2 id: viewing-and-analyzing-your-traces
5 This section describes how to visualize the data gathered after tracing
6 the Linux kernel or a user space application.
8 Many ways exist to read your LTTng traces:
10 * **`babeltrace`** is a command line utility which converts trace formats;
11 it supports the format used by LTTng,
12 <abbr title="Common Trace Format">CTF</abbr>, as well as a basic
13 text output which may be `grep`ed. The `babeltrace` command is
15 <a href="http://www.efficios.com/babeltrace" class="ext">Babeltrace</a> project.
16 * Babeltrace also includes a **Python binding** so that you may
17 easily open and read an LTTng trace with your own script, benefiting
18 from the power of Python.
19 * The **<a href="https://eclipse.org/downloads/packages/eclipse-ide-cc-developers/lunar" class="ext">
20 Eclise IDE for C/C++ Developers</a>**
21 includes the Tracing and Monitoring Framework (TMF) plugin which
22 supports LTTng traces, amongst others.
24 LTTng trace files are usually recorded in the `~/lttng-traces` directory.
25 Let's now view the trace and perform a basic analysis using
28 The simplest way to list all the recorded events of a trace is to pass its
29 path to `babeltrace` with no options:
32 babeltrace ~/lttng-traces/my-session
35 `babeltrace` will find all traces within the given path recursively and
36 output all their events, merging them intelligently.
38 Listing all the system calls of a Linux kernel trace with their arguments is
39 easy with `babeltrace` and `grep`:
42 babeltrace ~/lttng-traces/my-kernel-session | grep sys_
45 Counting events is also straightforward:
48 babeltrace ~/lttng-traces/my-kernel-session | grep sys_read | wc -l
51 The text output of `babeltrace` is useful for isolating events by simple
52 matching using `grep` and similar utilities. However, more elaborate filters
53 such as keeping only events with a field value falling within a specific range
54 are not trivial to write using a shell. Moreover, reductions and even the
55 most basic computations involving multiple events are virtually impossible
58 Fortunately, Babeltrace ships with a Python 3 binding which makes it
59 really easy to read the events of an LTTng trace sequentially and compute
60 the desired information.
62 Here's a simple example using the Babeltrace Python binding. The following
63 script accepts an LTTng Linux kernel trace path as its first argument and
64 outputs the short names of the top 5 running processes on CPU 0 during the
69 from collections import Counter
74 if len(sys.argv) != 2:
75 msg = 'Usage: python {} TRACEPATH'.format(sys.argv[0])
78 # a trace collection holds one to many traces
79 col = babeltrace.TraceCollection()
81 # add the trace provided by the user
82 # (LTTng traces always have the 'ctf' format)
83 if col.add_trace(sys.argv[1], 'ctf') is None:
84 raise RuntimeError('Cannot add trace')
86 # this counter dict will hold execution times:
88 # task command name -> total execution time (ns)
89 exec_times = Counter()
91 # this holds the last `sched_switch` timestamp
95 for event in col.events:
96 # keep only `sched_switch` events
97 if event.name != 'sched_switch':
100 # keep only events which happened on CPU 0
101 if event['cpu_id'] != 0:
105 cur_ts = event.timestamp
111 # previous task command (short) name
112 prev_comm = event['prev_comm']
114 # initialize entry in our dict if not yet done
115 if prev_comm not in exec_times:
116 exec_times[prev_comm] = 0
118 # compute previous command execution time
119 diff = cur_ts - last_ts
121 # update execution time of this command
122 exec_times[prev_comm] += diff
124 # update last timestamp
128 for name, ns in exec_times.most_common()[:5]:
130 print('{:20}{} s'.format(name, s))
133 if __name__ == '__main__':
137 Save this script as `top5proc.py` and run it with Python 3, providing the
138 path to an LTTng Linux kernel trace as the first argument:
141 python3 top5proc.py ~/lttng-sessions/my-session-.../kernel
144 Make sure the path you provide is the directory containing actual trace
145 files (`channel0_0`, `metadata`, etc.): the `babeltrace` utility recurses
146 directories, but the Python binding does not.
148 Here's an example of output:
151 swapper/0 48.607245889 s
152 chromium 7.192738188 s
153 pavucontrol 0.709894415 s
154 Compositor 0.660867933 s
155 Xorg.bin 0.616753786 s
158 Note that `swapper/0` is the "idle" process of CPU 0 on Linux; since we
159 weren't using the CPU that much when tracing, its first position in the list