[lttng-dev] delay in Java/python events

[Adel Belkhiri]

Hello LTTng community,

Currently, I'm tracing a java (and Python) application using LTTng. I
noticed there is a considerable delay between the time the tracepoint
instruction is executed and the time its corresponding event is recorded by
LTTng. It seems that the problem comes from the fact that LTTng uses agents
to handle java/Python events. Is there any benchmark measuring such delay?
Is it possible to minimize this delay (for example by flushing the logger)?
Currently, I'm considering putting my tracepoints in a shared library and
calling them from within Java code using JNI so that events recording will
be "synchronous" with application code execution. Is there any other
technique that I can use? Thank you for your help.

Adel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.lttng.org/pipermail/lttng-dev/attachments/20230710/a77fd2cd/attachment.htm>