Red Hat Training

A Red Hat training course is available for RHEL 8

Chapter 26. Adding and removing tracepoints from a running perf collector without stopping or restarting perf

By using the control pipe interface to enable and disable different tracepoints in a running perf collector, you can dynamically adjust what data you are collecting without having to stop or restart perf. This ensures you do not lose performance data that would have otherwise been recorded during the stopping or restarting process.

26.1. Adding tracepoints to a running perf collector without stopping or restarting perf

Add tracepoints to a running perf collector using the control pipe interface to adjust the data you are recording without having to stop perf and losing performance data.

Prerequisites

  • The perf user space tool is installed. For more information, see Installing perf.

Procedure

  1. Configure the control pipe interface:

    # mkfifo control ack perf.pipe
  2. Run perf record with the control file setup and events you are interested in enabling:

    # perf record --control=fifo:control,ack -D -1 --no-buffering -e 'sched:*' -o - > perf.pipe

    In this example, declaring 'sched:*' after the -e option starts perf record with scheduler events.

  3. In a second terminal, start the read side of the control pipe:

    # cat perf.pipe | perf --no-pager script -i -

    Starting the read side of the control pipe triggers the following message in the first terminal:

    Events disabled
  4. In a third terminal, enable a tracepoint using the control file:

    # echo 'enable sched:sched_process_fork' > control

    This command triggers perf to scan the current event list in the control file for the declared event. If the event is present, the tracepoint is enabled and the following message appears in the first terminal:

    event sched:sched_process_fork enabled

    Once the tracepoint is enabled, the second terminal displays the output from perf detecting the tracepoint:

    bash 33349 [034] 149587.674295: sched:sched_process_fork: comm=bash pid=33349 child_comm=bash child_pid=34056

26.2. Removing tracepoints from a running perf collector without stopping or restarting perf

Remove tracepoints from a running perf collector using the control pipe interface to reduce the scope of data you are collecting without having to stop perf and losing performance data.

Prerequisites

Procedure

  • Remove the tracepoint:

    # echo 'disable sched:sched_process_fork' > control
    Note

    This example assumes you have previously loaded scheduler events into the control file and enabled the tracepoint sched:sched_process_fork.

    This command triggers perf to scan the current event list in the control file for the declared event. If the event is present, the tracepoint is disabled and the following message appears in the terminal used to configure the control pipe:

    event sched:sched_process_fork disabled