Transient exception handling


Keywords
transient, exception, handling, error, try, catch, promise
License
MIT
Install
npm install polly-js@1.8.3

Documentation

polly-js

Greenkeeper badge Transient exception handling for JavaScript made easy.

npm version Bower npm downloads Dependency Status Build Status codecov.io

Polly-js is a library to help developers recover from transient errors using policies like retry or wait and retry.

Polly-js

The typical use case for polly-js are retrying actions after they fail. These actions often include some form of IO which can fail. Examples of these IO actions are AJAX requests to other services, file operations on disk or interacting with a database. All these actions share the characteristics that they can occasionally fail because of circumstances outside of the control of your application like the network connection might be briefly unavailable or the file might be in use by another process. When any of these actions fail there is a change that just retrying the same action, possibly after a short delay, will work. Polly-js makes these retry actions easy to code.

Detecting failures

Depending on the function being executed different ways of detecting failure are used.

When you call polly().execute(<your function>) the code assumes that your code failed and needs to be retried when your function throws an Error.

When you call polly().executeForPromise(<your function>) the code assumes that your code failed and needs to be retried when your function returns a Promise that is rejected.

When you call polly().executeForNode(<your function that accepts a callback>) the code assumes that your code failed and needs to be retried when your function calls the callback with a first non null parameter indicating an error.

Deciding what to do on failure

Whenever a failure is detected Polly-js will attempt to retry your action.

You get to control how a failure is retried by calling either polly().retry() or polly().waitAndRetry(). Both will retry the failing action but the polly().waitAndRetry() option adds a small delay before retrying. In general polly().waitAndRetry() is probably the more appropriate policy to use as retrying without apause could cause a lot requests. By default both will retry once where polly().waitAndRetry() waits 100 ms before retrying. With either function you can specify a number of retries to attempt on repeated failures. With polly().waitAndRetry() it will double the time between each try. If you want to control the exact delays you can also specify and array with the individual delay values so polly().waitAndRetry(5) is equivalent to polly().waitAndRetry([100, 200, 400, 800, 1600]).

##Deciding what failures to retry Using polly().handle(<function>) you can decide if you want to retry a specific failure. For example you might want to retry an AJAX request returning a 404 response code but not one resulting in a 500 response code. The callback function specified in polly().handle() is called with watever error was received so you can use any property from there and you return true if you want to retry the action or false to stop retrying.

Usage

Try to load the Google home page and retry twice if it fails.

polly()
    .retry(2)
    .executeForPromise(function () {
        return requestPromise('http://www.google.com');
    })
    .then(function(result) {
        console.log(result)
    }, function(err) {
        console.error('Failed trying three times', err)
    });

Try to read a file from disk and retry twice if this fails.

polly()
    .retry(2)
    .executeForNode(function (cb) {
        fs.readFile(path.join(__dirname, './hello.txt'), cb);
    }, function (err, data) {
        if (err) {
            console.error('Failed trying three times', err)
        } else {
            console.log(data)
        }
    });

Only retry 'no such file or directory' errors. Wait 100 ms before retrying.

polly()
    .handle(function(err) {
        return err.code === 'ENOENT';
    })
    .waitAndRetry()
    .executeForNode(function (cb) {
        fs.readFile(path.join(__dirname, './not-there.txt'), cb);
    }, function (err, data) {
        if (err) {
            console.error('Failed trying twice with a 100ms delay', err)
        } else {
            console.log(data)
        }
    });

Use async await with the browsers fetch API.

const loadData = url => {
  return polly()
    .waitAndRetry(2)
    .executeForPromise(async () => {
      const rsp = await fetch(url);
      if (rsp.ok) {
        return rsp.json();
      }
      return Promise.reject(rsp);
    });
};

// Using the function somewhere else:
const movies = await loadData("http://localhost:3000/movies.json");

Logging errors

You can set a logger function to be called every time an error is detected.

polly()
    .logger(function(err){
        console.error(err); //will be hit 2 times
    })
    .retry(2)
    .executeForPromise(function () {
        return requestPromise('http://foo.bar.com');
    })
    .then(function(result) {
        //do some important stuff here
    }, function(err) {
        console.error('Failed trying three times', err);//third error is passed back to the caller
    });

Acknowledgements

The library is based on the Polly NuGet package by Michael Wolfenden

See Also

  • Cockatiel: A more feature rich package that includes polies like Circuit Breaker, Timeout, Bulkhead Isolation, and Fallback.