1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
|
LTTng Live trace reading how-to
Julien Desfossez
September 27th, 2013
This document presents a summary on how to use the live trace reading feature
introduced in LTTng 2.4. For the details about the protocol, please refer to
the live-reading-protocol.txt document.
Live trace reading refers to the feature of reading the trace while it is being
recorded. In order to do that, the trace must be streamed a relay even if the
viewer is running on the same machine as the tracer.
So, the first thing to do is to start a lttng-relayd process. It can be
anywhere on the network (including localhost) as long as the sessiond/consumerd
and the viewer can communicate with it over TCP/IP.
$ lttng-relayd -d
Then, we can create a session configured for streaming with the new --live
parameter.
$ lttng create --live 1000000 -U net://localhost
The --live parameter activates a session-wide timer (usec) that is responsible
for checking at a user-defined rate if new data is available. If there is new
data, it is flushed automatically, otherwise a beacon is sent to the relayd to
inform it that the stream is currently empty and the viewer can ignore this
stream up to a certain point in time.
Once the session is created, the user can activate events as usual.
In order to view the live trace, the viewer must implement the live-reading
protocol. As of now, Babeltrace[1] and LTTngTop[2] implement the protocol.
[1] https://git.efficios.com/babeltrace.git
[2] https://git.lttng.org/lttngtop.git
|
