babeltrace2-intro — Introduction to Babeltrace 2
This manual page is an introduction to the Babeltrace 2 project.
The WHAT IS BABELTRACE 2? section describes the parts of the project and shows the major changes from Babeltrace 1 to Babeltrace 2 while the BABELTRACE 2 CONCEPTS section defines the core concepts of Babeltrace 2.
The TRACE PROCESSING GRAPH REPRESENTATION section shows how some concepts are visually represented in other Babeltrace 2 manual pages.
Babeltrace 2 is an open-source software project of which the purpose is to process or convert traces.
The Babeltrace 2 project includes the following parts:
A shared library with a C API.
With libbabeltrace2, you can programmatically create plugins and component classes, build and run trace processing graphs, and more (see the BABELTRACE 2 CONCEPTS section for more details about those concepts).
All the other Babeltrace 2 parts rely on this library.
A command-line interface which uses libbabeltrace2 to load plugins, create a trace processing graph, create components, connect their ports correctly, and run the graph.
You can also use
babeltrace2 to list the available plugins or to
query an object from a component class.
A Python 3 package (
bt2) which offers a Pythonic interface of
You can perform the same operations which are available in libbabeltrace2 with the Python bindings, but more conveniently and with less code. However, the Python bindings are less performant than libbabeltrace2.
The Babeltrace 2 plugins shipped with the project.
Those plugins are not special in that they only rely on libbabeltrace2 and you don’t need them to use libbabeltrace2, babeltrace2(1), or the Python bindings. However, the project’s plugins provide many widely used trace format encoders/decoders as well as common trace processing graph utilities.
The Babeltrace 2 project’s plugins are:
Common Trace Format (CTF) input/output, including the LTTng live source.
Graph utilities specific to LTTng traces.
Plain text input/output.
Common graph utilities (muxer, trimmer, counter, dummy sink).
This manual page is an introduction to Babeltrace 2, a rewrite of Babeltrace 1 with a focus on extensibility, flexibility, and interoperability.
Babeltrace 1 exists since 2010.
You can install both projects on the same file system as there are no file name conflicts.
The major improvements brought by Babeltrace 2 are:
Plugins are not just trace format encoders and decoders: they package source, filter, and sink component classes so that you can connect specialized, reusable components together in a trace processing graph to create a customized trace conversion or analysis device.
All the parts of the Babeltrace 2 project run on the major operating systems, including Windows and macOS.
Some component classes, such as sink.text.pretty
(similar to the
text output format of babeltrace(1)) and
sink.text.details, can write color codes to the standard
output when it’s connected to a color-enabled terminal.
The Babeltrace 2 log, printed to the standard output, can also be colorized.
Whereas you can convert traces from one format to another with Babeltrace 1’s CLI tool, babeltrace(1), you can also execute a custom trace manipulation task with babeltrace2(1) thanks to the babeltrace2-run(1) command.
The babeltrace2-convert(1) command features an automatic source component discovery algorithm to find the best suited components to create for a given non-option argument (file or directory path, or custom string like an LTTng live URL).
$ babeltrace2 /path/to/ctf/trace
$ babeltrace2 net://localhost/host/myhost/my-session
The source.ctf.fs component class, which is more or less the
equivalent of Babeltrace 1’s
ctf input format, has features
not found in Babeltrace 1:
The component handles many trace quirks which are the results of known tracer bugs and corner cases (LTTng-UST, LTTng-modules, and barectf), making it possible to decode malformed packets.
The component merges CTF traces sharing the same UUID into a single, logical trace.
This feature supports LTTng 2.11’s tracing session rotation trace chunks.
With a sink.ctf.fs component, you can create CTF traces on the file system.
$ babeltrace2 /path/to/input/trace \ --output-format=ctf --output=trace-dir
The babeltrace(1) command exits successfully when it cannot find
an LTTng live (
--input-format=lttng-live option) tracing session.
If the action is
end, the message iterator does like babeltrace(1)
and simply ends successfully.
If the action is
continue (the default), the message iterator never
ends: it keeps on trying until the tracing session exists, indeed
subscribing to the session.
libbabeltrace2 shares nothing with libbabeltrace.
The Babeltrace 2 library C API has features such as:
A single header file.
Function precondition and postcondition checking.
Object-oriented model with shared and unique objects.
Strict C typing and
Rich, thread-safe error reporting.
Per-component and per-subsystem logging levels.
Trace intermediate representation (IR) objects to make the API trace-format-agnostic.
A versioned protocol for message interchange between components to enable forward and backward compatibility.
You can build the library in developer mode to enable an extensive set of function precondition and postcondition checks.
The developer mode can help detect programming errors early when you develop a Babeltrace 2 plugin or an application using libbabeltrace2.
See the project’s
README for build-time requirements and detailed
This section defines the main concepts of the Babeltrace 2 project.
These concepts translate into types and functions in
libbabeltrace2 and its Python bindings, but also as command-line actions and options in the
babeltrace2 program. The other Babeltrace 2
manual pages assume that you are familiar with the following
Some Babeltrace 2 concepts are interdependent: it is normal to jump from one definition to another to understand the big picture.
A reusable class which you can instantiate as one or more components within a trace processing graph.
There are three types of component classes used to create the three types of components: source, filter, and sink.
A component class implements methods, one of which is an initialization
method, or constructor, to create a component. You pass initialization
parameters to this method to customize the created component. For
example, the initialization method of the source.ctf.fs
component class accepts a mandatory
parameter which is an array of file system path(s) to the CTF trace(s).
It also accepts an optional
parameter which is an offset, in nanoseconds, to add to all the clock
classes (descriptors of stream clocks) found in the traces’s metadata.
A component class can have a description and a help text.
A node within a trace processing graph.
There are three types of components:
An input component which produces messages.
Examples: CTF files input, log file input, LTTng live input, random event generator.
An intermediate component which can transform the messages it consumes, augment them, sort them, discard them, or create new ones.
Examples: filter which removes messages based on an expression, filter which adds debugging information to selected events, message muxer, trace trimmer.
An output component which consumes messages and usually writes them to one or more formatted files.
Examples: log file output, CTF files output, pretty-printed plain text output.
Components are connected together within a trace processing graph through their ports. Source components have output ports, sink components have input ports, and filter components have both.
A component is the instance of a component class. The terms component and component class instance are equivalent.
Within a trace processing graph, each component has a unique name. This
is not the name of its component class, but an instance name. If
is a component class name, than
John could be component
Once a graph is configured (the first time it runs), you cannot add components to it for the remaining graph’s lifetime.
A connection point, on a component, from which are sent or where are received messages when the trace processing graph runs.
An output port is from where messages are sent. An input port is where messages are received. Source components have output ports, sink components have input ports, and filter components have both.
You can only connect an output port to a single input port.
All ports do not need to be connected.
A filter or sink component receiving messages from its input ports is said to consume messages.
The link between an output port and input port is a connection.
Once a graph is configured (the first time it runs), you cannot connect ports for the remaining graph’s lifetime.
The link between an output port and an input port through which messages flow when a trace processing graph runs.
An iterator on an input port of which the returned elements are messages.
A component or another message iterator can create many message iterators on a single input port, before or while the trace processing graph runs.
The element of a message iterator.
Messages flow from output ports to input ports.
A source component message iterator produces messages, while a sink component consumes them. A filter component message iterator can both consume and produce messages.
The main types of messages are:
A trace event record within a packet or within a stream.
The beginning of a packet within a stream.
A packet is a conceptual container of events.
The end of a packet within a stream.
The beginning of a stream.
A stream is a conceptual container of packets and/or events.
Usually, a given source component’s output port sends packet and event messages which belong to a single stream, but it’s not required.
The end of a stream.
A count of discarded events within a given time interval for a given stream.
A count of discarded packets within a given time interval for a given stream.
A filter graph where nodes are components and messages flow from output ports to input ports.
You can build a trace processing graph with libbabeltrace2, with the Babeltrace 2 Python bindings, or with the babeltrace2-run(1) and babeltrace2-convert(1) CLI commands.
When a trace processing graph runs, the sink components consume messages from their input ports, making all the graph’s message iterators work one message at a time to perform the trace conversion or analysis duty.
A container, or package, of component classes as a shared library or Python module.
Each component class within a plugin has a type (source, filter, or sink) and a name. The type and name pair is unique within a given plugin.
libbabeltrace2 can load a plugin (
.py file) at run time: the result is a plugin object in which you can
find a specific component class and instantiate it within a
trace processing graph as a component.
babeltrace2 program uses the
COMP-CLS-TYPE.PLUGIN-NAME.COMP-CLS-NAME format to identify a specific
component class within a specific plugin.
COMP-CLS-TYPE is either
You can list the available Babeltrace 2 plugins with the babeltrace2-list-plugins(1) command.
An operation with which you can get a named object from a component class, possibly with custom query parameters.
The plain text metadata stream of a CTF trace and the available LTTng live sessions of a given LTTng relay daemon are examples of query objects.
You can use libbabeltrace2, the Babeltrace 2 Python bindings, or the babeltrace2-query(1) CLI command to query a component class’s object.
In the Babeltrace 2 manual pages, a component is represented with a
box. The box has the component class type,
plugin name, and component class name at the top. Just below,
between square brackets, is its component name within the trace processing graph. Each port is represented with an
on the border(s) of the component box with its name inside the box.
Output ports are on the box’s right border while input ports are on the
box’s left border.
For example, here’s a source component box:
+------------+ | src.ctf.fs | | [my-src] | | | | stream0 @ | stream1 @ | stream2 @ +------------+
This one is an instance of the source.ctf.fs component class
my-src. It has three output ports named
A trace processing graph is represented with multiple component boxes connected together. The connections are arrows from output ports to input ports.
For example, here’s a simple conversion graph:
+------------+ +-----------------+ +------------------+ | src.ctf.fs | | flt.utils.muxer | | sink.text.pretty | | [ctf] | | [muxer] | | [text] | | | | | | | | stream0 @--->@ in0 out @--->@ in | | stream1 @--->@ in1 | +------------------+ | stream2 @--->@ in2 | +------------+ @ in3 | +-----------------+
Note that input port
in3 of component
muxer is not connected in this
Sometimes, we symbolically represent other resources which are consumed
from or produced by components. In this case, arrows are used, but they
do not go to or from port symbols (
@), except for messages. For
example, in the graph above, the
ctf source component consumes a CTF
trace and the
text sink component prints plain text to the terminal,
so here’s a more complete diagram:
CTF trace | | +------------+ +-----------------+ +------------------+ | | src.ctf.fs | | flt.utils.muxer | | sink.text.pretty | '-->| [ctf] | | [muxer] | | [text] | | | | | | | | stream0 @--->@ in0 out @--->@ in | | stream1 @--->@ in1 | +-----+------------+ | stream2 @--->@ in2 | | +------------+ @ in3 | '--> Terminal +-----------------+
Here’s another example of a more complex graph which splits a specific stream using some criteria:
+------------+ +-----------------+ +------------------+ | src.ctf.fs | | flt.utils.muxer | | sink.text.pretty | | [ctf-in] | | [muxer] | | [text] | | | | | | | | stream0 @--->@ in0 out @--->@ in | | stream1 @--->@ in1 | +------------------+ | stream2 @-. @ in2 | +------------+ | +-----------------+ +-------------+ | | sink.ctf.fs | | | [ctf-out0] | | +-------------------+ | | | | flt.some.splitter | .->@ in | | | [splitter] | | +-------------+ | | | | '->@ in A @-' +-------------+ | B @-. | sink.ctf.fs | +-------------------+ | | [ctf-out1] | | | | '->@ in | +-------------+