint provider header file_ contains the tracepoint
2198 definitions of a tracepoint provider.
2200 To create a tracepoint provider header file:
2202 . Start from this template:
2206 .Tracepoint provider header file template (`.h` file extension).
2208 #undef TRACEPOINT_PROVIDER
2209 #define TRACEPOINT_PROVIDER provider_name
2211 #undef TRACEPOINT_INCLUDE
2212 #define TRACEPOINT_INCLUDE "./tp.h"
2214 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
2217 #include <lttng/tracepoint.h>
2220 * Use TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
2221 * TRACEPOINT_EVENT_INSTANCE(), and TRACEPOINT_LOGLEVEL() here.
2226 #include <lttng/tracepoint-event.h>
2232 * `provider_name` with the name of your tracepoint provider.
2233 * `"tp.h"` with the name of your tracepoint provider header file.
2235 . Below the `#include <lttng/tracepoint.h>` line, put your
2236 <<defining-tracepoints,tracepoint definitions>>.
2238 Your tracepoint provider name must be unique amongst all the possible
2239 tracepoint provider names used on the same target system. We
2240 suggest to include the name of your project or company in the name,
2241 for example, `org_lttng_my_project_tpp`.
2243 TIP: [[lttng-gen-tp]]You can use the man:lttng-gen-tp(1) tool to create
2244 this boilerplate for you. When using cmd:lttng-gen-tp, all you need to
2245 write are the <<defining-tracepoints,tracepoint definitions>>.
2248 [[defining-tracepoints]]
2249 ===== Create a tracepoint definition
2251 A _tracepoint definition_ defines, for a given tracepoint:
2253 * Its **input arguments**. They are the macro parameters that the
2254 `tracepoint()` macro accepts for this particular tracepoint
2255 in the user application's source code.
2256 * Its **output event fields**. They are the sources of event fields
2257 that form the payload of any event that the execution of the
2258 `tracepoint()` macro emits for this particular tracepoint.
2260 You can create a tracepoint definition by using the
2261 `TRACEPOINT_EVENT()` macro below the `#include <lttng/tracepoint.h>`
2263 <<tpp-header,tracepoint provider header file template>>.
2265 The syntax of the `TRACEPOINT_EVENT()` macro is:
2268 .`TRACEPOINT_EVENT()` macro syntax.
2271 /* Tracepoint provider name */
2274 /* Tracepoint name */
2277 /* Input arguments */
2282 /* Output event fields */
2291 * `provider_name` with your tracepoint provider name.
2292 * `tracepoint_name` with your tracepoint name.
2293 * `arguments` with the <<tpp-def-input-args,input arguments>>.
2294 * `fields` with the <<tpp-def-output-fields,output event field>>
2297 This tracepoint emits events named `provider_name:tracepoint_name`.
2300 .Event name's length limitation
2302 The concatenation of the tracepoint provider name and the
2303 tracepoint name must not exceed **254 characters**. If it does, the
2304 instrumented application compiles and runs, but LTTng throws multiple
2305 warnings and you could experience serious issues.
2308 [[tpp-def-input-args]]The syntax of the `TP_ARGS()` macro is:
2311 .`TP_ARGS()` macro syntax.
2320 * `type` with the C type of the argument.
2321 * `arg_name` with the argument name.
2323 You can repeat `type` and `arg_name` up to 10 times to have
2324 more than one argument.
2326 .`TP_ARGS()` usage with three arguments.
2338 The `TP_ARGS()` and `TP_ARGS(void)` forms are valid to create a
2339 tracepoint definition with no input arguments.
2341 [[tpp-def-output-fields]]The `TP_FIELDS()` macro contains a list of
2342 `ctf_*()` macros. Each `ctf_*()` macro defines one event field. See
2343 man:lttng-ust(3) for a complete description of the available `ctf_*()`
2344 macros. A `ctf_*()` macro specifies the type, size, and byte order of
2347 Each `ctf_*()` macro takes an _argument expression_ parameter. This is a
2348 C expression that the tracer evalutes at the `tracepoint()` macro site
2349 in the application's source code. This expression provides a field's
2350 source of data. The argument expression can include input argument names
2351 listed in the `TP_ARGS()` macro.
2353 Each `ctf_*()` macro also takes a _field name_ parameter. Field names
2354 must be unique within a given tracepoint definition.
2356 Here's a complete tracepoint definition example:
2358 .Tracepoint definition.
2360 The following tracepoint definition defines a tracepoint which takes
2361 three input arguments and has four output event fields.
2365 #include "my-custom-structure.h"
2371 const struct my_custom_structure*, my_custom_structure,
2376 ctf_string(query_field, query)
2377 ctf_float(double, ratio_field, ratio)
2378 ctf_integer(int, recv_size, my_custom_structure->recv_size)
2379 ctf_integer(int, send_size, my_custom_structure->send_size)
2384 You can refer to this tracepoint definition with the `tracepoint()`
2385 macro in your application's source code like this:
2389 tracepoint(my_provider, my_tracepoint,
2390 my_structure, some_ratio, the_query);
2394 NOTE: The LTTng tracer only evaluates tracepoint arguments at run time
2395 if they satisfy an enabled <<event,event rule>>.
2398 [[using-tracepoint-classes]]
2399 ===== Use a tracepoint class
2401 A _tracepoint class_ is a class of tracepoints which share the same
2402 output event field definitions. A _tracepoint instance_ is one
2403 instance of such a defined tracepoint class, with its own tracepoint
2406 The <<defining-tracepoints,`TRACEPOINT_EVENT()` macro>> is actually a
2407 shorthand which defines both a tracepoint class and a tracepoint
2408 instance at the same time.
2410 When you build a tracepoint provider package, the C or $$C++$$ compiler
2411 creates one serialization function for each **tracepoint class**. A
2412 serialization function is responsible for serializing the event fields
2413 of a tracepoint to a sub-buffer when tracing.
2415 For various performance reasons, when your situation requires multiple
2416 tracepoint definitions with different names, but with the same event
2417 fields, we recommend that you manually create a tracepoint class
2418 and instantiate as many tracepoint instances as needed. One positive
2419 effect of such a design, amongst other advantages, is that all
2420 tracepoint instances of the same tracepoint class reuse the same
2421 serialization function, thus reducing
2422 https://en.wikipedia.org/wiki/Cache_pollution[cache pollution].
2424 .Use a tracepoint class and tracepoint instances.
2426 Consider the following three tracepoint definitions:
2438 ctf_integer(int, userid, userid)
2439 ctf_integer(size_t, len, len)
2451 ctf_integer(int, userid, userid)
2452 ctf_integer(size_t, len, len)
2464 ctf_integer(int, userid, userid)
2465 ctf_integer(size_t, len, len)
2470 In this case, we create three tracepoint classes, with one implicit
2471 tracepoint instance for each of them: `get_account`, `get_settings`, and
2472 `get_transaction`. However, they all share the same event field names
2473 and types. Hence three identical, yet independent serialization
2474 functions are created when you build the tracepoint provider package.
2476 A better design choice is to define a single tracepoint class and three
2477 tracepoint instances:
2481 /* The tracepoint class */
2482 TRACEPOINT_EVENT_CLASS(
2483 /* Tracepoint provider name */
2486 /* Tracepoint class name */
2489 /* Input arguments */
2495 /* Output event fields */
2497 ctf_integer(int, userid, userid)
2498 ctf_integer(size_t, len, len)
2502 /* The tracepoint instances */
2503 TRACEPOINT_EVENT_INSTANCE(
2504 /* Tracepoint provider name */
2507 /* Tracepoint class name */
2510 /* Tracepoint name */
2513 /* Input arguments */
2519 TRACEPOINT_EVENT_INSTANCE(
2528 TRACEPOINT_EVENT_INSTANCE(
2541 [[assigning-log-levels]]
2542 ===== Assign a log level to a tracepoint definition
2544 You can assign an optional _log level_ to a
2545 <<defining-tracepoints,tracepoint definition>>.
2547 Assigning different levels of severity to tracepoint definitions can
2548 be useful: when you <<enabling-disabling-events,create an event rule>>,
2549 you can target tracepoints having a log level as severe as a specific
2552 The concept of LTTng-UST log levels is similar to the levels found
2553 in typical logging frameworks:
2555 * In a logging framework, the log level is given by the function
2556 or method name you use at the log statement site: `debug()`,
2557 `info()`, `warn()`, `error()`, and so on.
2558 * In LTTng-UST, you statically assign the log level to a tracepoint
2559 definition; any `tracepoint()` macro invocation which refers to
2560 this definition has this log level.
2562 You can assign a log level to a tracepoint definition with the
2563 `TRACEPOINT_LOGLEVEL()` macro. You must use this macro _after_ the
2564 <<defining-tracepoints,`TRACEPOINT_EVENT()`>> or
2565 <<using-tracepoint-classes,`TRACEPOINT_INSTANCE()`>> macro for a given
2568 The syntax of the `TRACEPOINT_LOGLEVEL()` macro is:
2571 .`TRACEPOINT_LOGLEVEL()` macro syntax.
2573 TRACEPOINT_LOGLEVEL(provider_name, tracepoint_name, log_level)
2578 * `provider_name` with the tracepoint provider name.
2579 * `tracepoint_name` with the tracepoint name.
2580 * `log_level` with the log level to assign to the tracepoint
2581 definition named `tracepoint_name` in the `provider_name`
2582 tracepoint provider.
2584 See man:lttng-ust(3) for a list of available log level names.
2586 .Assign the `TRACE_DEBUG_UNIT` log level to a tracepoint definition.
2590 /* Tracepoint definition */
2599 ctf_integer(int, userid, userid)
2600 ctf_integer(size_t, len, len)
2604 /* Log level assignment */
2605 TRACEPOINT_LOGLEVEL(my_app, get_transaction, TRACE_DEBUG_UNIT)
2611 ===== Create a tracepoint provider package source file
2613 A _tracepoint provider package source file_ is a C source file which
2614 includes a <<tpp-header,tracepoint provider header file>> to expand its
2615 macros into event serialization and other functions.
2617 You can always use the following tracepoint provider package source
2621 .Tracepoint provider package source file template.
2623 #define TRACEPOINT_CREATE_PROBES
2628 Replace `tp.h` with the name of your <<tpp-header,tracepoint provider
2629 header file>> name. You may also include more than one tracepoint
2630 provider header file here to create a tracepoint provider package
2631 holding more than one tracepoint providers.
2634 [[probing-the-application-source-code]]
2635 ==== Add tracepoints to an application's source code
2637 Once you <<tpp-header,create a tracepoint provider header file>>, you
2638 can use the `tracepoint()` macro in your application's
2639 source code to insert the tracepoints that this header
2640 <<defining-tracepoints,defines>>.
2642 The `tracepoint()` macro takes at least two parameters: the tracepoint
2643 provider name and the tracepoint name. The corresponding tracepoint
2644 definition defines the other parameters.
2646 .`tracepoint()` usage.
2648 The following <<defining-tracepoints,tracepoint definition>> defines a
2649 tracepoint which takes two input arguments and has two output event
2653 .Tracepoint provider header file.
2655 #include "my-custom-structure.h"
2662 const char*, cmd_name
2665 ctf_string(cmd_name, cmd_name)
2666 ctf_integer(int, number_of_args, argc)
2671 You can refer to this tracepoint definition with the `tracepoint()`
2672 macro in your application's source code like this:
2675 .Application's source file.
2679 int main(int argc, char* argv[])
2681 tracepoint(my_provider, my_tracepoint, argc, argv[0]);
2687 Note how the application's source code includes
2688 the tracepoint provider header file containing the tracepoint
2689 definitions to use, path:{tp.h}.
2692 .`tracepoint()` usage with a complex tracepoint definition.
2694 Consider this complex tracepoint definition, where multiple event
2695 fields refer to the same input arguments in their argument expression
2699 .Tracepoint provider header file.
2701 /* For `struct stat` */
2702 #include <sys/types.h>
2703 #include <sys/stat.h>
2715 ctf_integer(int, my_constant_field, 23 + 17)
2716 ctf_integer(int, my_int_arg_field, my_int_arg)
2717 ctf_integer(int, my_int_arg_field2, my_int_arg * my_int_arg)
2718 ctf_integer(int, sum4_field, my_str_arg[0] + my_str_arg[1] +
2719 my_str_arg[2] + my_str_arg[3])
2720 ctf_string(my_str_arg_field, my_str_arg)
2721 ctf_integer_hex(off_t, size_field, st->st_size)
2722 ctf_float(double, size_dbl_field, (double) st->st_size)
2723 ctf_sequence_text(char, half_my_str_arg_field, my_str_arg,
2724 size_t, strlen(my_str_arg) / 2)
2729 You can refer to this tracepoint definition with the `tracepoint()`
2730 macro in your application's source code like this:
2733 .Application's source file.
2735 #define TRACEPOINT_DEFINE
2742 stat("/etc/fstab", &s);
2743 tracepoint(my_provider, my_tracepoint, 23, "Hello, World!", &s);
2749 If you look at the event record that LTTng writes when tracing this
2750 program, assuming the file size of path:{/etc/fstab} is 301{nbsp}bytes,
2751 it should look like this:
2753 .Event record fields
2755 |Field's name |Field's value
2756 |`my_constant_field` |40
2757 |`my_int_arg_field` |23
2758 |`my_int_arg_field2` |529
2760 |`my_str_arg_field` |`Hello, World!`
2761 |`size_field` |0x12d
2762 |`size_dbl_field` |301.0
2763 |`half_my_str_arg_field` |`Hello,`
2767 Sometimes, the arguments you pass to `tracepoint()` are expensive to
2768 compute--they use the call stack, for example. To avoid this
2769 computation when the tracepoint is disabled, you can use the
2770 `tracepoint_enabled()` and `do_tracepoint()` macros.
2772 The syntax of the `tracepoint_enabled()` and `do_tracepoint()` macros
2776 .`tracepoint_enabled()` and `do_tracepoint()` macros syntax.
2778 tracepoint_enabled(provider_name, tracepoint_name)
2779 do_tracepoint(provider_name, tracepoint_name, ...)
2784 * `provider_name` with the tracepoint provider name.
2785 * `tracepoint_name` with the tracepoint name.
2787 `tracepoint_enabled()` returns a non-zero value if the tracepoint named
2788 `tracepoint_name` from the provider named `provider_name` is enabled
2791 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
2792 if the tracepoint is enabled. Using `tracepoint()` with
2793 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
2794 the `tracepoint_enabled()` check, thus a race condition is
2795 possible in this situation:
2798 .Possible race condition when using `tracepoint_enabled()` with `tracepoint()`.
2800 if (tracepoint_enabled(my_provider, my_tracepoint)) {
2801 stuff = prepare_stuff();
2804 tracepoint(my_provider, my_tracepoint, stuff);
2807 If the tracepoint is enabled after the condition, then `stuff` is not
2808 prepared: the emitted event will either contain wrong data, or the whole
2809 application could crash (segmentation fault, for example).
2811 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have an
2812 `STAP_PROBEV()` call. If you need it, you must emit
2816 [[building-tracepoint-providers-and-user-application]]
2817 ==== Build and link a tracepoint provider package and an application
2819 Once you have one or more <<tpp-header,tracepoint provider header
2820 files>> and a <<tpp-source,tracepoint provider package source file>>,
2821 you can create the tracepoint provider package by compiling its source
2822 file. From here, multiple build and run scenarios are possible. The
2823 following table shows common application and library configurations
2824 along with the required command lines to achieve them.
2826 In the following diagrams, we use the following file names:
2829 Executable application.
2832 Application's object file.
2835 Tracepoint provider package object file.
2838 Tracepoint provider package archive file.
2841 Tracepoint provider package shared object file.
2844 User library object file.
2847 User library shared object file.
2849 We use the following symbols in the diagrams of table below:
2852 .Symbols used in the build scenario diagrams.
2853 image::ust-sit-symbols.png[]
2855 We assume that path:{.} is part of the env:LD_LIBRARY_PATH environment
2856 variable in the following instructions.
2858 [role="growable ust-scenarios",cols="asciidoc,asciidoc"]
2859 .Common tracepoint provider package scenarios.
2861 |Scenario |Instructions
2864 The instrumented application is statically linked with
2865 the tracepoint provider package object.
2867 image::ust-sit+app-linked-with-tp-o+app-instrumented.png[]
2870 include::../common/ust-sit-step-tp-o.txt[]
2872 To build the instrumented application:
2874 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2879 #define TRACEPOINT_DEFINE
2883 . Compile the application source file:
2892 . Build the application:
2897 $ gcc -o app app.o tpp.o -llttng-ust -ldl
2901 To run the instrumented application:
2903 * Start the application:
2913 The instrumented application is statically linked with the
2914 tracepoint provider package archive file.
2916 image::ust-sit+app-linked-with-tp-a+app-instrumented.png[]
2919 To create the tracepoint provider package archive file:
2921 . Compile the <<tpp-source,tracepoint provider package source file>>:
2930 . Create the tracepoint provider package archive file:
2935 $ ar rcs tpp.a tpp.o
2939 To build the instrumented application:
2941 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2946 #define TRACEPOINT_DEFINE
2950 . Compile the application source file:
2959 . Build the application:
2964 $ gcc -o app app.o tpp.a -llttng-ust -ldl
2968 To run the instrumented application:
2970 * Start the application:
2980 The instrumented application is linked with the tracepoint provider
2981 package shared object.
2983 image::ust-sit+app-linked-with-tp-so+app-instrumented.png[]
2986 include::../common/ust-sit-step-tp-so.txt[]
2988 To build the instrumented application:
2990 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2995 #define TRACEPOINT_DEFINE
2999 . Compile the application source file:
3008 . Build the application:
3013 $ gcc -o app app.o -ldl -L. -ltpp
3017 To run the instrumented application:
3019 * Start the application:
3029 The tracepoint provider package shared object is preloaded before the
3030 instrumented application starts.
3032 image::ust-sit+tp-so-preloaded+app-instrumented.png[]
3035 include::../common/ust-sit-step-tp-so.txt[]
3037 To build the instrumented application:
3039 . In path:{app.c}, before including path:{tpp.h}, add the
3045 #define TRACEPOINT_DEFINE
3046 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3050 . Compile the application source file:
3059 . Build the application:
3064 $ gcc -o app app.o -ldl
3068 To run the instrumented application with tracing support:
3070 * Preload the tracepoint provider package shared object and
3071 start the application:
3076 $ LD_PRELOAD=./libtpp.so ./app
3080 To run the instrumented application without tracing support:
3082 * Start the application:
3092 The instrumented application dynamically loads the tracepoint provider
3093 package shared object.
3095 See the <<dlclose-warning,warning about `dlclose()`>>.
3097 image::ust-sit+app-dlopens-tp-so+app-instrumented.png[]
3100 include::../common/ust-sit-step-tp-so.txt[]
3102 To build the instrumented application:
3104 . In path:{app.c}, before including path:{tpp.h}, add the
3110 #define TRACEPOINT_DEFINE
3111 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3115 . Compile the application source file:
3124 . Build the application:
3129 $ gcc -o app app.o -ldl
3133 To run the instrumented application:
3135 * Start the application:
3145 The application is linked with the instrumented user library.
3147 The instrumented user library is statically linked with the tracepoint
3148 provider package object file.
3150 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-o+lib-instrumented.png[]
3153 include::../common/ust-sit-step-tp-o-fpic.txt[]
3155 To build the instrumented user library:
3157 . In path:{emon.c}, before including path:{tpp.h}, add the
3163 #define TRACEPOINT_DEFINE
3167 . Compile the user library source file:
3172 $ gcc -I. -fpic -c emon.c
3176 . Build the user library shared object:
3181 $ gcc -shared -o libemon.so emon.o tpp.o -llttng-ust -ldl
3185 To build the application:
3187 . Compile the application source file:
3196 . Build the application:
3201 $ gcc -o app app.o -L. -lemon
3205 To run the application:
3207 * Start the application:
3217 The application is linked with the instrumented user library.
3219 The instrumented user library is linked with the tracepoint provider
3220 package shared object.
3222 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3225 include::../common/ust-sit-step-tp-so.txt[]
3227 To build the instrumented user library:
3229 . In path:{emon.c}, before including path:{tpp.h}, add the
3235 #define TRACEPOINT_DEFINE
3239 . Compile the user library source file:
3244 $ gcc -I. -fpic -c emon.c
3248 . Build the user library shared object:
3253 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3257 To build the application:
3259 . Compile the application source file:
3268 . Build the application:
3273 $ gcc -o app app.o -L. -lemon
3277 To run the application:
3279 * Start the application:
3289 The tracepoint provider package shared object is preloaded before the
3292 The application is linked with the instrumented user library.
3294 image::ust-sit+tp-so-preloaded+app-linked-with-lib+lib-instrumented.png[]
3297 include::../common/ust-sit-step-tp-so.txt[]
3299 To build the instrumented user library:
3301 . In path:{emon.c}, before including path:{tpp.h}, add the
3307 #define TRACEPOINT_DEFINE
3308 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3312 . Compile the user library source file:
3317 $ gcc -I. -fpic -c emon.c
3321 . Build the user library shared object:
3326 $ gcc -shared -o libemon.so emon.o -ldl
3330 To build the application:
3332 . Compile the application source file:
3341 . Build the application:
3346 $ gcc -o app app.o -L. -lemon
3350 To run the application with tracing support:
3352 * Preload the tracepoint provider package shared object and
3353 start the application:
3358 $ LD_PRELOAD=./libtpp.so ./app
3362 To run the application without tracing support:
3364 * Start the application:
3374 The application is linked with the instrumented user library.
3376 The instrumented user library dynamically loads the tracepoint provider
3377 package shared object.
3379 See the <<dlclose-warning,warning about `dlclose()`>>.
3381 image::ust-sit+app-linked-with-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3384 include::../common/ust-sit-step-tp-so.txt[]
3386 To build the instrumented user library:
3388 . In path:{emon.c}, before including path:{tpp.h}, add the
3394 #define TRACEPOINT_DEFINE
3395 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3399 . Compile the user library source file:
3404 $ gcc -I. -fpic -c emon.c
3408 . Build the user library shared object:
3413 $ gcc -shared -o libemon.so emon.o -ldl
3417 To build the application:
3419 . Compile the application source file:
3428 . Build the application:
3433 $ gcc -o app app.o -L. -lemon
3437 To run the application:
3439 * Start the application:
3449 The application dynamically loads the instrumented user library.
3451 The instrumented user library is linked with the tracepoint provider
3452 package shared object.
3454 See the <<dlclose-warning,warning about `dlclose()`>>.
3456 image::ust-sit+app-dlopens-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3459 include::../common/ust-sit-step-tp-so.txt[]
3461 To build the instrumented user library:
3463 . In path:{emon.c}, before including path:{tpp.h}, add the
3469 #define TRACEPOINT_DEFINE
3473 . Compile the user library source file:
3478 $ gcc -I. -fpic -c emon.c
3482 . Build the user library shared object:
3487 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3491 To build the application:
3493 . Compile the application source file:
3502 . Build the application:
3507 $ gcc -o app app.o -ldl -L. -lemon
3511 To run the application:
3513 * Start the application:
3523 The application dynamically loads the instrumented user library.
3525 The instrumented user library dynamically loads the tracepoint provider
3526 package shared object.
3528 See the <<dlclose-warning,warning about `dlclose()`>>.
3530 image::ust-sit+app-dlopens-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3533 include::../common/ust-sit-step-tp-so.txt[]
3535 To build the instrumented user library:
3537 . In path:{emon.c}, before including path:{tpp.h}, add the
3543 #define TRACEPOINT_DEFINE
3544 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3548 . Compile the user library source file:
3553 $ gcc -I. -fpic -c emon.c
3557 . Build the user library shared object:
3562 $ gcc -shared -o libemon.so emon.o -ldl
3566 To build the application:
3568 . Compile the application source file:
3577 . Build the application:
3582 $ gcc -o app app.o -ldl -L. -lemon
3586 To run the application:
3588 * Start the application:
3598 The tracepoint provider package shared object is preloaded before the
3601 The application dynamically loads the instrumented user library.
3603 image::ust-sit+tp-so-preloaded+app-dlopens-lib+lib-instrumented.png[]
3606 include::../common/ust-sit-step-tp-so.txt[]
3608 To build the instrumented user library:
3610 . In path:{emon.c}, before including path:{tpp.h}, add the
3616 #define TRACEPOINT_DEFINE
3617 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3621 . Compile the user library source file:
3626 $ gcc -I. -fpic -c emon.c
3630 . Build the user library shared object:
3635 $ gcc -shared -o libemon.so emon.o -ldl
3639 To build the application:
3641 . Compile the application source file:
3650 . Build the application:
3655 $ gcc -o app app.o -L. -lemon
3659 To run the application with tracing support:
3661 * Preload the tracepoint provider package shared object and
3662 start the application:
3667 $ LD_PRELOAD=./libtpp.so ./app
3671 To run the application without tracing support:
3673 * Start the application:
3683 The application is statically linked with the tracepoint provider
3684 package object file.
3686 The application is linked with the instrumented user library.
3688 image::ust-sit+app-linked-with-tp-o+app-linked-with-lib+lib-instrumented.png[]
3691 include::../common/ust-sit-step-tp-o.txt[]
3693 To build the instrumented user library:
3695 . In path:{emon.c}, before including path:{tpp.h}, add the
3701 #define TRACEPOINT_DEFINE
3705 . Compile the user library source file:
3710 $ gcc -I. -fpic -c emon.c
3714 . Build the user library shared object:
3719 $ gcc -shared -o libemon.so emon.o
3723 To build the application:
3725 . Compile the application source file:
3734 . Build the application:
3739 $ gcc -o app app.o tpp.o -llttng-ust -ldl -L. -lemon
3743 To run the instrumented application:
3745 * Start the application:
3755 The application is statically linked with the tracepoint provider
3756 package object file.
3758 The application dynamically loads the instrumented user library.
3760 image::ust-sit+app-linked-with-tp-o+app-dlopens-lib+lib-instrumented.png[]
3763 include::../common/ust-sit-step-tp-o.txt[]
3765 To build the application:
3767 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3772 #define TRACEPOINT_DEFINE
3776 . Compile the application source file:
3785 . Build the application:
3790 $ gcc -Wl,--export-dynamic -o app app.o tpp.o \
3795 The `--export-dynamic` option passed to the linker is necessary for the
3796 dynamically loaded library to ``see'' the tracepoint symbols defined in
3799 To build the instrumented user library:
3801 . Compile the user library source file:
3806 $ gcc -I. -fpic -c emon.c
3810 . Build the user library shared object:
3815 $ gcc -shared -o libemon.so emon.o
3819 To run the application:
3821 * Start the application:
3833 .Do not use man:dlclose(3) on a tracepoint provider package
3835 Never use man:dlclose(3) on any shared object which:
3837 * Is linked with, statically or dynamically, a tracepoint provider
3839 * Calls man:dlopen(3) itself to dynamically open a tracepoint provider
3840 package shared object.
3842 This is currently considered **unsafe** due to a lack of reference
3843 counting from LTTng-UST to the shared object.
3845 A known workaround (available since glibc 2.2) is to use the
3846 `RTLD_NODELETE` flag when calling man:dlopen(3) initially. This has the
3847 effect of not unloading the loaded shared object, even if man:dlclose(3)
3850 You can also preload the tracepoint provider package shared object with
3851 the env:LD_PRELOAD environment variable to overcome this limitation.
3855 [[using-lttng-ust-with-daemons]]
3856 ===== Use noch:{LTTng-UST} with daemons
3858 If your instrumented application calls man:fork(2), man:clone(2),
3859 or BSD's man:rfork(2), without a following man:exec(3)-family
3860 system call, you must preload the path:{liblttng-ust-fork.so} shared
3861 object when starting the application.
3865 $ LD_PRELOAD=liblttng-ust-fork.so ./my-app
3868 If your tracepoint provider package is
3869 a shared library which you also preload, you must put both
3870 shared objects in env:LD_PRELOAD:
3874 $ LD_PRELOAD=liblttng-ust-fork.so:/path/to/tp.so ./my-app
3878 [[lttng-ust-pkg-config]]
3879 ===== Use noch:{pkg-config}
3881 On some distributions, LTTng-UST ships with a
3882 https://www.freedesktop.org/wiki/Software/pkg-config/[pkg-config]
3883 metadata file. If this is your case, then you can use cmd:pkg-config to
3884 build an application on the command line:
3888 $ gcc -o my-app my-app.o tp.o $(pkg-config --cflags --libs lttng-ust)
3892 [[instrumenting-32-bit-app-on-64-bit-system]]
3893 ===== [[advanced-instrumenting-techniques]]Build a 32-bit instrumented application for a 64-bit target system
3895 In order to trace a 32-bit application running on a 64-bit system,
3896 LTTng must use a dedicated 32-bit
3897 <<lttng-consumerd,consumer daemon>>.
3899 The following steps show how to build and install a 32-bit consumer
3900 daemon, which is _not_ part of the default 64-bit LTTng build, how to
3901 build and install the 32-bit LTTng-UST libraries, and how to build and
3902 link an instrumented 32-bit application in that context.
3904 To build a 32-bit instrumented application for a 64-bit target system,
3905 assuming you have a fresh target system with no installed Userspace RCU
3908 . Download, build, and install a 32-bit version of Userspace RCU:
3913 $ cd $(mktemp -d) &&
3914 wget http://lttng.org/files/urcu/userspace-rcu-latest-0.9.tar.bz2 &&
3915 tar -xf userspace-rcu-latest-0.9.tar.bz2 &&
3916 cd userspace-rcu-0.9.* &&
3917 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 &&
3919 sudo make install &&
3924 . Using your distribution's package manager, or from source, install
3925 the following 32-bit versions of the following dependencies of
3926 LTTng-tools and LTTng-UST:
3929 * https://sourceforge.net/projects/libuuid/[libuuid]
3930 * http://directory.fsf.org/wiki/Popt[popt]
3931 * http://www.xmlsoft.org/[libxml2]
3934 . Download, build, and install a 32-bit version of the latest
3935 LTTng-UST{nbsp}{revision}:
3940 $ cd $(mktemp -d) &&
3941 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.8.tar.bz2 &&
3942 tar -xf lttng-ust-latest-2.8.tar.bz2 &&
3943 cd lttng-ust-2.8.* &&
3944 ./configure --libdir=/usr/local/lib32 \
3945 CFLAGS=-m32 CXXFLAGS=-m32 \
3946 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
3948 sudo make install &&
3955 Depending on your distribution,
3956 32-bit libraries could be installed at a different location than
3957 `/usr/lib32`. For example, Debian is known to install
3958 some 32-bit libraries in `/usr/lib/i386-linux-gnu`.
3960 In this case, make sure to set `LDFLAGS` to all the
3961 relevant 32-bit library paths, for example:
3965 $ LDFLAGS='-L/usr/lib/i386-linux-gnu -L/usr/lib32'
3969 . Download the latest LTTng-tools{nbsp}{revision}, build, and install
3970 the 32-bit consumer daemon:
3975 $ cd $(mktemp -d) &&
3976 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.8.tar.bz2 &&
3977 tar -xf lttng-tools-latest-2.8.tar.bz2 &&
3978 cd lttng-tools-2.8.* &&
3979 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 CXXFLAGS=-m32 \
3980 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' \
3981 --disable-bin-lttng --disable-bin-lttng-crash \
3982 --disable-bin-lttng-relayd --disable-bin-lttng-sessiond &&
3984 cd src/bin/lttng-consumerd &&
3985 sudo make install &&
3990 . From your distribution or from source,
3991 <<installing-lttng,install>> the 64-bit versions of
3992 LTTng-UST and Userspace RCU.
3993 . Download, build, and install the 64-bit version of the
3994 latest LTTng-tools{nbsp}{revision}:
3999 $ cd $(mktemp -d) &&
4000 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.8.tar.bz2 &&
4001 tar -xf lttng-tools-latest-2.8.tar.bz2 &&
4002 cd lttng-tools-2.8.* &&
4003 ./configure --with-consumerd32-libdir=/usr/local/lib32 \
4004 --with-consumerd32-bin=/usr/local/lib32/lttng/libexec/lttng-consumerd &&
4006 sudo make install &&
4011 . Pass the following options to man:gcc(1), man:g++(1), or man:clang(1)
4012 when linking your 32-bit application:
4015 -m32 -L/usr/lib32 -L/usr/local/lib32 \
4016 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32
4019 For example, let's rebuild the quick start example in
4020 <<tracing-your-own-user-application,Trace a user application>> as an
4021 instrumented 32-bit application:
4026 $ gcc -m32 -c -I. hello-tp.c
4027 $ gcc -m32 -c hello.c
4028 $ gcc -m32 -o hello hello.o hello-tp.o \
4029 -L/usr/lib32 -L/usr/local/lib32 \
4030 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32 \
4035 No special action is required to execute the 32-bit application and
4036 to trace it: use the command-line man:lttng(1) tool as usual.
4043 man:tracef(3) is a small LTTng-UST API designed for quick,
4044 man:printf(3)-like instrumentation without the burden of
4045 <<tracepoint-provider,creating>> and
4046 <<building-tracepoint-providers-and-user-application,building>>
4047 a tracepoint provider package.
4049 To use `tracef()` in your application:
4051 . In the C or C++ source files where you need to use `tracef()`,
4052 include `<lttng/tracef.h>`:
4057 #include <lttng/tracef.h>
4061 . In the application's source code, use `tracef()` like you would use
4069 tracef("my message: %d (%s)", my_integer, my_string);
4075 . Link your application with `liblttng-ust`:
4080 $ gcc -o app app.c -llttng-ust
4084 To trace the events that `tracef()` calls emit:
4086 * <<enabling-disabling-events,Create an event rule>> which matches the
4087 `lttng_ust_tracef:*` event name:
4092 $ lttng enable-event --userspace 'lttng_ust_tracef:*'
4097 .Limitations of `tracef()`
4099 The `tracef()` utility function was developed to make user space tracing
4100 super simple, albeit with notable disadvantages compared to
4101 <<defining-tracepoints,user-defined tracepoints>>:
4103 * All the emitted events have the same tracepoint provider and
4104 tracepoint names, respectively `lttng_ust_tracef` and `event`.
4105 * There is no static type checking.
4106 * The only event record field you actually get, named `msg`, is a string
4107 potentially containing the values you passed to `tracef()`
4108 using your own format string. This also means that you cannot filter
4109 events with a custom expression at run time because there are no
4111 * Since `tracef()` uses the C standard library's man:vasprintf(3)
4112 function behind the scenes to format the strings at run time, its
4113 expected performance is lower than with user-defined tracepoints,
4114 which do not require a conversion to a string.
4116 Taking this into consideration, `tracef()` is useful for some quick
4117 prototyping and debugging, but you should not consider it for any
4118 permanent and serious applicative instrumentation.
4124 ==== Use `tracelog()`
4126 The man:tracelog(3) API is very similar to <<tracef,`tracef()`>>, with
4127 the difference that it accepts an additional log level parameter.
4129 The goal of `tracelog()` is to ease the migration from logging to
4132 To use `tracelog()` in your application:
4134 . In the C or C++ source files where you need to use `tracelog()`,
4135 include `<lttng/tracelog.h>`:
4140 #include <lttng/tracelog.h>
4144 . In the application's source code, use `tracelog()` like you would use
4145 man:printf(3), except for the first parameter which is the log
4153 tracelog(TRACE_WARNING, "my message: %d (%s)",
4154 my_integer, my_string);
4160 See man:lttng-ust(3) for a list of available log level names.
4162 . Link your application with `liblttng-ust`:
4167 $ gcc -o app app.c -llttng-ust
4171 To trace the events that `tracelog()` calls emit with a log level
4172 _as severe as_ a specific log level:
4174 * <<enabling-disabling-events,Create an event rule>> which matches the
4175 `lttng_ust_tracelog:*` event name and a minimum level
4181 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4182 --loglevel=TRACE_WARNING
4186 To trace the events that `tracelog()` calls emit with a
4187 _specific log level_:
4189 * Create an event rule which matches the `lttng_ust_tracelog:*`
4190 event name and a specific log level:
4195 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4196 --loglevel-only=TRACE_INFO
4201 [[prebuilt-ust-helpers]]
4202 === Prebuilt user space tracing helpers
4204 The LTTng-UST package provides a few helpers in the form or preloadable
4205 shared objects which automatically instrument system functions and
4208 The helper shared objects are normally found in dir:{/usr/lib}. If you
4209 built LTTng-UST <<building-from-source,from source>>, they are probably
4210 located in dir:{/usr/local/lib}.
4212 The installed user space tracing helpers in LTTng-UST{nbsp}{revision}
4215 path:{liblttng-ust-libc-wrapper.so}::
4216 path:{liblttng-ust-pthread-wrapper.so}::
4217 <<liblttng-ust-libc-pthread-wrapper,C{nbsp}standard library
4218 memory and POSIX threads function tracing>>.
4220 path:{liblttng-ust-cyg-profile.so}::
4221 path:{liblttng-ust-cyg-profile-fast.so}::
4222 <<liblttng-ust-cyg-profile,Function entry and exit tracing>>.
4224 path:{liblttng-ust-dl.so}::
4225 <<liblttng-ust-dl,Dynamic linker tracing>>.
4227 To use a user space tracing helper with any user application:
4229 * Preload the helper shared object when you start the application:
4234 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so my-app
4238 You can preload more than one helper:
4243 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so:liblttng-ust-dl.so my-app
4249 [[liblttng-ust-libc-pthread-wrapper]]
4250 ==== Instrument C standard library memory and POSIX threads functions
4252 The path:{liblttng-ust-libc-wrapper.so} and
4253 path:{liblttng-ust-pthread-wrapper.so} helpers
4254 add instrumentation to some C standard library and POSIX
4258 .Functions instrumented by preloading path:{liblttng-ust-libc-wrapper.so}.
4260 |TP provider name |TP name |Instrumented function
4262 .6+|`lttng_ust_libc` |`malloc` |man:malloc(3)
4263 |`calloc` |man:calloc(3)
4264 |`realloc` |man:realloc(3)
4265 |`free` |man:free(3)
4266 |`memalign` |man:memalign(3)
4267 |`posix_memalign` |man:posix_memalign(3)
4271 .Functions instrumented by preloading path:{liblttng-ust-pthread-wrapper.so}.
4273 |TP provider name |TP name |Instrumented function
4275 .4+|`lttng_ust_pthread` |`pthread_mutex_lock_req` |man:pthread_mutex_lock(3p) (request time)
4276 |`pthread_mutex_lock_acq` |man:pthread_mutex_lock(3p) (acquire time)
4277 |`pthread_mutex_trylock` |man:pthread_mutex_trylock(3p)
4278 |`pthread_mutex_unlock` |man:pthread_mutex_unlock(3p)
4281 When you preload the shared object, it replaces the functions listed
4282 in the previous tables by wrappers which contain tracepoints and call
4283 the replaced functions.
4286 [[liblttng-ust-cyg-profile]]
4287 ==== Instrument function entry and exit
4289 The path:{liblttng-ust-cyg-profile*.so} helpers can add instrumentation
4290 to the entry and exit points of functions.
4292 man:gcc(1) and man:clang(1) have an option named
4293 https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html[`-finstrument-functions`]
4294 which generates instrumentation calls for entry and exit to functions.
4295 The LTTng-UST function tracing helpers,
4296 path:{liblttng-ust-cyg-profile.so} and
4297 path:{liblttng-ust-cyg-profile-fast.so}, take advantage of this feature
4298 to add tracepoints to the two generated functions (which contain
4299 `cyg_profile` in their names, hence the helper's name).
4301 To use the LTTng-UST function tracing helper, the source files to
4302 instrument must be built using the `-finstrument-functions` compiler
4305 There are two versions of the LTTng-UST function tracing helper:
4307 * **path:{liblttng-ust-cyg-profile-fast.so}** is a lightweight variant
4308 that you should only use when it can be _guaranteed_ that the
4309 complete event stream is recorded without any lost event record.
4310 Any kind of duplicate information is left out.
4312 Assuming no event record is lost, having only the function addresses on
4313 entry is enough to create a call graph, since an event record always
4314 contains the ID of the CPU that generated it.
4316 You can use a tool like man:addr2line(1) to convert function addresses
4317 back to source file names and line numbers.
4319 * **path:{liblttng-ust-cyg-profile.so}** is a more robust variant
4320 which also works in use cases where event records might get discarded or
4321 not recorded from application startup.
4322 In these cases, the trace analyzer needs more information to be
4323 able to reconstruct the program flow.
4325 See man:lttng-ust-cyg-profile(3) to learn more about the instrumentation
4326 points of this helper.
4328 All the tracepoints that this helper provides have the
4329 log level `TRACE_DEBUG_FUNCTION` (see man:lttng-ust(3)).
4331 TIP: It's sometimes a good idea to limit the number of source files that
4332 you compile with the `-finstrument-functions` option to prevent LTTng
4333 from writing an excessive amount of trace data at run time. When using
4334 man:gcc(1), you can use the
4335 `-finstrument-functions-exclude-function-list` option to avoid
4336 instrument entries and exits of specific function names.
4341 ==== Instrument the dynamic linker
4343 The path:{liblttng-ust-dl.so} helper adds instrumentation to the
4344 man:dlopen(3) and man:dlclose(3) function calls.
4346 See man:lttng-ust-dl(3) to learn more about the instrumentation points
4351 [[java-application]]
4352 === User space Java agent
4354 You can instrument any Java application which uses one of the following
4357 * The https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[**`java.util.logging`**]
4358 (JUL) core logging facilities.
4359 * http://logging.apache.org/log4j/1.2/[**Apache log4j 1.2**], since
4360 LTTng 2.6. Note that Apache Log4j{nbsp}2 is not supported.
4363 .LTTng-UST Java agent imported by a Java application.
4364 image::java-app.png[]
4366 Note that the methods described below are new in LTTng{nbsp}{revision}.
4367 Previous LTTng versions use another technique.
4369 NOTE: We use http://openjdk.java.net/[OpenJDK]{nbsp}8 for development
4370 and https://ci.lttng.org/[continuous integration], thus this version is
4371 directly supported. However, the LTTng-UST Java agent is also tested
4372 with OpenJDK{nbsp}7.
4377 ==== Use the LTTng-UST Java agent for `java.util.logging`
4379 To use the LTTng-UST Java agent in a Java application which uses
4380 `java.util.logging` (JUL):
4382 . In the Java application's source code, import the LTTng-UST
4383 log handler package for `java.util.logging`:
4388 import org.lttng.ust.agent.jul.LttngLogHandler;
4392 . Create an LTTng-UST JUL log handler:
4397 Handler lttngUstLogHandler = new LttngLogHandler();
4401 . Add this handler to the JUL loggers which should emit LTTng events:
4406 Logger myLogger = Logger.getLogger("some-logger");
4408 myLogger.addHandler(lttngUstLogHandler);
4412 . Use `java.util.logging` log statements and configuration as usual.
4413 The loggers with an attached LTTng-UST log handler can emit
4416 . Before exiting the application, remove the LTTng-UST log handler from
4417 the loggers attached to it and call its `close()` method:
4422 myLogger.removeHandler(lttngUstLogHandler);
4423 lttngUstLogHandler.close();
4427 This is not strictly necessary, but it is recommended for a clean
4428 disposal of the handler's resources.
4430 . Include the LTTng-UST Java agent's common and JUL-specific JAR files,
4431 path:{lttng-ust-agent-common.jar} and path:{lttng-ust-agent-jul.jar},
4433 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4434 path] when you build the Java application.
4436 The JAR files are typically located in dir:{/usr/share/java}.
4438 IMPORTANT: The LTTng-UST Java agent must be
4439 <<installing-lttng,installed>> for the logging framework your
4442 .Use the LTTng-UST Java agent for `java.util.logging`.
4447 import java.io.IOException;
4448 import java.util.logging.Handler;
4449 import java.util.logging.Logger;
4450 import org.lttng.ust.agent.jul.LttngLogHandler;
4454 private static final int answer = 42;
4456 public static void main(String[] argv) throws Exception
4459 Logger logger = Logger.getLogger("jello");
4461 // Create an LTTng-UST log handler
4462 Handler lttngUstLogHandler = new LttngLogHandler();
4464 // Add the LTTng-UST log handler to our logger
4465 logger.addHandler(lttngUstLogHandler);
4468 logger.info("some info");
4469 logger.warning("some warning");
4471 logger.finer("finer information; the answer is " + answer);
4473 logger.severe("error!");
4475 // Not mandatory, but cleaner
4476 logger.removeHandler(lttngUstLogHandler);
4477 lttngUstLogHandler.close();
4486 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4489 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4490 <<enabling-disabling-events,create an event rule>> matching the
4491 `jello` JUL logger, and <<basic-tracing-session-control,start tracing>>:
4496 $ lttng enable-event --jul jello
4500 Run the compiled class:
4504 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4507 <<basic-tracing-session-control,Stop tracing>> and inspect the
4517 In the resulting trace, an <<event,event record>> generated by a Java
4518 application using `java.util.logging` is named `lttng_jul:event` and
4519 has the following fields:
4522 Log record's message.
4528 Name of the class in which the log statement was executed.
4531 Name of the method in which the log statement was executed.
4534 Logging time (timestamp in milliseconds).
4537 Log level integer value.
4540 ID of the thread in which the log statement was executed.
4542 You can use the opt:lttng-enable-event(1):--loglevel or
4543 opt:lttng-enable-event(1):--loglevel-only option of the
4544 man:lttng-enable-event(1) command to target a range of JUL log levels
4545 or a specific JUL log level.
4550 ==== Use the LTTng-UST Java agent for Apache log4j
4552 To use the LTTng-UST Java agent in a Java application which uses
4555 . In the Java application's source code, import the LTTng-UST
4556 log appender package for Apache log4j:
4561 import org.lttng.ust.agent.log4j.LttngLogAppender;
4565 . Create an LTTng-UST log4j log appender:
4570 Appender lttngUstLogAppender = new LttngLogAppender();
4574 . Add this appender to the log4j loggers which should emit LTTng events:
4579 Logger myLogger = Logger.getLogger("some-logger");
4581 myLogger.addAppender(lttngUstLogAppender);
4585 . Use Apache log4j log statements and configuration as usual. The
4586 loggers with an attached LTTng-UST log appender can emit LTTng events.
4588 . Before exiting the application, remove the LTTng-UST log appender from
4589 the loggers attached to it and call its `close()` method:
4594 myLogger.removeAppender(lttngUstLogAppender);
4595 lttngUstLogAppender.close();
4599 This is not strictly necessary, but it is recommended for a clean
4600 disposal of the appender's resources.
4602 . Include the LTTng-UST Java agent's common and log4j-specific JAR
4603 files, path:{lttng-ust-agent-common.jar} and
4604 path:{lttng-ust-agent-log4j.jar}, in the
4605 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4606 path] when you build the Java application.
4608 The JAR files are typically located in dir:{/usr/share/java}.
4610 IMPORTANT: The LTTng-UST Java agent must be
4611 <<installing-lttng,installed>> for the logging framework your
4614 .Use the LTTng-UST Java agent for Apache log4j.
4619 import org.apache.log4j.Appender;
4620 import org.apache.log4j.Logger;
4621 import org.lttng.ust.agent.log4j.LttngLogAppender;
4625 private static final int answer = 42;
4627 public static void main(String[] argv) throws Exception
4630 Logger logger = Logger.getLogger("jello");
4632 // Create an LTTng-UST log appender
4633 Appender lttngUstLogAppender = new LttngLogAppender();
4635 // Add the LTTng-UST log appender to our logger
4636 logger.addAppender(lttngUstLogAppender);
4639 logger.info("some info");
4640 logger.warn("some warning");
4642 logger.debug("debug information; the answer is " + answer);
4644 logger.fatal("error!");
4646 // Not mandatory, but cleaner
4647 logger.removeAppender(lttngUstLogAppender);
4648 lttngUstLogAppender.close();
4654 Build this example (`$LOG4JPATH` is the path to the Apache log4j JAR
4659 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH Test.java
4662 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4663 <<enabling-disabling-events,create an event rule>> matching the
4664 `jello` log4j logger, and <<basic-tracing-session-control,start tracing>>:
4669 $ lttng enable-event --log4j jello
4673 Run the compiled class:
4677 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH:. Test
4680 <<basic-tracing-session-control,Stop tracing>> and inspect the
4690 In the resulting trace, an <<event,event record>> generated by a Java
4691 application using log4j is named `lttng_log4j:event` and
4692 has the following fields:
4695 Log record's message.
4701 Name of the class in which the log statement was executed.
4704 Name of the method in which the log statement was executed.
4707 Name of the file in which the executed log statement is located.
4710 Line number at which the log statement was executed.
4716 Log level integer value.
4719 Name of the Java thread in which the log statement was executed.
4721 You can use the opt:lttng-enable-event(1):--loglevel or
4722 opt:lttng-enable-event(1):--loglevel-only option of the
4723 man:lttng-enable-event(1) command to target a range of Apache log4j log levels
4724 or a specific log4j log level.
4728 [[java-application-context]]
4729 ==== Provide application-specific context fields in a Java application
4731 A Java application-specific context field is a piece of state provided
4732 by the application which <<adding-context,you can add>>, using the
4733 man:lttng-add-context(1) command, to each <<event,event record>>
4734 produced by the log statements of this application.
4736 For example, a given object might have a current request ID variable.
4737 You can create a context information retriever for this object and
4738 assign a name to this current request ID. You can then, using the
4739 man:lttng-add-context(1) command, add this context field by name to
4740 the JUL or log4j <<channel,channel>>.
4742 To provide application-specific context fields in a Java application:
4744 . In the Java application's source code, import the LTTng-UST
4745 Java agent context classes and interfaces:
4750 import org.lttng.ust.agent.context.ContextInfoManager;
4751 import org.lttng.ust.agent.context.IContextInfoRetriever;
4755 . Create a context information retriever class, that is, a class which
4756 implements the `IContextInfoRetriever` interface:
4761 class MyContextInfoRetriever implements IContextInfoRetriever
4764 public Object retrieveContextInfo(String key)
4766 if (key.equals("intCtx")) {
4768 } else if (key.equals("strContext")) {
4769 return "context value!";
4778 This `retrieveContextInfo()` method is the only member of the
4779 `IContextInfoRetriever` interface. Its role is to return the current
4780 value of a state by name to create a context field. The names of the
4781 context fields and which state variables they return depends on your
4784 All primitive types and objects are supported as context fields.
4785 When `retrieveContextInfo()` returns an object, the context field
4786 serializer calls its `toString()` method to add a string field to
4787 event records. The method can also return `null`, which means that
4788 no context field is available for the required name.
4790 . Register an instance of your context information retriever class to
4791 the context information manager singleton:
4796 IContextInfoRetriever cir = new MyContextInfoRetriever();
4797 ContextInfoManager cim = ContextInfoManager.getInstance();
4798 cim.registerContextInfoRetriever("retrieverName", cir);
4802 . Before exiting the application, remove your context information
4803 retriever from the context information manager singleton:
4808 ContextInfoManager cim = ContextInfoManager.getInstance();
4809 cim.unregisterContextInfoRetriever("retrieverName");
4813 This is not strictly necessary, but it is recommended for a clean
4814 disposal of some manager's resources.
4816 . Build your Java application with LTTng-UST Java agent support as
4817 usual, following the procedure for either the <<jul,JUL>> or
4818 <<log4j,Apache log4j>> framework.
4821 .Provide application-specific context fields in a Java application.
4826 import java.util.logging.Handler;
4827 import java.util.logging.Logger;
4828 import org.lttng.ust.agent.jul.LttngLogHandler;
4829 import org.lttng.ust.agent.context.ContextInfoManager;
4830 import org.lttng.ust.agent.context.IContextInfoRetriever;
4834 // Our context information retriever class
4835 private static class MyContextInfoRetriever
4836 implements IContextInfoRetriever
4839 public Object retrieveContextInfo(String key) {
4840 if (key.equals("intCtx")) {
4842 } else if (key.equals("strContext")) {
4843 return "context value!";
4850 private static final int answer = 42;
4852 public static void main(String args[]) throws Exception
4854 // Get the context information manager instance
4855 ContextInfoManager cim = ContextInfoManager.getInstance();
4857 // Create and register our context information retriever
4858 IContextInfoRetriever cir = new MyContextInfoRetriever();
4859 cim.registerContextInfoRetriever("myRetriever", cir);
4862 Logger logger = Logger.getLogger("jello");
4864 // Create an LTTng-UST log handler
4865 Handler lttngUstLogHandler = new LttngLogHandler();
4867 // Add the LTTng-UST log handler to our logger
4868 logger.addHandler(lttngUstLogHandler);
4871 logger.info("some info");
4872 logger.warning("some warning");
4874 logger.finer("finer information; the answer is " + answer);
4876 logger.severe("error!");
4878 // Not mandatory, but cleaner
4879 logger.removeHandler(lttngUstLogHandler);
4880 lttngUstLogHandler.close();
4881 cim.unregisterContextInfoRetriever("myRetriever");
4890 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4893 <<creating-destroying-tracing-sessions,Create a tracing session>>
4894 and <<enabling-disabling-events,create an event rule>> matching the
4900 $ lttng enable-event --jul jello
4903 <<adding-context,Add the application-specific context fields>> to the
4908 $ lttng add-context --jul --type='$app.myRetriever:intCtx'
4909 $ lttng add-context --jul --type='$app.myRetriever:strContext'
4912 <<basic-tracing-session-control,Start tracing>>:
4919 Run the compiled class:
4923 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4926 <<basic-tracing-session-control,Stop tracing>> and inspect the
4938 [[python-application]]
4939 === User space Python agent
4941 You can instrument a Python 2 or Python 3 application which uses the
4942 standard https://docs.python.org/3/library/logging.html[`logging`]
4945 Each log statement emits an LTTng event once the
4946 application module imports the
4947 <<lttng-ust-agents,LTTng-UST Python agent>> package.
4950 .A Python application importing the LTTng-UST Python agent.
4951 image::python-app.png[]
4953 To use the LTTng-UST Python agent:
4955 . In the Python application's source code, import the LTTng-UST Python
4965 The LTTng-UST Python agent automatically adds its logging handler to the
4966 root logger at import time.
4968 Any log statement that the application executes before this import does
4969 not emit an LTTng event.
4971 IMPORTANT: The LTTng-UST Python agent must be
4972 <<installing-lttng,installed>>.
4974 . Use log statements and logging configuration as usual.
4975 Since the LTTng-UST Python agent adds a handler to the _root_
4976 logger, you can trace any log statement from any logger.
4978 .Use the LTTng-UST Python agent.
4989 logging.basicConfig()
4990 logger = logging.getLogger('my-logger')
4993 logger.debug('debug message')
4994 logger.info('info message')
4995 logger.warn('warn message')
4996 logger.error('error message')
4997 logger.critical('critical message')
5001 if __name__ == '__main__':
5005 NOTE: `logging.basicConfig()`, which adds to the root logger a basic
5006 logging handler which prints to the standard error stream, is not
5007 strictly required for LTTng-UST tracing to work, but in versions of
5008 Python preceding 3.2, you could see a warning message which indicates
5009 that no handler exists for the logger `my-logger`.
5011 <<creating-destroying-tracing-sessions,Create a tracing session>>,
5012 <<enabling-disabling-events,create an event rule>> matching the
5013 `my-logger` Python logger, and <<basic-tracing-session-control,start
5019 $ lttng enable-event --python my-logger
5023 Run the Python script:
5030 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
5040 In the resulting trace, an <<event,event record>> generated by a Python
5041 application is named `lttng_python:event` and has the following fields:
5044 Logging time (string).
5047 Log record's message.
5053 Name of the function in which the log statement was executed.
5056 Line number at which the log statement was executed.
5059 Log level integer value.
5062 ID of the Python thread in which the log statement was executed.
5065 Name of the Python thread in which the log statement was executed.
5067 You can use the opt:lttng-enable-event(1):--loglevel or
5068 opt:lttng-enable-event(1):--loglevel-only option of the
5069 man:lttng-enable-event(1) command to target a range of Python log levels
5070 or a specific Python log level.
5072 When an application imports the LTTng-UST Python agent, the agent tries
5073 to register to a <<lttng-sessiond,session daemon>>. Note that you must
5074 <<start-sessiond,start the session daemon>> _before_ you run the Python
5075 application. If a session daemon is found, the agent tries to register
5076 to it during 5{nbsp}seconds, after which the application continues
5077 without LTTng tracing support. You can override this timeout value with
5078 the env:LTTNG_UST_PYTHON_REGISTER_TIMEOUT environment variable
5081 If the session daemon stops while a Python application with an imported
5082 LTTng-UST Python agent runs, the agent retries to connect and to
5083 register to a session daemon every 3{nbsp}seconds. You can override this
5084 delay with the env:LTTNG_UST_PYTHON_REGISTER_RETRY_DELAY environment
5089 [[proc-lttng-logger-abi]]
5092 The `lttng-tracer` Linux kernel module, part of
5093 <<lttng-modules,LTTng-modules>>, creates the special LTTng logger file
5094 path:{/proc/lttng-logger} when it's loaded. Any application can write
5095 text data to this file to emit an LTTng event.
5098 .An application writes to the LTTng logger file to emit an LTTng event.
5099 image::lttng-logger.png[]
5101 The LTTng logger is the quickest method--not the most efficient,
5102 however--to add instrumentation to an application. It is designed
5103 mostly to instrument shell scripts:
5107 $ echo "Some message, some $variable" > /proc/lttng-logger
5110 Any event that the LTTng logger emits is named `lttng_logger` and
5111 belongs to the Linux kernel <<domain,tracing domain>>. However, unlike
5112 other instrumentation points in the kernel tracing domain, **any Unix
5113 user** can <<enabling-disabling-events,create an event rule>> which
5114 matches its event name, not only the root user or users in the
5115 <<tracing-group,tracing group>>.
5117 To use the LTTng logger:
5119 * From any application, write text data to the path:{/proc/lttng-logger}
5122 The `msg` field of `lttng_logger` event records contains the
5125 NOTE: The maximum message length of an LTTng logger event is
5126 1024{nbsp}bytes. Writing more than this makes the LTTng logger emit more
5127 than one event to contain the remaining data.
5129 You should not use the LTTng logger to trace a user application which
5130 can be instrumented in a more efficient way, namely:
5132 * <<c-application,C and $$C++$$ applications>>.
5133 * <<java-application,Java applications>>.
5134 * <<python-application,Python applications>>.
5136 .Use the LTTng logger.
5141 echo 'Hello, World!' > /proc/lttng-logger
5143 df --human-readable --print-type / > /proc/lttng-logger
5146 <<creating-destroying-tracing-sessions,Create a tracing session>>,
5147 <<enabling-disabling-events,create an event rule>> matching the
5148 `lttng_logger` Linux kernel tracepoint, and
5149 <<basic-tracing-session-control,start tracing>>:
5154 $ lttng enable-event --kernel lttng_logger
5158 Run the Bash script:
5165 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
5176 [[instrumenting-linux-kernel]]
5177 === LTTng kernel tracepoints
5179 NOTE: This section shows how to _add_ instrumentation points to the
5180 Linux kernel. The kernel's subsystems are already thoroughly
5181 instrumented at strategic places for LTTng when you
5182 <<installing-lttng,install>> the <<lttng-modules,LTTng-modules>>
5186 There are two methods to instrument the Linux kernel:
5188 . <<linux-add-lttng-layer,Add an LTTng layer>> over an existing ftrace
5189 tracepoint which uses the `TRACE_EVENT()` API.
5191 Choose this if you want to instrumentation a Linux kernel tree with an
5192 instrumentation point compatible with ftrace, perf, and SystemTap.
5194 . Use an <<linux-lttng-tracepoint-event,LTTng-only approach>> to
5195 instrument an out-of-tree kernel module.
5197 Choose this if you don't need ftrace, perf, or SystemTap support.
5201 [[linux-add-lttng-layer]]
5202 ==== [[instrumenting-linux-kernel-itself]][[mainline-trace-event]][[lttng-adaptation-layer]]Add an LTTng layer to an existing ftrace tracepoint
5204 This section shows how to add an LTTng layer to existing ftrace
5205 instrumentation using the `TRACE_EVENT()` API.
5207 This section does not document the `TRACE_EVENT()` macro. You can
5208 read the following articles to learn more about this API:
5210 * http://lwn.net/Articles/379903/[Using the TRACE_EVENT() macro (Part 1)]
5211 * http://lwn.net/Articles/381064/[Using the TRACE_EVENT() macro (Part 2)]
5212 * http://lwn.net/Articles/383362/[Using the TRACE_EVENT() macro (Part 3)]
5214 The following procedure assumes that your ftrace tracepoints are
5215 correctly defined in their own header and that they are created in
5216 one source file using the `CREATE_TRACE_POINTS` definition.
5218 To add an LTTng layer over an existing ftrace tracepoint:
5220 . Make sure the following kernel configuration options are
5226 * `CONFIG_HIGH_RES_TIMERS`
5227 * `CONFIG_TRACEPOINTS`
5230 . Build the Linux source tree with your custom ftrace tracepoints.
5231 . Boot the resulting Linux image on your target system.
5233 Confirm that the tracepoints exist by looking for their names in the
5234 dir:{/sys/kernel/debug/tracing/events/subsys} directory, where `subsys`
5235 is your subsystem's name.
5237 . Get a copy of the latest LTTng-modules{nbsp}{revision}:
5242 $ cd $(mktemp -d) &&
5243 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.8.tar.bz2 &&
5244 tar -xf lttng-modules-latest-2.8.tar.bz2 &&
5245 cd lttng-modules-2.8.*
5249 . In dir:{instrumentation/events/lttng-module}, relative to the root
5250 of the LTTng-modules source tree, create a header file named
5251 +__subsys__.h+ for your custom subsystem +__subsys__+ and write your
5252 LTTng-modules tracepoint definitions using the LTTng-modules
5255 Start with this template:
5259 .path:{instrumentation/events/lttng-module/my_subsys.h}
5262 #define TRACE_SYSTEM my_subsys
5264 #if !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ)
5265 #define _LTTNG_MY_SUBSYS_H
5267 #include "../../../probes/lttng-tracepoint-event.h"
5268 #include <linux/tracepoint.h>
5270 LTTNG_TRACEPOINT_EVENT(
5272 * Format is identical to TRACE_EVENT()'s version for the three
5273 * following macro parameters:
5276 TP_PROTO(int my_int, const char *my_string),
5277 TP_ARGS(my_int, my_string),
5279 /* LTTng-modules specific macros */
5281 ctf_integer(int, my_int_field, my_int)
5282 ctf_string(my_bar_field, my_bar)
5286 #endif /* !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ) */
5288 #include "../../../probes/define_trace.h"
5292 The entries in the `TP_FIELDS()` section are the list of fields for the
5293 LTTng tracepoint. This is similar to the `TP_STRUCT__entry()` part of
5294 ftrace's `TRACE_EVENT()` macro.
5296 See <<lttng-modules-tp-fields,Tracepoint fields macros>> for a
5297 complete description of the available `ctf_*()` macros.
5299 . Create the LTTng-modules probe's kernel module C source file,
5300 +probes/lttng-probe-__subsys__.c+, where +__subsys__+ is your
5305 .path:{probes/lttng-probe-my-subsys.c}
5307 #include <linux/module.h>
5308 #include "../lttng-tracer.h"
5311 * Build-time verification of mismatch between mainline
5312 * TRACE_EVENT() arguments and the LTTng-modules adaptation
5313 * layer LTTNG_TRACEPOINT_EVENT() arguments.
5315 #include <trace/events/my_subsys.h>
5317 /* Create LTTng tracepoint probes */
5318 #define LTTNG_PACKAGE_BUILD
5319 #define CREATE_TRACE_POINTS
5320 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
5322 #include "../instrumentation/events/lttng-module/my_subsys.h"
5324 MODULE_LICENSE("GPL and additional rights");
5325 MODULE_AUTHOR("Your name <your-email>");
5326 MODULE_DESCRIPTION("LTTng my_subsys probes");
5327 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
5328 __stringify(LTTNG_MODULES_MINOR_VERSION) "."
5329 __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
5330 LTTNG_MODULES_EXTRAVERSION);
5334 . Edit path:{probes/Makefile} and add your new kernel module object
5335 next to the existing ones:
5339 .path:{probes/Makefile}
5343 obj-m += lttng-probe-module.o
5344 obj-m += lttng-probe-power.o
5346 obj-m += lttng-probe-my-subsys.o
5352 . Build and install the LTTng kernel modules:
5357 $ make KERNELDIR=/path/to/linux
5358 # make modules_install && depmod -a
5362 Replace `/path/to/linux` with the path to the Linux source tree where
5363 you defined and used tracepoints with ftrace's `TRACE_EVENT()` macro.
5365 Note that you can also use the
5366 <<lttng-tracepoint-event-code,`LTTNG_TRACEPOINT_EVENT_CODE()` macro>>
5367 instead of `LTTNG_TRACEPOINT_EVENT()` to use custom local variables and
5368 C code that need to be executed before the event fields are recorded.
5370 The best way to learn how to use the previous LTTng-modules macros is to
5371 inspect the existing LTTng-modules tracepoint definitions in the
5372 dir:{instrumentation/events/lttng-module} header files. Compare them
5373 with the Linux kernel mainline versions in the
5374 dir:{include/trace/events} directory of the Linux source tree.
5378 [[lttng-tracepoint-event-code]]
5379 ===== Use custom C code to access the data for tracepoint fields
5381 Although we recommended to always use the
5382 <<lttng-adaptation-layer,`LTTNG_TRACEPOINT_EVENT()`>> macro to describe
5383 the arguments and fields of an LTTng-modules tracepoint when possible,
5384 sometimes you need a more complex process to access the data that the
5385 tracer records as event record fields. In other words, you need local
5386 variables and multiple C{nbsp}statements instead of simple
5387 argument-based expressions that you pass to the
5388 <<lttng-modules-tp-fields,`ctf_*()` macros of `TP_FIELDS()`>>.
5390 You can use the `LTTNG_TRACEPOINT_EVENT_CODE()` macro instead of
5391 `LTTNG_TRACEPOINT_EVENT()` to declare custom local variables and define
5392 a block of C{nbsp}code to be executed before LTTng records the fields.
5393 The structure of this macro is:
5396 .`LTTNG_TRACEPOINT_EVENT_CODE()` macro syntax.
5398 LTTNG_TRACEPOINT_EVENT_CODE(
5400 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5401 * version for the following three macro parameters:
5404 TP_PROTO(int my_int, const char *my_string),
5405 TP_ARGS(my_int, my_string),
5407 /* Declarations of custom local variables */
5410 unsigned long b = 0;
5411 const char *name = "(undefined)";
5412 struct my_struct *my_struct;
5416 * Custom code which uses both tracepoint arguments
5417 * (in TP_ARGS()) and local variables (in TP_locvar()).
5419 * Local variables are actually members of a structure pointed
5420 * to by the special variable tp_locvar.
5424 tp_locvar->a = my_int + 17;
5425 tp_locvar->my_struct = get_my_struct_at(tp_locvar->a);
5426 tp_locvar->b = my_struct_compute_b(tp_locvar->my_struct);
5427 tp_locvar->name = my_struct_get_name(tp_locvar->my_struct);
5428 put_my_struct(tp_locvar->my_struct);
5437 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5438 * version for this, except that tp_locvar members can be
5439 * used in the argument expression parameters of
5440 * the ctf_*() macros.
5443 ctf_integer(unsigned long, my_struct_b, tp_locvar->b)
5444 ctf_integer(int, my_struct_a, tp_locvar->a)
5445 ctf_string(my_string_field, my_string)
5446 ctf_string(my_struct_name, tp_locvar->name)
5451 IMPORTANT: The C code defined in `TP_code()` must not have any side
5452 effects when executed. In particular, the code must not allocate
5453 memory or get resources without deallocating this memory or putting
5454 those resources afterwards.
5457 [[instrumenting-linux-kernel-tracing]]
5458 ==== Load and unload a custom probe kernel module
5460 You must load a <<lttng-adaptation-layer,created LTTng-modules probe
5461 kernel module>> in the kernel before it can emit LTTng events.
5463 To load the default probe kernel modules and a custom probe kernel
5466 * Use the opt:lttng-sessiond(8):--extra-kmod-probes option to give extra
5467 probe modules to load when starting a root <<lttng-sessiond,session
5471 .Load the `my_subsys`, `usb`, and the default probe modules.
5475 # lttng-sessiond --extra-kmod-probes=my_subsys,usb
5480 You only need to pass the subsystem name, not the whole kernel module
5483 To load _only_ a given custom probe kernel module:
5485 * Use the opt:lttng-sessiond(8):--kmod-probes option to give the probe
5486 modules to load when starting a root session daemon:
5489 .Load only the `my_subsys` and `usb` probe modules.
5493 # lttng-sessiond --kmod-probes=my_subsys,usb
5498 To confirm that a probe module is loaded:
5505 $ lsmod | grep lttng_probe_usb
5509 To unload the loaded probe modules:
5511 * Kill the session daemon with `SIGTERM`:
5516 # pkill lttng-sessiond
5520 You can also use man:modprobe(8)'s `--remove` option if the session
5521 daemon terminates abnormally.
5524 [[controlling-tracing]]
5527 Once an application or a Linux kernel is
5528 <<instrumenting,instrumented>> for LTTng tracing,
5531 This section is divided in topics on how to use the various
5532 <<plumbing,components of LTTng>>, in particular the <<lttng-cli,cmd:lttng
5533 command-line tool>>, to _control_ the LTTng daemons and tracers.
5535 NOTE: In the following subsections, we refer to an man:lttng(1) command
5536 using its man page name. For example, instead of _Run the `create`
5537 command to..._, we use _Run the man:lttng-create(1) command to..._.
5541 === Start a session daemon
5543 In some situations, you need to run a <<lttng-sessiond,session daemon>>
5544 (man:lttng-sessiond(8)) _before_ you can use the man:lttng(1)
5547 You will see the following error when you run a command while no session
5551 Error: No session daemon is available
5554 The only command that automatically runs a session daemon is
5555 man:lttng-create(1), which you use to
5556 <<creating-destroying-tracing-sessions,create a tracing session>>. While
5557 this is most of the time the first operation that you do, sometimes it's
5558 not. Some examples are:
5560 * <<list-instrumentation-points,List the available instrumentation points>>.
5561 * <<saving-loading-tracing-session,Load a tracing session configuration>>.
5563 [[tracing-group]] Each Unix user must have its own running session
5564 daemon to trace user applications. The session daemon that the root user
5565 starts is the only one allowed to control the LTTng kernel tracer. Users
5566 that are part of the _tracing group_ can control the root session
5567 daemon. The default tracing group name is `tracing`; you can set it to
5568 something else with the opt:lttng-sessiond(8):--group option when you
5569 start the root session daemon.
5571 To start a user session daemon:
5573 * Run man:lttng-sessiond(8):
5578 $ lttng-sessiond --daemonize
5582 To start the root session daemon:
5584 * Run man:lttng-sessiond(8) as the root user:
5589 # lttng-sessiond --daemonize
5593 In both cases, remove the opt:lttng-sessiond(8):--daemonize option to
5594 start the session daemon in foreground.
5596 To stop a session daemon, use man:kill(1) on its process ID (standard
5599 Note that some Linux distributions could manage the LTTng session daemon
5600 as a service. In this case, you should use the service manager to
5601 start, restart, and stop session daemons.
5604 [[creating-destroying-tracing-sessions]]
5605 === Create and destroy a tracing session
5607 Almost all the LTTng control operations happen in the scope of
5608 a <<tracing-session,tracing session>>, which is the dialogue between the
5609 <<lttng-sessiond,session daemon>> and you.
5611 To create a tracing session with a generated name:
5613 * Use the man:lttng-create(1) command:
5622 The created tracing session's name is `auto` followed by the
5625 To create a tracing session with a specific name:
5627 * Use the optional argument of the man:lttng-create(1) command:
5632 $ lttng create my-session
5636 Replace `my-session` with the specific tracing session name.
5638 LTTng appends the creation date to the created tracing session's name.
5640 LTTng writes the traces of a tracing session in
5641 +$LTTNG_HOME/lttng-trace/__name__+ by default, where +__name__+ is the
5642 name of the tracing session. Note that the env:LTTNG_HOME environment
5643 variable defaults to `$HOME` if not set.
5645 To output LTTng traces to a non-default location:
5647 * Use the opt:lttng-create(1):--output option of the man:lttng-create(1) command:
5652 $ lttng create my-session --output=/tmp/some-directory
5656 You may create as many tracing sessions as you wish.
5658 To list all the existing tracing sessions for your Unix user:
5660 * Use the man:lttng-list(1) command:
5669 When you create a tracing session, it is set as the _current tracing
5670 session_. The following man:lttng(1) commands operate on the current
5671 tracing session when you don't specify one:
5673 [role="list-3-cols"]
5689 To change the current tracing session:
5691 * Use the man:lttng-set-session(1) command:
5696 $ lttng set-session new-session
5700 Replace `new-session` by the name of the new current tracing session.
5702 When you are done tracing in a given tracing session, you can destroy
5703 it. This operation frees the resources taken by the tracing session
5704 to destroy; it does not destroy the trace data that LTTng wrote for
5705 this tracing session.
5707 To destroy the current tracing session:
5709 * Use the man:lttng-destroy(1) command:
5719 [[list-instrumentation-points]]
5720 === List the available instrumentation points
5722 The <<lttng-sessiond,session daemon>> can query the running instrumented
5723 user applications and the Linux kernel to get a list of available
5724 instrumentation points. For the Linux kernel <<domain,tracing domain>>,
5725 they are tracepoints and system calls. For the user space tracing
5726 domain, they are tracepoints. For the other tracing domains, they are
5729 To list the available instrumentation points:
5731 * Use the man:lttng-list(1) command with the requested tracing domain's
5735 * opt:lttng-list(1):--kernel: Linux kernel tracepoints (your Unix user
5736 must be a root user, or it must be a member of the
5737 <<tracing-group,tracing group>>).
5738 * opt:lttng-list(1):--kernel with opt:lttng-list(1):--syscall: Linux
5739 kernel system calls (your Unix user must be a root user, or it must be
5740 a member of the tracing group).
5741 * opt:lttng-list(1):--userspace: user space tracepoints.
5742 * opt:lttng-list(1):--jul: `java.util.logging` loggers.
5743 * opt:lttng-list(1):--log4j: Apache log4j loggers.
5744 * opt:lttng-list(1):--python: Python loggers.
5747 .List the available user space tracepoints.
5751 $ lttng list --userspace
5755 .List the available Linux kernel system call tracepoints.
5759 $ lttng list --kernel --syscall
5764 [[enabling-disabling-events]]
5765 === Create and enable an event rule
5767 Once you <<creating-destroying-tracing-sessions,create a tracing
5768 session>>, you can create <<event,event rules>> with the
5769 man:lttng-enable-event(1) command.
5771 You specify each condition with a command-line option. The available
5772 condition options are shown in the following table.
5774 [role="growable",cols="asciidoc,asciidoc,default"]
5775 .Condition command-line options for the man:lttng-enable-event(1) command.
5777 |Option |Description |Applicable tracing domains
5783 . +--probe=__ADDR__+
5784 . +--function=__ADDR__+
5787 Instead of using the default _tracepoint_ instrumentation type, use:
5789 . A Linux system call.
5790 . A Linux https://lwn.net/Articles/132196/[KProbe] (symbol or address).
5791 . The entry and return points of a Linux function (symbol or address).
5795 |First positional argument.
5798 Tracepoint or system call name. In the case of a Linux KProbe or
5799 function, this is a custom name given to the event rule. With the
5800 JUL, log4j, and Python domains, this is a logger name.
5802 With a tracepoint, logger, or system call name, the last character
5803 can be `*` to match anything that remains.
5810 . +--loglevel=__LEVEL__+
5811 . +--loglevel-only=__LEVEL__+
5814 . Match only tracepoints or log statements with a logging level at
5815 least as severe as +__LEVEL__+.
5816 . Match only tracepoints or log statements with a logging level
5817 equal to +__LEVEL__+.
5819 See man:lttng-enable-event(1) for the list of available logging level
5822 |User space, JUL, log4j, and Python.
5824 |+--exclude=__EXCLUSIONS__+
5827 When you use a `*` character at the end of the tracepoint or logger
5828 name (first positional argument), exclude the specific names in the
5829 comma-delimited list +__EXCLUSIONS__+.
5832 User space, JUL, log4j, and Python.
5834 |+--filter=__EXPR__+
5837 Match only events which satisfy the expression +__EXPR__+.
5839 See man:lttng-enable-event(1) to learn more about the syntax of a
5846 You attach an event rule to a <<channel,channel>> on creation. If you do
5847 not specify the channel with the opt:lttng-enable-event(1):--channel
5848 option, and if the event rule to create is the first in its
5849 <<domain,tracing domain>> for a given tracing session, then LTTng
5850 creates a _default channel_ for you. This default channel is reused in
5851 subsequent invocations of the man:lttng-enable-event(1) command for the
5852 same tracing domain.
5854 An event rule is always enabled at creation time.
5856 The following examples show how you can combine the previous
5857 command-line options to create simple to more complex event rules.
5859 .Create an event rule targetting a Linux kernel tracepoint (default channel).
5863 $ lttng enable-event --kernel sched_switch
5867 .Create an event rule matching four Linux kernel system calls (default channel).
5871 $ lttng enable-event --kernel --syscall open,write,read,close
5875 .Create event rules matching tracepoints with filter expressions (default channel).
5879 $ lttng enable-event --kernel sched_switch --filter='prev_comm == "bash"'
5884 $ lttng enable-event --kernel --all \
5885 --filter='$ctx.tid == 1988 || $ctx.tid == 1534'
5890 $ lttng enable-event --jul my_logger \
5891 --filter='$app.retriever:cur_msg_id > 3'
5894 IMPORTANT: Make sure to always quote the filter string when you
5895 use man:lttng(1) from a shell.
5898 .Create an event rule matching any user space tracepoint of a given tracepoint provider with a log level range (default channel).
5902 $ lttng enable-event --userspace my_app:'*' --loglevel=TRACE_INFO
5905 IMPORTANT: Make sure to always quote the wildcard character when you
5906 use man:lttng(1) from a shell.
5909 .Create an event rule matching multiple Python loggers with a wildcard and with exclusions (default channel).
5913 $ lttng enable-event --python my-app.'*' \
5914 --exclude='my-app.module,my-app.hello'
5918 .Create an event rule matching any Apache log4j logger with a specific log level (default channel).
5922 $ lttng enable-event --log4j --all --loglevel-only=LOG4J_WARN
5926 .Create an event rule attached to a specific channel matching a specific user space tracepoint provider and tracepoint.
5930 $ lttng enable-event --userspace my_app:my_tracepoint --channel=my-channel
5934 The event rules of a given channel form a whitelist: as soon as an
5935 emitted event passes one of them, LTTng can record the event. For
5936 example, an event named `my_app:my_tracepoint` emitted from a user space
5937 tracepoint with a `TRACE_ERROR` log level passes both of the following
5942 $ lttng enable-event --userspace my_app:my_tracepoint
5943 $ lttng enable-event --userspace my_app:my_tracepoint \
5944 --loglevel=TRACE_INFO
5947 The second event rule is redundant: the first one includes
5951 [[disable-event-rule]]
5952 === Disable an event rule
5954 To disable an event rule that you <<enabling-disabling-events,created>>
5955 previously, use the man:lttng-disable-event(1) command. This command
5956 disables _all_ the event rules (of a given tracing domain and channel)
5957 which match an instrumentation point. The other conditions are not
5958 supported as of LTTng{nbsp}{revision}.
5960 The LTTng tracer does not record an emitted event which passes
5961 a _disabled_ event rule.
5963 .Disable an event rule matching a Python logger (default channel).
5967 $ lttng disable-event --python my-logger
5971 .Disable an event rule matching all `java.util.logging` loggers (default channel).
5975 $ lttng disable-event --jul '*'
5979 .Disable _all_ the event rules of the default channel.
5981 The opt:lttng-disable-event(1):--all-events option is not, like the
5982 opt:lttng-enable-event(1):--all option of man:lttng-enable-event(1), the
5983 equivalent of the event name `*` (wildcard): it disables _all_ the event
5984 rules of a given channel.
5988 $ lttng disable-event --jul --all-events
5992 NOTE: You cannot delete an event rule once you create it.
5996 === Get the status of a tracing session
5998 To get the status of the current tracing session, that is, its
5999 parameters, its channels, event rules, and their attributes:
6001 * Use the man:lttng-status(1) command:
6011 To get the status of any tracing session:
6013 * Use the man:lttng-list(1) command with the tracing session's name:
6018 $ lttng list my-session
6022 Replace `my-session` with the desired tracing session's name.
6025 [[basic-tracing-session-control]]
6026 === Start and stop a tracing session
6028 Once you <<creating-destroying-tracing-sessions,create a tracing
6030 <<enabling-disabling-events,create one or more event rules>>,
6031 you can start and stop the tracers for this tracing session.
6033 To start tracing in the current tracing session:
6035 * Use the man:lttng-start(1) command:
6044 LTTng is very flexible: you can launch user applications before
6045 or after the you start the tracers. The tracers only record the events
6046 if they pass enabled event rules and if they occur while the tracers are
6049 To stop tracing in the current tracing session:
6051 * Use the man:lttng-stop(1) command:
6060 If there were <<channel-overwrite-mode-vs-discard-mode,lost event
6061 records>> or lost sub-buffers since the last time you ran
6062 man:lttng-start(1), warnings are printed when you run the
6063 man:lttng-stop(1) command.
6066 [[enabling-disabling-channels]]
6067 === Create a channel
6069 Once you create a tracing session, you can create a <<channel,channel>>
6070 with the man:lttng-enable-channel(1) command.
6072 Note that LTTng automatically creates a default channel when, for a
6073 given <<domain,tracing domain>>, no channels exist and you
6074 <<enabling-disabling-events,create>> the first event rule. This default
6075 channel is named `channel0` and its attributes are set to reasonable
6076 values. Therefore, you only need to create a channel when you need
6077 non-default attributes.
6079 You specify each non-default channel attribute with a command-line
6080 option when you use the man:lttng-enable-channel(1) command. The
6081 available command-line options are:
6083 [role="growable",cols="asciidoc,asciidoc"]
6084 .Command-line options for the man:lttng-enable-channel(1) command.
6086 |Option |Description
6092 <<channel-overwrite-mode-vs-discard-mode,event loss mode>> instead of
6093 the default _discard_ mode.
6095 |`--buffers-pid` (user space tracing domain only)
6098 Use the per-process <<channel-buffering-schemes,buffering scheme>>
6099 instead of the default per-user buffering scheme.
6101 |+--subbuf-size=__SIZE__+
6104 Allocate sub-buffers of +__SIZE__+ bytes (power of two), for each CPU,
6105 either for each Unix user (default), or for each instrumented process.
6107 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
6109 |+--num-subbuf=__COUNT__+
6112 Allocate +__COUNT__+ sub-buffers (power of two), for each CPU, either
6113 for each Unix user (default), or for each instrumented process.
6115 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
6117 |+--tracefile-size=__SIZE__+
6120 Set the maximum size of each trace file that this channel writes within
6121 a stream to +__SIZE__+ bytes instead of no maximum.
6123 See <<tracefile-rotation,Trace file count and size>>.
6125 |+--tracefile-count=__COUNT__+
6128 Limit the number of trace files that this channel creates to
6129 +__COUNT__+ channels instead of no limit.
6131 See <<tracefile-rotation,Trace file count and size>>.
6133 |+--switch-timer=__PERIODUS__+
6136 Set the <<channel-switch-timer,switch timer period>>
6137 to +__PERIODUS__+{nbsp}µs.
6139 |+--read-timer=__PERIODUS__+
6142 Set the <<channel-read-timer,read timer period>>
6143 to +__PERIODUS__+{nbsp}µs.
6145 |+--output=__TYPE__+ (Linux kernel tracing domain only)
6148 Set the channel's output type to +__TYPE__+, either `mmap` or `splice`.
6152 You can only create a channel in the Linux kernel and user space
6153 <<domain,tracing domains>>: other tracing domains have their own channel
6154 created on the fly when <<enabling-disabling-events,creating event
6159 Because of a current LTTng limitation, you must create all channels
6160 _before_ you <<basic-tracing-session-control,start tracing>> in a given
6161 tracing session, that is, before the first time you run
6164 Since LTTng automatically creates a default channel when you use the
6165 man:lttng-enable-event(1) command with a specific tracing domain, you
6166 cannot, for example, create a Linux kernel event rule, start tracing,
6167 and then create a user space event rule, because no user space channel
6168 exists yet and it's too late to create one.
6170 For this reason, make sure to configure your channels properly
6171 before starting the tracers for the first time!
6174 The following examples show how you can combine the previous
6175 command-line options to create simple to more complex channels.
6177 .Create a Linux kernel channel with default attributes.
6181 $ lttng enable-channel --kernel my-channel
6185 .Create a user space channel with 4 sub-buffers or 1{nbsp}MiB each, per CPU, per instrumented process.
6189 $ lttng enable-channel --userspace --num-subbuf=4 --subbuf-size=1M \
6190 --buffers-pid my-channel
6194 .Create a Linux kernel channel which rotates 8 trace files of 4{nbsp}MiB each for each stream
6198 $ lttng enable-channel --kernel --tracefile-count=8 \
6199 --tracefile-size=4194304 my-channel
6203 .Create a user space channel in overwrite (or _flight recorder_) mode.
6207 $ lttng enable-channel --userspace --overwrite my-channel
6211 You can <<enabling-disabling-events,create>> the same event rule in
6212 two different channels:
6216 $ lttng enable-event --userspace --channel=my-channel app:tp
6217 $ lttng enable-event --userspace --channel=other-channel app:tp
6220 If both channels are enabled, when a tracepoint named `app:tp` is
6221 reached, LTTng records two events, one for each channel.
6225 === Disable a channel
6227 To disable a specific channel that you <<enabling-disabling-channels,created>>
6228 previously, use the man:lttng-disable-channel(1) command.
6230 .Disable a specific Linux kernel channel.
6234 $ lttng disable-channel --kernel my-channel
6238 The state of a channel precedes the individual states of event rules
6239 attached to it: event rules which belong to a disabled channel, even if
6240 they are enabled, are also considered disabled.
6244 === Add context fields to a channel
6246 Event record fields in trace files provide important information about
6247 events that occured previously, but sometimes some external context may
6248 help you solve a problem faster. Examples of context fields are:
6250 * The **process ID**, **thread ID**, **process name**, and
6251 **process priority** of the thread in which the event occurs.
6252 * The **hostname** of the system on which the event occurs.
6253 * The current values of many possible **performance counters** using
6255 ** CPU cycles, stalled cycles, idle cycles, and the other cycle types.
6257 ** Branch instructions, misses, and loads.
6259 * Any context defined at the application level (supported for the
6260 JUL and log4j <<domain,tracing domains>>).
6262 To get the full list of available context fields, see
6263 `lttng add-context --list`. Some context fields are reserved for a
6264 specific <<domain,tracing domain>> (Linux kernel or user space).
6266 You add context fields to <<channel,channels>>. All the events
6267 that a channel with added context fields records contain those fields.
6269 To add context fields to one or all the channels of a given tracing
6272 * Use the man:lttng-add-context(1) command.
6274 .Add context fields to all the channels of the current tracing session.
6276 The following command line adds the virtual process identifier and
6277 the per-thread CPU cycles count fields to all the user space channels
6278 of the current tracing session.
6282 $ lttng add-context --userspace --type=vpid --type=perf:thread:cpu-cycles
6286 .Add a context field to a specific channel.
6288 The following command line adds the thread identifier context field
6289 to the Linux kernel channel named `my-channel` in the current
6294 $ lttng add-context --kernel --channel=my-channel --type=tid
6298 .Add an application-specific context field to a specific channel.
6300 The following command line adds the `cur_msg_id` context field of the
6301 `retriever` context retriever for all the instrumented
6302 <<java-application,Java applications>> recording <<event,event records>>
6303 in the channel named `my-channel`:
6307 $ lttng add-context --kernel --channel=my-channel \
6308 --type='$app:retriever:cur_msg_id'
6311 IMPORTANT: Make sure to always quote the `$` character when you
6312 use man:lttng-add-context(1) from a shell.
6315 NOTE: You cannot remove context fields from a channel once you add it.
6320 === Track process IDs
6322 It's often useful to allow only specific process IDs (PIDs) to emit
6323 events. For example, you may wish to record all the system calls made by
6324 a given process (à la http://linux.die.net/man/1/strace[strace]).
6326 The man:lttng-track(1) and man:lttng-untrack(1) commands serve this
6327 purpose. Both commands operate on a whitelist of process IDs. You _add_
6328 entries to this whitelist with the man:lttng-track(1) command and remove
6329 entries with the man:lttng-untrack(1) command. Any process which has one
6330 of the PIDs in the whitelist is allowed to emit LTTng events which pass
6331 an enabled <<event,event rule>>.
6333 NOTE: The PID tracker tracks the _numeric process IDs_. Should a
6334 process with a given tracked ID exit and another process be given this
6335 ID, then the latter would also be allowed to emit events.
6337 .Track and untrack process IDs.
6339 For the sake of the following example, assume the target system has 16
6343 <<creating-destroying-tracing-sessions,create a tracing session>>,
6344 the whitelist contains all the possible PIDs:
6347 .All PIDs are tracked.
6348 image::track-all.png[]
6350 When the whitelist is full and you use the man:lttng-track(1) command to
6351 specify some PIDs to track, LTTng first clears the whitelist, then it
6352 tracks the specific PIDs. After:
6356 $ lttng track --pid=3,4,7,10,13
6362 .PIDs 3, 4, 7, 10, and 13 are tracked.
6363 image::track-3-4-7-10-13.png[]
6365 You can add more PIDs to the whitelist afterwards:
6369 $ lttng track --pid=1,15,16
6375 .PIDs 1, 15, and 16 are added to the whitelist.
6376 image::track-1-3-4-7-10-13-15-16.png[]
6378 The man:lttng-untrack(1) command removes entries from the PID tracker's
6379 whitelist. Given the previous example, the following command:
6383 $ lttng untrack --pid=3,7,10,13
6386 leads to this whitelist:
6389 .PIDs 3, 7, 10, and 13 are removed from the whitelist.
6390 image::track-1-4-15-16.png[]
6392 LTTng can track all possible PIDs again using the opt:track(1):--all
6397 $ lttng track --pid --all
6400 The result is, again:
6403 .All PIDs are tracked.
6404 image::track-all.png[]
6407 .Track only specific PIDs
6409 A very typical use case with PID tracking is to start with an empty
6410 whitelist, then <<basic-tracing-session-control,start the tracers>>, and
6411 then add PIDs manually while tracers are active. You can accomplish this
6412 by using the opt:lttng-untrack(1):--all option of the
6413 man:lttng-untrack(1) command to clear the whitelist after you
6414 <<creating-destroying-tracing-sessions,create a tracing session>>:
6418 $ lttng untrack --pid --all
6424 .No PIDs are tracked.
6425 image::untrack-all.png[]
6427 If you trace with this whitelist configuration, the tracer records no
6428 events for this <<domain,tracing domain>> because no processes are
6429 tracked. You can use the man:lttng-track(1) command as usual to track
6430 specific PIDs, for example:
6434 $ lttng track --pid=6,11
6440 .PIDs 6 and 11 are tracked.
6441 image::track-6-11.png[]
6446 [[saving-loading-tracing-session]]
6447 === Save and load tracing session configurations
6449 Configuring a <<tracing-session,tracing session>> can be long. Some of
6450 the tasks involved are:
6452 * <<enabling-disabling-channels,Create channels>> with
6453 specific attributes.
6454 * <<adding-context,Add context fields>> to specific channels.
6455 * <<enabling-disabling-events,Create event rules>> with specific log
6456 level and filter conditions.
6458 If you use LTTng to solve real world problems, chances are you have to
6459 record events using the same tracing session setup over and over,
6460 modifying a few variables each time in your instrumented program
6461 or environment. To avoid constant tracing session reconfiguration,
6462 the man:lttng(1) command-line tool can save and load tracing session
6463 configurations to/from XML files.
6465 To save a given tracing session configuration:
6467 * Use the man:lttng-save(1) command:
6472 $ lttng save my-session
6476 Replace `my-session` with the name of the tracing session to save.
6478 LTTng saves tracing session configurations to
6479 dir:{$LTTNG_HOME/.lttng/sessions} by default. Note that the
6480 env:LTTNG_HOME environment variable defaults to `$HOME` if not set. Use
6481 the opt:lttng-save(1):--output-path option to change this destination
6484 LTTng saves all configuration parameters, for example:
6486 * The tracing session name.
6487 * The trace data output path.
6488 * The channels with their state and all their attributes.
6489 * The context fields you added to channels.
6490 * The event rules with their state, log level and filter conditions.
6492 To load a tracing session:
6494 * Use the man:lttng-load(1) command:
6499 $ lttng load my-session
6503 Replace `my-session` with the name of the tracing session to load.
6505 When LTTng loads a configuration, it restores your saved tracing session
6506 as if you just configured it manually.
6508 See man:lttng(1) for the complete list of command-line options. You
6509 can also save and load all many sessions at a time, and decide in which
6510 directory to output the XML files.
6513 [[sending-trace-data-over-the-network]]
6514 === Send trace data over the network
6516 LTTng can send the recorded trace data to a remote system over the
6517 network instead of writing it to the local file system.
6519 To send the trace data over the network:
6521 . On the _remote_ system (which can also be the target system),
6522 start an LTTng <<lttng-relayd,relay daemon>> (man:lttng-relayd(8)):
6531 . On the _target_ system, create a tracing session configured to
6532 send trace data over the network:
6537 $ lttng create my-session --set-url=net://remote-system
6541 Replace `remote-system` by the host name or IP address of the
6542 remote system. See man:lttng-create(1) for the exact URL format.
6544 . On the target system, use the man:lttng(1) command-line tool as usual.
6545 When tracing is active, the target's consumer daemon sends sub-buffers
6546 to the relay daemon running on the remote system instead of flushing
6547 them to the local file system. The relay daemon writes the received
6548 packets to the local file system.
6550 The relay daemon writes trace files to
6551 +$LTTNG_HOME/lttng-traces/__hostname__/__session__+ by default, where
6552 +__hostname__+ is the host name of the target system and +__session__+
6553 is the tracing session name. Note that the env:LTTNG_HOME environment
6554 variable defaults to `$HOME` if not set. Use the
6555 opt:lttng-relayd(8):--output option of man:lttng-relayd(8) to write
6556 trace files to another base directory.
6561 === View events as LTTng emits them (noch:{LTTng} live)
6563 LTTng live is a network protocol implemented by the <<lttng-relayd,relay
6564 daemon>> (man:lttng-relayd(8)) to allow compatible trace viewers to
6565 display events as LTTng emits them on the target system while tracing is
6568 The relay daemon creates a _tee_: it forwards the trace data to both
6569 the local file system and to connected live viewers:
6572 .The relay daemon creates a _tee_, forwarding the trace data to both trace files and a connected live viewer.
6577 . On the _target system_, create a <<tracing-session,tracing session>>
6583 $ lttng create my-session --live
6587 This spawns a local relay daemon.
6589 . Start the live viewer and configure it to connect to the relay
6590 daemon. For example, with http://diamon.org/babeltrace[Babeltrace]:
6595 $ babeltrace --input-format=lttng-live
6596 net://localhost/host/hostname/my-session
6603 * `hostname` with the host name of the target system.
6604 * `my-session` with the name of the tracing session to view.
6607 . Configure the tracing session as usual with the man:lttng(1)
6608 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6610 You can list the available live tracing sessions with Babeltrace:
6614 $ babeltrace --input-format=lttng-live net://localhost
6617 You can start the relay daemon on another system. In this case, you need
6618 to specify the relay daemon's URL when you create the tracing session
6619 with the opt:lttng-create(1):--set-url option. You also need to replace
6620 `localhost` in the procedure above with the host name of the system on
6621 which the relay daemon is running.
6623 See man:lttng-create(1) and man:lttng-relayd(8) for the complete list of
6624 command-line options.
6628 [[taking-a-snapshot]]
6629 === Take a snapshot of the current sub-buffers of a tracing session
6631 The normal behavior of LTTng is to append full sub-buffers to growing
6632 trace data files. This is ideal to keep a full history of the events
6633 that occurred on the target system, but it can
6634 represent too much data in some situations. For example, you may wish
6635 to trace your application continuously until some critical situation
6636 happens, in which case you only need the latest few recorded
6637 events to perform the desired analysis, not multi-gigabyte trace files.
6639 With the man:lttng-snapshot(1) command, you can take a snapshot of the
6640 current sub-buffers of a given <<tracing-session,tracing session>>.
6641 LTTng can write the snapshot to the local file system or send it over
6646 . Create a tracing session in _snapshot mode_:
6651 $ lttng create my-session --snapshot
6655 The <<channel-overwrite-mode-vs-discard-mode,event loss mode>> of
6656 <<channel,channels>> created in this mode is automatically set to
6657 _overwrite_ (flight recorder mode).
6659 . Configure the tracing session as usual with the man:lttng(1)
6660 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6662 . **Optional**: When you need to take a snapshot,
6663 <<basic-tracing-session-control,stop tracing>>.
6665 You can take a snapshot when the tracers are active, but if you stop
6666 them first, you are sure that the data in the sub-buffers does not
6667 change before you actually take the snapshot.
6674 $ lttng snapshot record --name=my-first-snapshot
6678 LTTng writes the current sub-buffers of all the current tracing
6679 session's channels to trace files on the local file system. Those trace
6680 files have `my-first-snapshot` in their name.
6682 There is no difference between the format of a normal trace file and the
6683 format of a snapshot: viewers of LTTng traces also support LTTng
6686 By default, LTTng writes snapshot files to the path shown by
6687 `lttng snapshot list-output`. You can change this path or decide to send
6688 snapshots over the network using either:
6690 . An output path or URL that you specify when you create the
6692 . An snapshot output path or URL that you add using
6693 `lttng snapshot add-output`
6694 . An output path or URL that you provide directly to the
6695 `lttng snapshot record` command.
6697 Method 3 overrides method 2, which overrides method 1. When you
6698 specify a URL, a relay daemon must listen on a remote system (see
6699 <<sending-trace-data-over-the-network,Send trace data over the network>>).
6704 === Use the machine interface
6706 With any command of the man:lttng(1) command-line tool, you can set the
6707 opt:lttng(1):--mi option to `xml` (before the command name) to get an
6708 XML machine interface output, for example:
6712 $ lttng --mi=xml enable-event --kernel --syscall open
6715 A schema definition (XSD) is
6716 https://github.com/lttng/lttng-tools/blob/stable-2.8/src/common/mi-lttng-3.0.xsd[available]
6717 to ease the integration with external tools as much as possible.
6721 [[metadata-regenerate]]
6722 === Regenerate the metadata of an LTTng trace
6724 An LTTng trace, which is a http://diamon.org/ctf[CTF] trace, has both
6725 data stream files and a metadata file. This metadata file contains,
6726 amongst other things, information about the offset of the clock sources
6727 used to timestamp <<event,event records>> when tracing.
6729 If, once a <<tracing-session,tracing session>> is
6730 <<basic-tracing-session-control,started>>, a major
6731 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
6732 happens, the trace's clock offset also needs to be updated. You
6733 can use the man:lttng-metadata(1) command to do so.
6735 The main use case of this command is to allow a system to boot with
6736 an incorrect wall time and trace it with LTTng before its wall time
6737 is corrected. Once the system is known to be in a state where its
6738 wall time is correct, it can run `lttng metadata regenerate`.
6740 To regenerate the metadata of an LTTng trace:
6742 * Use the `regenerate` action of the man:lttng-metadata(1) command:
6747 $ lttng metadata regenerate
6753 `lttng metadata regenerate` has the following limitations:
6755 * Tracing session <<creating-destroying-tracing-sessions,created>>
6757 * User space <<channel,channels>>, if any, using
6758 <<channel-buffering-schemes,per-user buffering>>.
6763 [[persistent-memory-file-systems]]
6764 === Record trace data on persistent memory file systems
6766 https://en.wikipedia.org/wiki/Non-volatile_random-access_memory[Non-volatile random-access memory]
6767 (NVRAM) is random-access memory that retains its information when power
6768 is turned off (non-volatile). Systems with such memory can store data
6769 structures in RAM and retrieve them after a reboot, without flushing
6770 to typical _storage_.
6772 Linux supports NVRAM file systems thanks to either
6773 http://pramfs.sourceforge.net/[PRAMFS] or
6774 https://www.kernel.org/doc/Documentation/filesystems/dax.txt[DAX]{nbsp}+{nbsp}http://lkml.iu.edu/hypermail/linux/kernel/1504.1/03463.html[pmem]
6775 (requires Linux 4.1+).
6777 This section does not describe how to operate such file systems;
6778 we assume that you have a working persistent memory file system.
6780 When you create a <<tracing-session,tracing session>>, you can specify
6781 the path of the shared memory holding the sub-buffers. If you specify a
6782 location on an NVRAM file system, then you can retrieve the latest
6783 recorded trace data when the system reboots after a crash.
6785 To record trace data on a persistent memory file system and retrieve the
6786 trace data after a system crash:
6788 . Create a tracing session with a sub-buffer shared memory path located
6789 on an NVRAM file system:
6794 $ lttng create my-session --shm-path=/path/to/shm
6798 . Configure the tracing session as usual with the man:lttng(1)
6799 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6801 . After a system crash, use the man:lttng-crash(1) command-line tool to
6802 view the trace data recorded on the NVRAM file system:
6807 $ lttng-crash /path/to/shm
6811 The binary layout of the ring buffer files is not exactly the same as
6812 the trace files layout. This is why you need to use man:lttng-crash(1)
6813 instead of your preferred trace viewer directly.
6815 To convert the ring buffer files to LTTng trace files:
6817 * Use the opt:lttng-crash(1):--extract option of man:lttng-crash(1):
6822 $ lttng-crash --extract=/path/to/trace /path/to/shm
6830 [[lttng-modules-ref]]
6831 === noch:{LTTng-modules}
6834 [[lttng-modules-tp-fields]]
6835 ==== Tracepoint fields macros (for `TP_FIELDS()`)
6837 [[tp-fast-assign]][[tp-struct-entry]]The available macros to define
6838 tracepoint fields, which must be listed within `TP_FIELDS()` in
6839 `LTTNG_TRACEPOINT_EVENT()`, are:
6841 [role="func-desc growable",cols="asciidoc,asciidoc"]
6842 .Available macros to define LTTng-modules tracepoint fields
6844 |Macro |Description and parameters
6847 +ctf_integer(__t__, __n__, __e__)+
6849 +ctf_integer_nowrite(__t__, __n__, __e__)+
6851 +ctf_user_integer(__t__, __n__, __e__)+
6853 +ctf_user_integer_nowrite(__t__, __n__, __e__)+
6855 Standard integer, displayed in base 10.
6858 Integer C type (`int`, `long`, `size_t`, ...).
6864 Argument expression.
6867 +ctf_integer_hex(__t__, __n__, __e__)+
6869 +ctf_user_integer_hex(__t__, __n__, __e__)+
6871 Standard integer, displayed in base 16.
6880 Argument expression.
6882 |+ctf_integer_oct(__t__, __n__, __e__)+
6884 Standard integer, displayed in base 8.
6893 Argument expression.
6896 +ctf_integer_network(__t__, __n__, __e__)+
6898 +ctf_user_integer_network(__t__, __n__, __e__)+
6900 Integer in network byte order (big-endian), displayed in base 10.
6909 Argument expression.
6912 +ctf_integer_network_hex(__t__, __n__, __e__)+
6914 +ctf_user_integer_network_hex(__t__, __n__, __e__)+
6916 Integer in network byte order, displayed in base 16.
6925 Argument expression.
6928 +ctf_string(__n__, __e__)+
6930 +ctf_string_nowrite(__n__, __e__)+
6932 +ctf_user_string(__n__, __e__)+
6934 +ctf_user_string_nowrite(__n__, __e__)+
6936 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
6942 Argument expression.
6945 +ctf_array(__t__, __n__, __e__, __s__)+
6947 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
6949 +ctf_user_array(__t__, __n__, __e__, __s__)+
6951 +ctf_user_array_nowrite(__t__, __n__, __e__, __s__)+
6953 Statically-sized array of integers.
6956 Array element C type.
6962 Argument expression.
6968 +ctf_array_bitfield(__t__, __n__, __e__, __s__)+
6970 +ctf_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
6972 +ctf_user_array_bitfield(__t__, __n__, __e__, __s__)+
6974 +ctf_user_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
6976 Statically-sized array of bits.
6978 The type of +__e__+ must be an integer type. +__s__+ is the number
6979 of elements of such type in +__e__+, not the number of bits.
6982 Array element C type.
6988 Argument expression.
6994 +ctf_array_text(__t__, __n__, __e__, __s__)+
6996 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
6998 +ctf_user_array_text(__t__, __n__, __e__, __s__)+
7000 +ctf_user_array_text_nowrite(__t__, __n__, __e__, __s__)+
7002 Statically-sized array, printed as text.
7004 The string does not need to be null-terminated.
7007 Array element C type (always `char`).
7013 Argument expression.
7019 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
7021 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7023 +ctf_user_sequence(__t__, __n__, __e__, __T__, __E__)+
7025 +ctf_user_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7027 Dynamically-sized array of integers.
7029 The type of +__E__+ must be unsigned.
7032 Array element C type.
7038 Argument expression.
7041 Length expression C type.
7047 +ctf_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7049 +ctf_user_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7051 Dynamically-sized array of integers, displayed in base 16.
7053 The type of +__E__+ must be unsigned.
7056 Array element C type.
7062 Argument expression.
7065 Length expression C type.
7070 |+ctf_sequence_network(__t__, __n__, __e__, __T__, __E__)+
7072 Dynamically-sized array of integers in network byte order (big-endian),
7073 displayed in base 10.
7075 The type of +__E__+ must be unsigned.
7078 Array element C type.
7084 Argument expression.
7087 Length expression C type.
7093 +ctf_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7095 +ctf_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7097 +ctf_user_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7099 +ctf_user_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7101 Dynamically-sized array of bits.
7103 The type of +__e__+ must be an integer type. +__s__+ is the number
7104 of elements of such type in +__e__+, not the number of bits.
7106 The type of +__E__+ must be unsigned.
7109 Array element C type.
7115 Argument expression.
7118 Length expression C type.
7124 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7126 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7128 +ctf_user_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7130 +ctf_user_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7132 Dynamically-sized array, displayed as text.
7134 The string does not need to be null-terminated.
7136 The type of +__E__+ must be unsigned.
7138 The behaviour is undefined if +__e__+ is `NULL`.
7141 Sequence element C type (always `char`).
7147 Argument expression.
7150 Length expression C type.
7156 Use the `_user` versions when the argument expression, `e`, is
7157 a user space address. In the cases of `ctf_user_integer*()` and
7158 `ctf_user_float*()`, `&e` must be a user space address, thus `e` must
7161 The `_nowrite` versions omit themselves from the session trace, but are
7162 otherwise identical. This means the `_nowrite` fields won't be written
7163 in the recorded trace. Their primary purpose is to make some
7164 of the event context available to the
7165 <<enabling-disabling-events,event filters>> without having to
7166 commit the data to sub-buffers.
7172 Terms related to LTTng and to tracing in general:
7175 The http://diamon.org/babeltrace[Babeltrace] project, which includes
7176 the cmd:babeltrace command, some libraries, and Python bindings.
7178 <<channel-buffering-schemes,buffering scheme>>::
7179 A layout of sub-buffers applied to a given channel.
7181 <<channel,channel>>::
7182 An entity which is responsible for a set of ring buffers.
7184 <<event,Event rules>> are always attached to a specific channel.
7187 A reference of time for a tracer.
7189 <<lttng-consumerd,consumer daemon>>::
7190 A process which is responsible for consuming the full sub-buffers
7191 and write them to a file system or send them over the network.
7193 <<channel-overwrite-mode-vs-discard-mode,discard mode>>:: The event loss
7194 mode in which the tracer _discards_ new event records when there's no
7195 sub-buffer space left to store them.
7198 The consequence of the execution of an instrumentation
7199 point, like a tracepoint that you manually place in some source code,
7200 or a Linux kernel KProbe.
7202 An event is said to _occur_ at a specific time. Different actions can
7203 be taken upon the occurance of an event, like record the event's payload
7206 <<channel-overwrite-mode-vs-discard-mode,event loss mode>>::
7207 The mechanism by which event records of a given channel are lost
7208 (not recorded) when there is no sub-buffer space left to store them.
7210 [[def-event-name]]event name::
7211 The name of an event, which is also the name of the event record.
7212 This is also called the _instrumentation point name_.
7215 A record, in a trace, of the payload of an event which occured.
7217 <<event,event rule>>::
7218 Set of conditions which must be satisfied for one or more occuring
7219 events to be recorded.
7221 `java.util.logging`::
7223 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[core logging facilities].
7225 <<instrumenting,instrumentation>>::
7226 The use of LTTng probes to make a piece of software traceable.
7228 instrumentation point::
7229 A point in the execution path of a piece of software that, when
7230 reached by this execution, can emit an event.
7232 instrumentation point name::
7233 See _<<def-event-name,event name>>_.
7236 A http://logging.apache.org/log4j/1.2/[logging library] for Java
7237 developed by the Apache Software Foundation.
7240 Level of severity of a log statement or user space
7241 instrumentation point.
7244 The _Linux Trace Toolkit: next generation_ project.
7246 <<lttng-cli,cmd:lttng>>::
7247 A command-line tool provided by the LTTng-tools project which you
7248 can use to send and receive control messages to and from a
7252 The https://github.com/lttng/lttng-analyses[LTTng analyses] project,
7253 which is a set of analyzing programs that are used to obtain a
7254 higher level view of an LTTng trace.
7256 cmd:lttng-consumerd::
7257 The name of the consumer daemon program.
7260 A utility provided by the LTTng-tools project which can convert
7261 ring buffer files (usually
7262 <<persistent-memory-file-systems,saved on a persistent memory file system>>)
7265 LTTng Documentation::
7268 <<lttng-live,LTTng live>>::
7269 A communication protocol between the relay daemon and live viewers
7270 which makes it possible to see events "live", as they are received by
7273 <<lttng-modules,LTTng-modules>>::
7274 The https://github.com/lttng/lttng-modules[LTTng-modules] project,
7275 which contains the Linux kernel modules to make the Linux kernel
7276 instrumentation points available for LTTng tracing.
7279 The name of the relay daemon program.
7281 cmd:lttng-sessiond::
7282 The name of the session daemon program.
7285 The https://github.com/lttng/lttng-tools[LTTng-tools] project, which
7286 contains the various programs and libraries used to
7287 <<controlling-tracing,control tracing>>.
7289 <<lttng-ust,LTTng-UST>>::
7290 The https://github.com/lttng/lttng-ust[LTTng-UST] project, which
7291 contains libraries to instrument user applications.
7293 <<lttng-ust-agents,LTTng-UST Java agent>>::
7294 A Java package provided by the LTTng-UST project to allow the
7295 LTTng instrumentation of `java.util.logging` and Apache log4j 1.2
7298 <<lttng-ust-agents,LTTng-UST Python agent>>::
7299 A Python package provided by the LTTng-UST project to allow the
7300 LTTng instrumentation of Python logging statements.
7302 <<channel-overwrite-mode-vs-discard-mode,overwrite mode>>::
7303 The event loss mode in which new event records overwrite older
7304 event records when there's no sub-buffer space left to store them.
7306 <<channel-buffering-schemes,per-process buffering>>::
7307 A buffering scheme in which each instrumented process has its own
7308 sub-buffers for a given user space channel.
7310 <<channel-buffering-schemes,per-user buffering>>::
7311 A buffering scheme in which all the processes of a Unix user share the
7312 same sub-buffer for a given user space channel.
7314 <<lttng-relayd,relay daemon>>::
7315 A process which is responsible for receiving the trace data sent by
7316 a distant consumer daemon.
7319 A set of sub-buffers.
7321 <<lttng-sessiond,session daemon>>::
7322 A process which receives control commands from you and orchestrates
7323 the tracers and various LTTng daemons.
7325 <<taking-a-snapshot,snapshot>>::
7326 A copy of the current data of all the sub-buffers of a given tracing
7327 session, saved as trace files.
7330 One part of an LTTng ring buffer which contains event records.
7333 The time information attached to an event when it is emitted.
7336 A set of files which are the concatenations of one or more
7337 flushed sub-buffers.
7340 The action of recording the events emitted by an application
7341 or by a system, or to initiate such recording by controlling
7345 The http://tracecompass.org[Trace Compass] project and application.
7348 An instrumentation point using the tracepoint mechanism of the Linux
7349 kernel or of LTTng-UST.
7351 tracepoint definition::
7352 The definition of a single tracepoint.
7355 The name of a tracepoint.
7357 tracepoint provider::
7358 A set of functions providing tracepoints to an instrumented user
7361 Not to be confused with a _tracepoint provider package_: many tracepoint
7362 providers can exist within a tracepoint provider package.
7364 tracepoint provider package::
7365 One or more tracepoint providers compiled as an object file or as
7369 A software which records emitted events.
7371 <<domain,tracing domain>>::
7372 A namespace for event sources.
7374 <<tracing-group,tracing group>>::
7375 The Unix group in which a Unix user can be to be allowed to trace the
7378 <<tracing-session,tracing session>>::
7379 A stateful dialogue between you and a <<lttng-sessiond,session
7383 An application running in user space, as opposed to a Linux kernel
7384 module, for example.