Set up Continuous Test

Setting up a continuous testing service that finds web memory leaks with memlab consists of a few parts:

1. Prepare test scenarios that cover key interactions of the web app under test.
2. Trigger test runs either through memlab command line interface in bash or through memlab API.
3. Collect memory leak results dumped in the disk.

Write Test Scenarios

Follow the instructions in this tutorial.

Run Tests

Run in CLI: Your test job can run the test scenarios via the memlab command line interface, for example:

memlab run --scenario /path/to/test/scenario/file.js \
  --work-dir /path/to/save/memlab/run/results/

Use the --work-dir option to specify a directory where you want memlab to dump browser interaction meta data (e.g., screenshots and heap snapshots).

Run in Node.js: Alternatively, if your test job is written in node.js, you can invoke memlab APIs:

const {run} = require('@memlab/api');
const scenario = require('/path/to/test/scenario/file.js');
const fs = require('fs-extra');

(async function () {
  const workDir = '/path/to/save/memlab/run/results/';
  // make sure the working directory exists
  fs.mkdirsSync(workDir);
  const result = await run({scenario, workDir});
})();

Collect Results

After the memlab run completes, all results and meta data will be saved in the specified working directory, which includes the following sub-directories and files:

/path/to/save/memlab/run/results/
├── data
│   ├── cur
│   │   ├── browser-info.txt    # browser web console output
│   │   ├── console-log.txt     # memlab terminal logging
│   │   ├── leaks.txt           # text summary of clustered memory leaks
│   │   ├── run-meta.json       # meta data of memlab run and browser configuration
│   │   ├── s1.heapsnapshot     # heap snapshot after the url callback (initial page load)
│   │   ├── s2.heapsnapshot     # heap snapshot after the action callback (after target interaction)
│   │   ├── s3.heapsnapshot     # heap snapshot after the back callback (after reverting target interaction)
│   │   ├── screenshot-1.png    # web page screenshot after the url callback (initial page load)
│   │   ├── screenshot-2.png    # web page screenshot after the action callback (after target interaction)
│   │   ├── screenshot-3.png    # web page screenshot after the back callback (after reverting target interaction)
│   │   └── snap-seq.json       # meta data about each browser interaction step
│   ├── logger
│   │   ├── trace-clusters
│   │   │   ├── @1846905.json   # meta data for one of memory leak trace cluster
│   │   │   ...
│   │   └── trace-jsons
│   │       ├── @1846905.json   # detailed JSON trace of a representative memory leak trace
│   │       ...                 # of a leak trace cluster, this can be used for trace visualization
...

To read results, use the built-in result reader or write your own script or library if the test job is running on a different runtime.

const {BrowserInteractionResultReader} = require('@memlab/api');

const workDir = '/path/to/save/memlab/run/results/';
const result = BrowserInteractionResultReader.from(workDir);

// get absolute paths of all snapshot files
const files = result.getSnapshotFiles();

// print all browser web console output
const metaInfo = result.getRunMetaInfo();
console.log(metaInfo.browserInfo._consoleMessages.join('\n'));

// clean up the results
result.cleanup();

For more APIs on the built-in result reader, click here.