Container-aware Tracing Infrastructure

Session information has not yet been published for this event.

*

One Line Summary

Discussion on enabling tracing infrastructure support for containers.

Abstract

Container technology is increasingly becoming important thanks to the
advent of Docker. Over the years, work on containers has largely focused
on security, checkpoint/restart etc. However, given the huge uptake, it
is imperative to focus on tracing infrastructure support around
containers. Such support is required to effectively troubleshoot, debug
and optimize applications running inside containers.

Work has already been done to enhance perf stat/record to filter
container specific events from host 1. It is essential to extend such
support to uprobes, kprobes and ftrace. However, it is also essential to
extend existing Linux infrastructure including PMUs, tracepoints,
probes, ftrace to seamlessly work inside containers. Importantly, such
tracing support from within a container should provide strong security
isolation by suitably filtering events.

One approach to container-aware tracing is to make kernel tracing
namespace aware. We would like to discuss the approach, current state
and future work related to tracing support for containers and solicit
feedback and suggestions for improvement.

1 https://lkml.org/lkml/2011/2/14/40

Initial RFC patches posted: lkml.org/lkml/2015/7/15/192

Tags

tracing, perf, docker, container

Speaker

  • Biography

    Aravinda is a Software Engineer at IBM Linux Technology Center.