Michael Jeanson

Follow up on the work that was done since last year's 'LTTng: The road to container awareness' presentation. A quick glance at the base features that were merged upstream, what is still missing and the tooling integration with container runtimes we plan to build on those features.

The LTTng kernel and user-space tracer were designed for traditional Linux hosts but can also be used in a container environment. While we currently lack high-level functions and integration with container technologies, the raw data can be gathered and analysed by the seasoned lttng user. This talk will detail what is currently available, what we would like to add and hopefully what the community would expect from a container aware version of LTTng.

These are the majors topics that will be covered :

• LTTng kernel tracer

  • Namespace contexts
  • Namespaces in the statedump

• LTTng UST userspace tracer

  • Running the session daemon on the host or in the containers
  • Adding namespace contexts
  • Trace correlation between host and containers

• LTTng analyses, python analysis scripts

  • Adding container awareness to current analyses
  • Containers specific analyses

LTTng is an open source kernel and userspace tracing framework for Linux. This talk will introduce developers and system administrators alike to the toolkit and show how it can help them get visibility into development and production systems. The following themes will be covered: - A description of the different tools involved - LTTng compared to other tracing tools like *perf* and *strace* - The state of LTTng in Debian - Basic use cases and workflows - Analysis of kernel traces You should leave this talk with an uncontrollable desire to use LTTng. **URLs**: https://lttng.org