@nearform/0x
Advanced tools
<img alt=0x src=assets/0x-logo.png width=350>
🔥 single-command flamegraph profiling 🔥
Discover the bottlenecks and hot paths in your code, with flamegraphs.
0x can profile and generate an interactive flamegraph for a Node process with a single command,
on any platform which Node runs on (macOs, Linux, Windows, Android...).
Older versions of Node are supported via previous 0x versions:
| 0x | Node | macOS/SmartOS | Linux | Windows |
|---|---|---|---|---|
| v4 | v8.5.0+ | ☑️ | ☑️ | ☑️ |
| v3 | v6 – v8.4.0 | ☑️ | ☑️ | ⤬ |
| v2 | v4 | ☑️ | ☑️ | ⤬ |
An example interactive flamegraph can be viewed at http://davidmarkclements.github.io/0x-demo/
npm install -g 0x
Use 0x to run a script:
0x my-app.js
Immediately open the flamegraph in the browser:
0x -o my-app.js
Automatically execute profiling command against the first port opened by profiled process:
0x -P 'autocannon localhost:$PORT' my-app.js
Use a custom node executable:
0x -- /path/to/node my-app.js
Pass custom arguments to node:
0x -- node --zero-fill-buffers my-app.js
When ready to generate a flamegraph, send a SIGINT.
The simplest way to do this is pressing CTRL+C.
When 0x catches the SIGINT, it process the stacks and
generates a profile folder (<pid>.0x), containing flamegraph.html.
The flamegraph.html file contains the 0x UI, which is explained in
docs/ui.md.
A lightweight, production server friendly, approach to generating a flamegraph is described in docs/production-servers.md.
By default, a Profile Folder will be created and named after the PID, e.g.
3866.0x (we can set this name manually using the --output-dir flag).
The Profile Folder is explained in more detail in docs/profile-folder.md
Clone this repo, run npm i -g and from the repo root run
0x examples/rest-api
In another tab run
npm run stress-rest-example
To put some load on the rest server, once that's done use ctrl + c to kill the server.
Print usage info.
Open the flamegraph in the browser using open or xdg-open (see
https://www.npmjs.com/package/open for details).
Run a given command and then generate the flamegraph.
The command as specified has access to a $PORT variable.
The $PORT variable is set according to the first port that
profiled process opens.
For instance, here's an example of using autocannon to load-test the process:
0x -P 'autocannon localhost:$PORT' app.js
When the load-test completes, the profiled processed will be sent a SIGINT and the flamegraph will be automatically generated.
Remember to use single quotes to avoid bash interpolation,
or else escape variable (e.g. 0x -P "autocannon localhost:$PORT" app.js
won't work wheras 0x -P "autocannon localhost:\$PORT" app.js will).
Note: On Windows interpolation usually occurs with %PORT%, however
in this case the dollar-prefix $PORT is the correct syntax
(because the interpolation is not shell based).
Default: ''
The name of the HTML file, without the .html extension
Can be set to - to write HTML to STDOUT (note
due to the nature of CLI argument parsing, this must be set using =,
e.g. --name=-).
If either this flag or --output-html-file is set to -
then the HTML will go to STDOUT.
Default: flamegraph
Set the title to display in the flamegraph UI.
Default: the command that 0x ran to start the process
Specify artifact output directory. This can be specified in template
form with possible variables being {pid}, {timestamp}, {name}
(based on the --name flag) and {outputDir}(variables
must be specified without whitespace, e.g. { pid } is not supported).
Default: {pid}.0x
Specify destination of the generated flamegraph HTML file.
This can be specified in template form with possible variables
being {pid}, {timestamp}, {name} (based on the --name flag) and
{outputDir} (variables must be specified without whitespace,
e.g. { pid } is not supported). It can also be set to - to
send the HTML output to STDOUT (note
due to the nature of CLI argument parsing, this must be set using =,
e.g. --output-html=-).
If either this flag or --name is set to -
then the HTML will go to STDOUT.
Default: {outputDir}/{name}.html
Use an OS kernel tracing tool (perf on Linux or dtrace on macOS and SmartOS). This will capture native stack frames (C++ modules and Libuv I/O), but may result in missing stacks on Node 8.
See docs/kernel-tracing.md for more information.
Default: false
Limit output, the only output will be fatal errors or
the path to the flamegraph.html upon successful generation.
Default: false
Suppress all output, except fatal errors.
Default: false
Don't generate the flamegraph, only create the Profile Folder, with relevant outputs.
Default: false
Supply a path to a profile folder to build or rebuild visualization from original stacks.
Default: undefined
Show output from DTrace or perf(1) tools.
Default: false
Save the intermediate tree representation of captured trace output to a JSON file.
Default: false
0x can also be required as a Node module and scripted:
const zeroEks = require('0x')
const path = require('path')
async function capture () {
const opts = {
argv: [path.join(__dirname, 'my-app.js'), '--my-flag', '"value for my flag"'],
workingDir: __dirname
}
try {
const file = await zeroEks(opts)
console.log(`flamegraph in ${file}`)
} catch (e) {
console.error(e)
}
}
capture()
The Programmatic API is detailed in docs/api.md.
Very complex applications with lots of stacks may hit memory issues.
The --stack-size flag can be used to set the memory to the maximum 8GB
in order to work around this when profiling:
node --stack-size=8024 $(which 0x) my-app.js
There may still be a problem opening the flamegraph in Chrome. The same work
around can be used by opening Chrome from the command line (platform dependent)
and nesting the --stack-size flag within the --js-flags flag:
--js-flags="--stack-size 8024".
DEBUG=0x* 0x my-app.js
Sponsored by nearForm
This tool is inspired from various info and code sources and would have taken much longer without the following people and their Open Source/Info Sharing efforts:
MIT
FAQs
🔥 single-command flamegraph profiling 🔥
The npm package @nearform/0x receives a total of 1 weekly downloads. As such, @nearform/0x popularity was classified as not popular.
We found that @nearform/0x demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 33 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Research
A malicious npm package disguised as a WhatsApp client is exploiting authentication flows with a remote kill switch to exfiltrate data and destroy files.
Security News
PyPI now supports digital attestations, enhancing security and trust by allowing package maintainers to verify the authenticity of Python packages.
Security News
GitHub removed 27 malicious pull requests attempting to inject harmful code across multiple open source repositories, in another round of low-effort attacks.