Coder Social home page Coder Social logo

bt.sh's Introduction

bt.sh

bt.sh is a simple set of bash functions that can be incorporated into an existing set of shell scripts to provide a timechart-like trace view of different script phases.

Here's an example where BT_WIDTH is chosen to fit the output nicely into this README:

[simonsj@simonsj-lx2 : bt.sh] BT_WIDTH=50 ./example.sh
Build Trace Start (example.sh:7)

              ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▃▁▁▁▁▁▁▂▇▇▇▇▇███████▆▆▁▁▁▁ * CPU Utilization
[    8.111s ] ├────────────────────────────────────────────────┤ * full example
[    4.036s ] ├──────────────────────┤                           * serial cats
[    2.008s ] ├──────────┤                                       * cat + gzip 1
[    2.008s ]             ├──────────┤                           * cat + gzip 2
[    1.011s ]                          ├────┤                    * sleep 1.0
[    2.023s ]                                ├──────────┤        * parallel cats
[    2.010s ]                                ├──────────┤        * cat + gzip 3
[    2.009s ]                                ├──────────┤        * cat + gzip 4
[    1.012s ]                                            ├─────┤ * final sleep 1.0

     one '.' unit is less than:    0.162s
                    total time:    8.122s

Build Trace End (example.sh:7)

Time goes left-to-right, and tracepoints are ordered top-to-bottom chronologically according to when they started.

The cumulative time for a trace is emitted on the far-left column, and the trace's globally unique name is emitted on the far-right column.

CPU usage is sampled with mpstat and emitted using spark.

Motivation

bt.sh was originally created to aid reasoning about the numerous highly-parallelized and heavily-cached build and CI test phases used in GitHub's internal development of GitHub Enterprise.

Explicit goals:

  • single-file consumption
  • a small set of trivial, commonly-available Linux dependencies
  • replace "eyeball grepping" build timestamps and manually computing different phase durations
  • verify work intended to be run in parallel is indeed doing so
  • "stats in your face" to help with identifying regressions
  • identify phases which are CPU-bound (and not)

Non-goals:

  • not suitable for fine-grained perf measurement
    • if date overhead would matter in your context, bt.sh is not a good fit

Usage

See example.sh for an example script.

In a nutshell:

#!/bin/bash
. bt.sh                            # source bt.sh
bt_init                            # initialize

bt_start 'some phase of my build'  # start trace
...                                # (your actual script)
bt_end 'some phase of my build'    # end trace

bt_cleanup                         # cleanup once when done

The bt_init and bt_cleanup are okay to be invoked in a nested fashion from multiple scripts: a report will be emitted when the first script to have done bt_init finally reaches its bt_cleanup.

The bt_start and bt_end invocations for a given measurement each require their string description argument to match, and to be globally unique. The description name is used to key into the measurement database: simple files placed into a mktemp'd directory under /tmp, $BT_DIR. Files for each measurement are organized as:

  actual file:  $BT_DIR/<description checksum>.<start timestamp>
  symlink:      $BT_DIR/<description checksum> --> actual file

With this scheme, lookup for bt_end is constant-time via the symlink, and the stats can be sorted by chronological start time using the actual files' names.

Upon bt_end, the final line of the data file looks like:

  <end timestamp> <end_caller_source_file:lineno> <description text ...>

Where above, the first two columns are stable, and everything from the third word to the end of line is the string description.

Assumptions

  • some bash-isms are relied upon: export -f, BASH_SOURCE, BASH_LINENO

Dependencies

bt.sh is expected to work without much effort on a modern Linux or MacOS.

It depends on these utilities:

  • bash
  • date, yes, seq, cksum, head, mktemp (GNU coreutils)
  • mpstat (sysstat package)
  • GNU bc (math package)
  • POSIX: awk, tail, sed, basename

On MacOS with Homebrew-installed GNU coreutils, you can specify BT_HEAD=ghead and BT_DATE=gdate to ensure that those versions of head and date are used.

License

See the LICENSE file for license rights and limitations (MIT).

bt.sh's People

Contributors

alindeman avatar simonsj avatar zerowidth avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar

bt.sh's Issues

Thanks!

Hi, @simonsj!

I just wanted to say that I love bt.sh!

Thanks 😀

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.