Getting Started with Dependency Security and NodeJS

Internet security is a topic that receives more attention every day.  If you’re reading this article in early 2018, issues like Meltdown, Specter and the Equifax breach are no doubt fresh in your mind.

Cybersecurity is a massive concern and can seem overwhelming.  Where do you start?  Where do you go?  What do you do if you’re a small application team with limited resources?  How can you better engineer your projects with security in mind?

Tackling the full gestalt of this topic, including OWASP and more is beyond the scope of this article. However, if you’re involved in a small front end team, and if you leverage services like AWS, have gone through the OWASP checklist and are wondering, ‘what now?’, this article is for you (especially if you are developing in NodeJS, which we’ll focus on in this article).

Let’s Talk About Co(de) Dependency:

A tiger and her baby pigs

We’re talking about claiming dependents right?

 

 

 

 

 

 

 

 

 

 

 

One way we can quickly impact the security of our applications is through better dependency management.  Whether you develop in Python, JavaScript, Ruby or even compiled languages like Java and C#, library packages and modules are part of your workflow.

Sure, it’s 3rd party code but that doesn’t mean it can’t have a major impact on your project (everyone should remember the Leftpad ordeal from just a few years ago).

As it turns out, your app’s dependencies can be your app’s greatest vulnerability.

Managing Code you Didn’t Write:

Below, we’ll outline some steps you can take to tackle at least one facet of secure development – dependency management.  In this article, we’ll cover how you can implement a handful of useful tools into a standard NodeJS project to protect, remediate and warn against potential security issues you may introduced into your application via its dependency stack.

Your code has issues

We’ve all had to deal with other people’s problematic code at one time or another

 

 

 

 

 

 

 

 

 

 

Using Dependency-Check:

Even if you’re not using NodeJS, if you are building any project that inherits one or more libraries, you should have some form of dependency checking in place for that project.

For NodeJS apps, Dependency-Check is the easiest, lowest-hanging fruit you can likely reach for to better secure your development process.

Dependency-Check is a command-line tool that can scan your project and warn you of any modules being included in your app’s manifest but not actually being utilized in any functioning code (e.g. modules listed in your package.json file but never ‘required’ in any class within your app).  After all, why import code that you do not require?

Installation is a snap.  From your project directory you can:

Npm install –g dependency-check

dependency-check package.json –unused

If you have any unused packages, Dependency-Check will warn you via a notice such as:

Fail! Modules in package.json not used in code: chai, chai-http, mocha

Armed with a list of any unused packages in hand, you can prune your application before pushing it to production.  This tool can be triggered within your CI process per commit to master or even incorporated into your project’s pre-commit hooks to scan your applications further upstream.

RetireJS:

‘What you require, you must retire’ as the saying goes – at least as it does when it comes to RetireJS.  Retire is a suite of tools that can be used for a variety of application types to help identify dependencies your app has incorporated that have known security vulnerabilities.

Retire is suited for JavaScript based projects in general, not just the NodeJS framework.

For the purpose of this article, and since we are dealing primarily with the command line, we’ll be working with the CLI portion of Retire.

npm install -g retire

then from the root of your project just do:

retire 

If you are scanning a NodeJS project’s dependencies only, you can narrow the scan by using the following arguments:

retire  —nopdepath 

By default, the tool will return something like the following, provided you have vulnerabilities RetireJS can identify:

ICanHandlebarz/test/jquery-1.4.4.min.js 
↳ jquery 1.4.4.min has known vulnerabilities: 
http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2011-4969 
http://research.insecurelabs.org/jquery/test/ 
<a href="http://bugs.jquery.com/ticket/11290">
http://bugs.jquery.com/ticket/11290</a>

Retire scans your app and compares packages used within it with an updated online database of packages/package versions that have been identified by the security world as having exploitable code.

Note the scan output above identified that the version of jquery we have imported via the ICanHandlebarz module has a bug that is exploitable by hackers (note the link to CVE-2011-4969 – the exploit in question).

Armed with this list, we can identify any unnecessary dependencies within our app, as well as any exploits that could be utilized given the code we are currently importing and using.

NSP:

There are a multitude of package scanning utilities for NodeJS apps out on the web. Among them, NSP (Node Security Platform) is arguably the most popular.

Where the previously covered Retire is a dependency scanner suited for JavaScript projects in general, NSP, as the name implies, is specifically designed for Node applications.

The premise is identical: this command line tool can scan your project’s package manifest and identify any commonly known web exploits that can be leveraged given the 3rd party packages you have included in your app.

Utilizing NSP and Retire may sound redundant but, much like a diagnosing a serious condition via medical professional, it’s often worth seeking a second opinion.  It’s also equally easy to get up and running with NSP:

npm install -g nsp

nsp check —reporter &lt;Report type (e.g. HTML)&gt;

Running the above within the root of your node application will generate a report in your chosen output format

Again, wiring this up into a CI job should be straightforward.  You can even perform back-to-back scans using both Retire and NSP.

Snyk:

Yes, this is yet another dependency scanner – and like NSP, another one that specifically NodeJS oriented but with an added twist: Snyk.

Snyk home page

Snyk home page

 

 

 

 

 

 

 

 

 

With Retire and NSP, we can quickly and freely generate a list of vulnerabilities our app is capable of having leveraged against it.  However, what of remediation?  What if you have a massive Node project that you may not be able to patch or collate dependency issues on quickly?  This is where Snyk comes in handy.

Snyk can generate detailed and presentable reports (perfect for team members who may not be elbow deep in your project’s code).  The service also provides other features such as automated email notifications and monitoring for your app’s dependency issues.

Typical Snyk report

Typical Snyk report

 

 

 

 

 

 

 

 

 

 

 

Cost: Now, these features sound great (they also sound expensive).  Snyk is a free service, depending on your app’s size.  For small projects or open source apps, Snyk is essentially free.  For teams with larger needs, you will want to consult their current pricing.

To install and get running with Snyk, first visit https://snyk.io and register for a user account (if you have a private or open source project in github or bitbucket, you’ll want to register your Snyk account with your code management tool’s account.

Next, from your project root in the command line console:

npm install -g snyk

snyk –auth

Read through the console prompt.  Once you receive a success message, your project is now ready to report to your Snyk account.  Next:

Snyk test

Once Snyk is finished you will be presented with a URL will direct you to a report containing all vulnerable dependency information Snyk was able to find regarding your project.

The report is handy in that it not only identifies vulnerabilities and the packages they’re associated with but also steps for remediation (e.g. – update a given package to a specific version, remove it, etc.).

Snyk has a few more tricks up its sleeve as well.  If you wish to dive headlong into securing your application. Simply run:

Snyk wizard

This will rescan your application in an interactive wizard mode that will walk you through each vulnerable package and prompt you for actions to take for each vulnerable package.

Note, it is not recommended to use Snyk’s wizard mode when dealing with large scale applications due to it being very time consuming.

It will look like this:

Snyk monitor output

Snyk monitor output

 

 

 

 

Additionally, you can utilize Snyk’s monitoring feature which, in addition to scanning your application, will send email notifications to you or your team when there are updates to vulnerable packages associated with your project.

Putting it all together in CI:

Of course we’re going to arrange these tools into some form a CI instance.  And why not?  Given we have a series of easy-to-implement command line tools, adding these to part of our project’s build process should be straight forward.

Below is an example of a shell script we added as a build step in a Jenkins job to install and run each of these scan tools as well as output some of their responses to artifacts our job can archive:

#Installs our tools
npm install -g snyk
npm install -g dependency-check
npm install -g retire

#Runs dependency-check and outputs results to a text file
dependency-check ./package.json >> depcheck_results.txt

#Runs retire and outputs results to a text file
retire --nodepath node_modules >> retire_results.txt

#Authenticates use with Snyk
snyk auth $SNYK_TOKEN 

#Runs Snyk's monitor task and outputs results to a test file 
snyk monitor >> snyk_raw_monitor.txt

#Uses printf, grep and cut CLI tools to retrieve the Snyk report URL
#and saves that URL to a text file
printf http: >> snykurl.txt
grep https snyk_raw_monitor.txt | cut -d ":" -f 2 >> snykurl.txt
cat snykurl.txt

#Don't forget the set your post-job task in your CI service to save the above .txt
#files as artifacts for our security check run.

Note – depending on how you stage your application’s build lifecycle in your CI service, you may want to break the above script up into separate scripts within the job or separate jobs entirely. For example, you may want to scan your app using Dependency-Check for every commit but save your scan of your application using NSP or Snyk for only when the application is built nightly or deployed to your staging environment.  In which case, you can divide this script accordingly.

Note on Snyk:

In order to use Snyk in your CI instance, you will need to reference your Snyk authentication key when executing this tool (see the API key reference in the script).

This key can be passed as a static string to the job.  Your auth key for Snyk can be obtained by logging in at Snyk.io and retrieving your API key from the account settings page after clicking on the My Account link from Snyk’s main menu.

Next Steps:

This is only the tip of the iceberg when it comes to security and web application development.  Hopefully, this article has given you some ideas and hits of where to start in writing more secure software.  For more info on how to secure your Node-based apps, here’s some additional reading to check out:

Creating a Realtime Reactive App for a collaborative domain

Sampling is a Bazaarvoice product that allows consumers to join communities and claim a limited amount of free products. In return consumers provide honest & authentic product reviews for the products they sample. Products are released to consumers for reviews at the same time. This causes a rush to claim these products. This is an example of a collaborative domain problem, where many users are trying to act on the same data (as discussed in Eric Evens book Domain-Driven design).

 

Bazaarvoice ran a 2 day hackathon twice a year. Employees are free to use this time to explore any technologies or ideas they are interested in. From our hackathon events Bazaarvoice has developed significant new features and products like our advertising platform and our personalization capabilities. For the Bazaarvoice 2017.2 hackathon, the Belfast team demonstrated a solution to this collaborative domain problem using near real-time state synchronisation.    

Bazaarvoice uses React + Redux for our front end web development. These libraries use the concepts of unidirectional data flows and immutable state management. These mean there is always one source of truth, the store, and there is no confusion about how to mutate the application state. Typically, we use the side effect library redux-thunk to synchronise state between server and client via HTTP API calls. The problem here is that, the synchronisation one way, it is not reactive. The client can tell the server to mutate state, but not vice versa. In a collaborative domain where the data is changing all the time, near-real time synchronisation is critical to ensure a good UX.

To solve this we decided to use Google’s Firebase platform. This solution provided many features that work seamlessly together, such as OAuth authentication, CDN hosting and Realtime DB. One important thing to note about Firebase, it’s a backend as a service, there was no backend code in this project.

The Realtime Database provides a pub/sub model on nodes of the database, this allows clients to be always up-to-date with the latest state. With Firebase Realtime DB there is an important concept not to be overlooked, data can only be accessed by it’s key (point query).

You can think of the database as a cloud-hosted JSON tree. Unlike a SQL database, there are no tables or records. When you add data to the JSON tree, it becomes a node in the existing JSON structure with an associated key (Reference)

Hackathon Goals

  1. Realtime configurable UI
  2. Realtime campaign administration and participation
  3. Live Demo to the whole company for both of the above

Realtime configurable UI

During the hackathon demo we demonstrated updating the app’s style and content via the administration portal, this would allow clients to style the app to suite their branding. These updates were pushed in real time to 50+ client devices from Belfast to Austin (4,608 miles away). 

The code to achieve this state synchronisation across clients was deceptively easy!

Given the nature of react, once a style config update was received, every device just ‘reacted’.

 

Realtime campaign administration and participation

In the demo, we added 40 products to the live campaign. This pushed 40 products to the admin screen and to the 50+ mobile app. Participants were then instructed to claim items.

Admin view

Member view

All members were authenticated via OAuth providers (Facebook, Github or Gmail).

To my surprise the live demo went without a hitch. I’m pleased to add…. my team won the hackathon for the ‘Technical’ category.

Conclusion

Firebase was a pleasure to work with, everything worked as expected and it performed brilliantly in the live demo….even on their free tier. The patterns used in Firebase are a little unconventional for the more orthodox engineers, but if your objective is rapid development, Firebase is unparalleled by any other platform. Firebase Realtime Database produced a great UX for running Sampling campaigns. While Firebase will not be used in the production product, it provided great context for discussions on the benefits and possibilities of realtime data synchronisation.

Some years ago, I would have maintained that web development was the wild west of software engineering; solutions were being developed without discipline and were lacking a solid framework to build on. It just didn’t seem to have any of the characteristics I would associate with good engineering practices. Fast forward to today, we now have a wealth of tools, libraries and techniques that make web development feel sane.

In recent years front-end developers have embraced concepts like unidirectional dataflows, immutability, pure functions (no hidden side-effects), asynchronous coding and concurrency over threading. Im curious to see if these same concepts gain popularity in backend development as Node.js continues to grow as backend language.

Event Stream Modeling

Recently, during a holiday lull, I decided to look at another way of modeling event stream data (for the purposes of anomaly detection).

I’ve dabbled with (simplistic) event stream models before but this time I decided to take a deeper look at Twitter’s anomaly detection algorithm [1], which in turn is based (more or less) on a 1990 paper on seasonal-trend decomposition [2].

To round it all off, and because of my own personal preference for on-line algorithms with minimal storage and processing requirements, I decided to blend it all with my favorite on-line fading-window statistics math.  On-line algorithms operate with minimal state and no history, and are good choices for real-time systems or systems with a surplus of information flowing through them (https://en.wikipedia.org/wiki/Online_algorithm)

The Problem

The high-level overview of the problem is this: we have massive amounts of data in the form of web interaction events flowing into our system (half a billion page views on Black Friday, for example), and it would be nice to know when something goes wrong with those event streams. In order to recognize when something is wrong we have to be able to model the event stream to know what is right.

A typical event stream consists of several signal components:  a long-term trend (slow increase or decrease in activity over a long time scale, e.g. year over year), cyclic activity (regular change in behavior over a smaller time scale, e.g. 24 hours), noise (so much noise), and possible events of interest (which can look a lot like noise).

This blog post looks at extracting the basic trend, cycle, and noise components from a signal.

Note: The signals processed in this experiment are based on real historical data, but are not actual, current event counts.

The Benefit

The big payback of having a workable model of your system is that when you compare actual system behavior against expected behavior, you can quickly identify problems — significant spikes or drops in activity, for example.  These could be from DDOS attacks, broken client code, broken infrastructure, and so forth — all of which are better to detect sooner than later.

Having accurate models can also let you synthesize test data in a more realistic manner.

The Basic Model

Trend

A first approximation of modeling the event stream comes in the form of a simple average across the data, hour-by-hour:

The blue line is some fairly well-behaved sample data (events per hour) and the green lines are averages across different time windows. Note how the fastest 1-day average takes on the shape of the event stream but is phase-shifted; it doesn’t give us anything useful to test against. The slowest 21-day average gives us a decent trend line for the data, so if we wanted to do a rough de-trending of the signal we could subtract this mean from the raw signal.  Twitter reported better results using the median value for de-trending but since I’m not planning on de-trending the signal in this exploration, and median calculations don’t fit into my lightweight on-line philosophy, I’m going to stay with the mean as trend.

Cycle

While the fast 1-day average takes on the shape of the signal, it is phase shifted and is not a good model of the signal’s cyclic nature. The STL seasonal-trend technique(see [2]) models the cyclic component of the signal (“seasonal” in the paper, but our “season” is a 24-hour span) by creating multiple averages of the signal, one per cycle sub-series.  What this means for our event data is that there will be one average for event counts from 1:00am, another average for 2:00am, and so forth, for a total of 24 running averages:

The blue graph in Figure 2a shows the raw data, with red triangles at “hour 0″… these are all averaged together to get the “hour 0” averages in the green graph in Figure 2b.  The green graph is the history of all 24 hourly sub-cycle averages.  It is in-phase with and closely resembles the raw data, which is easier to see in Figure 2c where the signal and cyclic averages are super-imposed.

A side effect of using a moving-window average is that the trend information is automatically incorporated into the cyclic averages, so while we can calculate an overall average, we don’t necessarily need to.

Noise

The noise component is the signal minus the cyclic model and the seasonal trend. Since our online cyclic model incorporates trending already, we can simple subtract the two signals from Figure 2c to give us the noise component, shown in Figure 3:

Some Refinements

Now that we have a basic model in place we can look at some refinements.

Signal Compression

One problem with the basic model is that a huge spike in event counts can unduly distort the underlying cyclic model. A mild version of this can be seen in Figure 4a, where the event spike is clearly reflected in the model after the spike:

By taking a page out of the audio signal compression handbook, we can squeeze our data to better fit within a “standard of deviation”. There are many different ways to compress a signal, but the one illustrated here is arc-tangent compression that limits the signal to PI/2 times the limit factor (in this case, 4 times the standard deviation of the signal trend):

There is still a spike in the signal but its impact on the cyclic model is greatly reduced, so post-event comparisons are made against a more accurate baseline model.

Event Detection

Now we have a signal, a cyclic model of the signal, and some statistical information about the signal such as its standard deviation along the hourly trend or for each sub-cycle.

Throwing it all into a pot and mixing, we choose error boundaries that are a combination of the signal model and the sub-cycle standard deviation and note any point where the raw signal exceeds the expected bounds:

Event Floor

You may note in Figure 5 how the lower bound goes negative.  This is silly – you can’t have negative event counts. Also, for a weak signal where the lower bound tends to be negative, it would be hard to notice when the signal fails completely – as it’s still in bounds.

It can be useful to enforce an event floor so we can detect the failure of a weak signal. This floor is arbitrarily chosen to be 3% of the cyclic model (though it could also be a hard constant):

The End

Playing with data is a ton of fun, so all of the Python code that generated these examples (plus a range of sample data) can be found in GitHub here:

https://github.com/bazaarvoice/event_monitor_test

References

For your reading pleasure:

[1] Automatic Anomaly Detection in the Cloud Via Statistical Learning

[2] STL: A Seasonal-Trend Decomposition Procedure Based on Loess

Telling Your Data To “Back Off!” (or How To Effectively Use Streams)

Foreword

Our Curations engineering team makes heavy use of serverless architecture. While this typically gives us the benefit of reduced costs, flexibility, and rapid development, it also requires us to ensure that our processes will run within the tight memory and lifecycle constraints of serverless instances.

In this article, I will describe an actual case where a scheduled job had started to fail, the discovery of the root cause and the refactor that resolved the issue. I will assume you have at least a rudimentary knowledge of Node JS and Amazon Web Services.

Content Export

Months previously, we built a scheduled job using AWS Lambdas that would kick off an export of our social content every 24 hours. In a nutshell, its purpose was to:

  1. Query for social content documents stored in a MongoDB server.
  2. Transform each document to a row in CSV format.
  3. Write out the CSV rows to a file in S3.

The resulting CSV files were meant for ingestion into an analytics tool to measure web site conversion impact.

It all worked fine for months. Then recently, this job began to fail.

Drowning in Content

For your viewing pleasure, here is an excerpted, condensed section of the main code:

function runExport(query) {
  MongoClient.connect(MONGODB_URL)
    .then((db) => {
      let exportCount = 0;
      db
        .collection('content')
        .find(query)
        .forEach(
          // iterator function to perform on each document
          (content) => {
            // transform content doc to a csv row, then push to S3 writer
            s3.write(csvTransform(content));
            exportCount++;
          },
          // done function, no more documents
          () => {
            db.close();
            s3.close();
            logger.info(`documents exported: ${exportCount}`);
          }
        );
    });
}

Each document from the database query is transformed into a CSV row, then pushed to our S3 writer which buffered all content until an s3.close(). At this point, the CSV data would be flushed out to the S3 destination bucket.

It was obvious from this implementation that heap usage would grow unbounded, as documents from the database would be loaded into resident memory as CSV data. As we aggregated content over the previous months, the data pushed into our export process began to generate frequent “out of heap memory” errors in our logs.

Memory Profile

To gain visibility into the memory usage, I wrote a quick and dirty MemProfiler class whose purpose was to sample memory usage using process.memoryUsage(), collecting the maximum heap used over the process lifetime.

class MemProfiler {
  constructor(collectInterval) {
    this.heapUsed = {
      min : null,
      max : null,
    };
    if (collectInterval) {
      this.start(collectInterval);
    }
  }

  start(collectInterval = 500) {
    this.interval = setInterval(() => this.sample(), collectInterval);
  }

  stop() {
    if (this.interval) {
      clearInterval(this.interval);
    }
  }

  reset() {
    this.stop();
    this.heapUsed = {
      min : null,
      max : null,
    };
  }

  sample() {
    const mem = process.memoryUsage();
    if (this.heapUsed.min === null || mem.heapUsed < this.heapUsed.min) {
      this.heapUsed.min = mem.heapUsed;
    }
    if (this.heapUsed.max === null || mem.heapUsed > this.heapUsed.max) {
      this.heapUsed.max = mem.heapUsed;
    }
  }
}

I then added the MemProfiler to the runExport code:

function runExport(query) {
  memProf.start();
  MongoClient.connect(MONGODB_URL)
    .then((db) => {
      let exportCount = 0;
      db
        .collection('content')
        .find(query)
        .forEach(
          (content) => {
            // transform content doc to a csv row, then push to S3 writer
            s3.write(csvTransform(content));
            exportCount++;
            memProf.sample();
          },
          () => {
            db.close();
            s3.close();
            logger.info(`documents exported: ${exportCount}`);
            console.log('heapUsed:');
            console.log(` max: ${memProf.heapUsed.max/1024/1024} MB`);
            console.log(` min: ${memProf.heapUsed.min/1024/1024} MB`);
          }
        );
    });
}

Running the export against a small set of 16,900 documents yielded this result:

documents exported: 16900
heapUsed:
max: 30.32764434814453 MB
min: 8.863059997558594 MB

Then running against a larger set of 34,000 documents:

documents exported: 34000
heapUsed:
max: 36.204505920410156 MB
min: 8.962921142578125 MB

Testing with additional documents increased the heap memory usage linearly as we expected.

Enter the Stream

The necessary code rewrite had one goal in mind: ensure the memory profile was constant, whether a total of one document was processed or a million. Since this was a batch process, we could trade off memory usage for processing time.

Let’s reiterate the primary operations:

  1. Fetch a document
  2. Transform the document to a CSV row
  3. Write the CSV row to S3

If I may present an analogy, we have widgets moving down the assembly line from one station to the next. The conveyor belt on which the widgets are transported is moving at a constant rate, which means the throughput at each station is constant. We could increase the speed of that conveyor belt, but only as fast as can be handled by the slowest station.

In Node JS, streams are the assembly line stations, and pipes are the conveyor belt. Let’s examine the three primary types of streams:

  1. Readable streams: This is typically an upstream data source that feeds data to a writable or transform stream. In our case, this is the MongoDB find query.
  2. Transform streams: This typically takes upstream data, performs a calculation or transformation operation on it and feeds it out downstream. In our case, this would take a MongoDB document and transform it to a CSV row.
  3. Writable streams: This is typically the terminating point of a data flow. In our case this is the writing of a CSV row to S3.

Readable Stream

Conveniently, the MongoDB Node JS driver’s find() function returns a Cursor object that extends the Node JS Readable stream. How convenient!

const streamContentReader = db.collection('content').find(query);

Transform Stream

All we need to do is extend the Transform stream class, and override the _transform() method to transform the content to a CSV row, push it out the other end, and signal upstream that it is ready for the next content.

class CsvTransformer extends Transform {
  constructor(options) {
    super(options);
  }

  _transform(content, encoding, callback) {
    // write row
    const { _id, authoredAt, client, channel, mediaType, permalink , text, textLanguage } = content;
    const csvRow = `${_id},${authoredAt},${client},${channel},${mediaType},${permalink}${text},${textLanguage}\n`;
    this.push(csvRow);
    // signal upstream that we are ready to take more data
    callback();
  }
}

Writable Stream

Let’s purposely mock the S3 Writer for now. It does nothing here, but realistically, we would buffer up incoming data, then flush out the buffer in chunks over the network for best throughput.

class S3Writer extends Writable {
  constructor(options) {
    super(options);
  }

  _write(data, encoding, callback) {
    // TODO: probably use a fixed size buffer
    // that we write to and then flush out to S3 using
    // multipart data upload
    callback();
  }
}

Pipe ‘Em Up

We now create the three streams:

const streamContentReader = db.collection('content').find(query);
const streamCsvTransformer = new CsvTransformer({ objectMode: true });
const streamS3Writer = new S3Writer();

Streams alone do nothing. What makes them useful is connecting the output from one stream to the input of another. In stream terminology, this is called piping, and is accomplished using the stream’s pipe method:

streamContentReader.pipe(streamCsvTransformer).pipe(streamS3Writer)

Although the native pipe method is perfectly functional, I highly recommend using the pump library. In addition to being more readable by passing in an array of streams to pipe together, we can also invoke then() when the pipe has finished, as well as handle close or error events emitted by the streams:

pump([ streamContentReader, streamCsvTransformer, streamS3Writer ])
  .then(() => {
    console.log(`documents exported: ${exportCount}`);
  })
  .catch((err) => console.error(err));

CSV and S3 Write Streams

I purposely did not implement the S3Writer above, because npm is a treasure trove of solutions! The s3-write-stream library will take care of buffering, using multipart upload and handling retries, so we don’t have to get our hands too dirty. And wouldn’t you know it, there is also the csv-write-stream library that will generate properly escaped CSV rows.

As an exercise, you the reader may want to try using the s3-write-stream and csv-write-stream. Trust me, once you get the hang of streaming, you will enjoy it!

Back Off!

Let’s touch back on the issue of memory usage.

What if a write stream becomes blocked for a period of time? Maybe we are writing data to a third-party service over HTTP and network congestion or service throttling is slowing the write rate. The readable stream will happily pump data in as fast as it can. Won’t that cause increased heap usage as all that data gets backed up and buffered?

The quick answer is “no”. In the implementation of the _write() method of your writable stream, when done with the data chunk (after writing to a file for example), call callback(). This signals to the incoming stream that it is ready to receive the next chunk. You can also provide a highWaterMark to the writable stream constructor to give a specific number of objects to buffer before pausing the incoming stream. This is all handled internally by the Node JS streams implementation. No more unbounded buffering!

For a deep dive into the concept of backpressure control, read this great article on Backpressuring in Streams.

Stream ‘Em If You Got ’em!

Our team is now in the process of utilizing Node JS streams whenever we read in volumes of data, process that data, then write out the results. We are now seeing great gains in both stability and maintainability!

References

Node JS Streams APIhttps://nodejs.org/api/stream.html

Article: Backpressuring in Streams: https://nodejs.org/en/docs/guides/backpressuring-in-streams/

MongoDB Cursor Streamhttp://mongodb.github.io/node-mongodb-native/2.2/api/Cursor.html

Pump Modulehttps://www.npmjs.com/package/pump

CSV Write Stream Module: https://www.npmjs.com/package/csv-write-stream

S3 Write Stream Module: https://www.npmjs.com/package/s3-write-stream

 

Testing for Application Front End Performance with Web Page Test

taco-meter

‘Taco-meter’ – a device used to measure how quickly you can obtain tacos from your current location

If you’ve followed Bazaarvoice’s R&D blog, you’ve probably read some of our posts on web application performance testing with tools like Jmeter here and here.

In this post, we’ll continue our dive into web app performance, this time, focusing on testing front end applications.

API Response Time vs App Usability:

Application UI testing in general can be challenging as the cost of doing so (in terms of both time, labor and money) can become quite expensive.

In our previous posts highlighted above, we touched on testing RESTful APIs – which revolves around sending a set of requests to a target, measuring their response times and checking for errors.

For front-end testing, you could do the same – automate requesting the application’s front-end resources and then measuring their collective response times.  However, there are additional considerations for front end testing that we must account for – such as cross-browser issues, line speed constraints (i.e. broadband vs. a mobile 3G) as well as client side loading and execution (especially important for JavaScript).

Tools like Jmeter can address some of these concerns, though you may be limited in what can be measured.  Plugins for Jmeter that allow Webdriver API access exist but can be difficult to configure, especially in a CI environment (read through these threads from Stack Overflow on getting headless Webdriver to work in Jenkins for example).

There are 3rd party services available that are specialized for UI-focused load tests (e.g. Blazemeter) but these options can sometimes be expensive.  This may be too much for some teams in terms of time and money – especially if you’re looking for a solution for introductory, exploratory testing as opposed to something larger in scale.

Webpagetest.org:

Webpagetest.org is an open source project backed by multiple partners such as Google, Akamai and Fastly, among others.  You can check out the project’s github page here.

Webpagetest.org is a publicly available service.  Its main goal is to provide performance information for web applications, allowing designers and engineers to measure and improve application speed.

Let’s check out Webpagetest.org’s public facing page – note the URL form field.  We can try out the service’s basic reporting features by pasting the URL of a website or application into the field and submitting the form.  In this case, let’s test the performance of Bazaarvoice’s home page.

Before submitting the form however, note some of the additional options made available to us:

  • Browser type
  • Location
  • Number of tests
  • Connection
webpagetest.org form

Webpagetest.org’s initial submission form

This service allows us to configure our tests to simulate user requests from a slow, mobile, 3G device in the US to a desktop client with a fiber optic connection in Singapore (just to name a few).

Go ahead and submit the test form.  Once the test resolves, you’ll first notice the waterfall snapshot Webpagetest.org returns.  This is a reading of the application’s resources (images, JavaScript execution, external API calls, the works).  It should look very familiar if you have debugged a web site using a browser’s developer tools console.

webpagetest waterfall

Webpagetest.org’s waterfall report

This waterfall readout gives us a visual timeline of all the assets our web page must load for it to be active for the end user, in the order they are loaded, while recording the time taken to load and/or execute each element.

From here, we can determine which resources which have the longest load times, the largest physical foot prints and give us a clue as to how we might better optimize the application or web page.

Webpagetest.org’s content breakdown chart

In addition to our waterfall chart, our test results can provide any more data points, including but not limited to:

  • Content breakdown by asset type
  • CPU utilization
  • Bandwidth utilization
  • Screen shots and video capture of page load

Armed with this information from our test execution, you can begin to see where we might decide to make adjustment to our site to increase client-side performance.

Getting Programmatic:

Webpagetest.org’s web UI provides some great, easy-to-use features.  However, we can automated these tests using Webpagetest.org’s API.

In our next couple of steps, we’ll demonstrate how to do just that using Jenkins, some command line tools and Webpagetest.org’s public API, which you can read more about here.

Before Going Forward:

There are a few limitations or considerations regarding using Webpagetest.org and its API:

  • All test results are publicly viewable
  • API usage requires you to submit an email address to generate an API key
  • API requests are limited to 200 test runs per day
  • The maximum number of concurrent test runs allowed is 9

If test result confidentiality or scalability is a major concern, you may want to skip to the end of this article where we discuss using private instances of Webpagetest.org’s service.

Requirements:

 As mentioned above, we need an API key from Webpagetest.org. Just visit https://www.webpagetest.org/getkey.php and submit the form made available there.  You’ll then receive an email containing your API key.

Once you have your API key – lets set up a test using as CI service.  For this, like in our other test tutorials, we’ll be using Jenkins.

For this walkthrough, we’ll use the following:

  • Webpagetest’s public API
  • Jenkins
  • NodeJS (be sure your Jenkins instance has at least 1 server capable of using NodeJS 6 or above)
  • Webpagetest (a node module that provides a command-line wrapper for their API)
  • Jq-cli (a node module that provides JQ – a JSON query tool)
  • Wget-improved (a node implementation of the wget CLI tool – skip this if your Jenkins servers already have this binary installed).

Setting Up the Initial Job:

First, let’s get the Jenkins job up and running and make sure we can use it to reach our primary testing tool.

Do the following:

  • Log into Jenkins
  • Navigate to a project folder you wish your job to reside in
  • Click New Item
  • Give the job a name, select Freestyle Project from the menu and click OK

Now that the job is created, let’s set a few basic parameters:

  • Check ‘Discard old builds’
  • Set the Max # of Builds value to 100
  • Check the ‘Provide Node & npm bin/ folder to PATH’ option then choose your desired Node version (if this option is unavailable, see your system administrator).
jenkins-node

Specifying a NodeJs version for Jenkins to use

Next, let’s create a temporary build step to verify our job can reach Webpagetest.org

  • Click the ‘Add Build Step’ menu button
  • Select ‘Execute shell’
  • In the text box that appears, past in the following Curl command:
    •  curl http://webpagetest.org -s –f 
  • Save changes to your job and click the build option on the main job settings screen

Once the job has executed, click on the build icon and select the console output option.  Review the console output captured by Jenkins.  If you see HTML output captured at the end of the job, congratulations, your Jenkins instance can talk to Webpagetest.org (and its API).

Now we are ready to configure our actual job.

The Task at Hand:  

 The next steps will walk you through creating a test script.  The purpose of this script is as follows:

  • Install a few command line tools in our Jenkins instance when executed
  • Use these tools to execute a test using Webpagetest.org’s API
  • Determine when our test execution is complete
  • Collect and preserve our results in the following manner:
    • HAR (HTTP Archive)The entire test result data in JSON format
    • A PNG of our application’s waterfall
  • Filter our results for a specific element’s statistic we wish to know about

Adding Tools:

Click on the configure link for your job and scroll to the shell execution window.  Remove your curl command and copy the following into the window:

npm install -g webpagetest
npm install -g jq-cli-wrapper
npm install -g wget-improved

If you know that your Jenkins environments comes deployed with wget already installed, you can skip the last npm install command above.

Once this portion of the script executes, we’ll now have a command line tool to execute tests against Webpagetest’s API.  This wrapper makes it somewhat easier to work with the API, reducing the amount of syntax we’ll need for this script.

Next in our script, we can declare this command that will kick off our test:

webpagetest test $SITE -k $APIKEY -r $RUNS &gt; test_exec.json
&lt;span style="font-size: 1rem;"&gt;testId=$(cat test_exec.json | jq .data.testId)
&lt;/span&gt;testId="${testId%\"}"
testId="${testId#\"}"
echo $testId

This script block calls the API, passing it some variables (which we’ll get to in a minute) then saves the results to a JSON file.

Next, we use the command line tools cat and jq to sort our response file, assigning the testId key value from that file to a shell variable.  We then use some string manipulation to get rid of a few extraneous quotation marks in our variable.

Finally, we print the testId variable’s contents to the Jenkins console log.  The testId variable in this case should now be set to a unique string ID assigned to our test case by the API.  All future API requests for this test will require this ID.

Adding Parameters:

Before we go further, let’s address the 3 variables we mentioned above.  Using shell variables in the confines of the Jenkins task will give us a bit more flexibility in how we can execute our tests.  To finish setting this part up, do the following:

  • In the job configuration screen, scroll to and check the ‘This build is parameterized’ option.
  • From the dropdown menu, select String Parameter
  • In the name field, enter SITE
  • For the default field, enter a URL you wish to test (e.g. http://www.bazaarvoice.com)
  • Repeat this process for the APIKEY and RUNS variables
  • Set the RUNS parameter to a value of 1
  • Set the default value for the APIKEY parameter to your Webpagetest.org API key.

Save changes to your job and try running it.  See if it passes.  If you view the console output form the run, there’s probably not much you’ll find useful.  We have more work to do.

site-parameter

Your site parameter should look like this

Your runs parameter should look like this

And your key parameter should looks like this

Waiting in Line:

The next portion of our script is probably the trickiest part of this exercise.  When executing a test with the Webpagetest.org web form, you probably noticed that as your test run executes, your test is placed into a queue.

waiting

Cue the Jeopardy! theme…

Depending current API usage, we won’t be able to tell how long we may be waiting in line for our test execution to finish.  Nor will we know exactly how long a given test will take.

To account for that, next we’ll have to modify our script to poll the API for our test’s status until our test is complete.  Luckily, Webpagetest.org’s API provides a ‘status’ endpoint we can utilize for this.

Add this to your script:

webpagetest status $testId -o status.json
testStatus=$(cat status.json | jq .data.testsCompleted)

Using our API wrapper for webpage test and jq, we’re posting a request to the API’s status endpoint, with our test ID variable passed as an argument.  The response then is saved to the file, status.json.

Next, we use cat and jq to filter the content of status.json, returning the contents of the ‘testsCompleted’ (which is a child of the ‘data’ node) within that file.  This gets assigned to a new variable – testStatus.

Still Waiting in Line:

 

long line

Just like going to the DMV

The response returned when polling the status endpoint contains a key called ‘status’ with possible values set to ‘In progress’ or ‘Completed’.  This seems like the perfect argument we would want to check to monitor our test status.However, in practice, this key-value pair can be set prematurely – returning a value of ‘Completed’ when not necessarily every run within our test set has finished.  This can result in our script attempting to retrieve an unfinished test result – which would be partial or empty.

After some trial and error, it turns out that reading the ‘testsCompleted’ key from the status response is a more accurate read as to the status of our tests.  This key contains an integer value equal to the number of test runs you specified that have completed execution.

Add this to your script:

while [ "$testStatus" != "$RUNS" ]
do
sleep 10
webpagetest status $testId -o status.json
testStatus=$(cat status.json | jq .data.testsCompleted)
done

This loop compares the number of tests completed with the number of runs we specified via our Jenkins string variable, $RUNS.

While those two variables are not equal, we request our test’s status, overwrite the status.json file with our new response, filter that file again with jq and assign our updated ‘testsCompleted’ value to our shell variable.

Note that the 10 second sleep command within the loop is optional.  We added that as to not overload the API with requests every second for our test’s status.

Once the condition is met, we know that the number of tests runs we requested should be the same number completed. We can now move on.

Getting The Big Report:

Once we have passed our while loop, we can assume that our test has completed.  Now we simply need to retrieve our results:

webpagetest har $testId -o results.har

webpagetest results $testId -o results.json

Webpagetest’s API provides multiple options to retrieve result information for any given test.  In this case, we are asking the API to give us the test results via HAR format.

Note – if you wish to deliver your test results in HAR format, you will need a 3rd party application or service to view .har files.  There are several options available including stand-alone applications such as Charles Proxy and Har Viewer.

Additionally, we are going to retrieve the test results from our run using the API’s results endpoint and assign this to a JSON file – results.json.

Now would be a great time to save the changes we’ve made to your test.

Getting the Big Picture:

Next, we want to retrieve the web application’s waterfall image – similar to what was returned when executing our test via the Webpagetest.org web UI.

To do so, add this to your script:

waterfallImage=$(cat results.json | jq '.data.runs."1".firstView'.images.waterfall)
waterfallImage="${waterfallImage%\"}"
waterfallImage="${waterfallImage#\"}"
nwget $waterfallImage -O waterfall.png

Since we’ve already retrieved our test results and saved them to the results.json file, we simply need to filter this file for a key-value pair that contains the URL for where our waterfall snapshot resides online.

In the results.json file, the main parent that contains all test run information is the ‘data’ node.  Within that node, our results are divided up amongst each test run.  Here we will further filter our results.json object based on the 1st test run.

Within that test run’s ‘firstView’ node, we have an images node that contains our waterfall image.

We then assign the value of that node to another shell variable (and then use some string manipulation to trim off some unnecessary leading and ending quotation characters).

Finally, as we installed the nwget module at the beginning of this script, we invoke it, passing the URL of our waterfall image as an argument and the option to output the result to a PNG file.

Archiving Results:

Upon each execution, we wish to save our results files so that we can build a collection of historical data as we continue to build and test our application.  Doing this within Jenkins is easy

Just click on the ‘Add Post Build Action’ button within the Jenkins job configuration menu and select ‘Archive the artifacts’ from the menu.

In the text field provided for this new step, enter ‘*.har, results.json, waterfall.png’ and click save.

Now, when you run your test job – once the script succeeds, it will save each instance of our retrieved HAR, waterfall image and results.json file to each respective Jenkins run.

Once you save your job configuration, click the ‘Build with Parameters’ button and try running it.  Your artifacts should be appended to each job run once it completes.

results

Collecting your results in Jenkins

Like a Key-Value Pair in a JSON Haystack:

Next, we’re going to talk about result filtering.  The results.json provided initially by the API is exhaustive in size.  Go ahead and scroll through that file saved from your last successful test run (you will probably be scrolling for a while).  There’s a mountain of data there but let’s see if we can fish some information out of the pile.

The next JSON-relative trick we’re going to show you is how you can filter the results.json from Webpagetest.org down to a specific object you wish to measure statistics for.

The structure of the results.json lists every web element (from .js libraries to images, to .css files) called during the test and enumerates some statistics about it (e.g. time to load, size, etc.).  What if your goal is to monitor and measure statistics about a very specific element within your application (say, a custom .js library you ship with your app)?

needle-in-a-haystack

That’s one way of finding it…

For that purpose, we’ll use cat and js again to filter our results down to one, specific file/element’s results.  Try adding something like this to your script:

cat results.json | jq '.data.runs."1".firstView.requests[] | select(.url | contains(" <a unique identifying string for your web element> "))' > stats.json

This is similar to how we filtered our results to obtain our waterfall image above.  In this case, each test result from Webpagetest.org contains a JSON array called ‘requests’.  Each item in the array is delineated by the URL for each relative web element.

Our script command above parses the contests of results.json, pulling the ‘results’ array out of the initial test run, then filtering that array based on the ‘url’ key, provided that key matches the string we provided to jq’s ‘contains’ method.

These filtered results are then output to the stats.json file.  This file will contain the specific test result statistics for your specific web element.

Add stats.json to the list of artifacts you wish to archive per test run.  Now save and run your test again.  Note, you may need to experiment with the arguments passed to JQ’s contains method to filter your results based on your specific needs.

Next Steps:

At this point, we should have a Jenkins job that contains a script that will allow us to execute tests against Webpagetest.org’s public API and retrieve and archive test results in a set of files and formats.

This can be handy in of itself but what if you have team or organization members who need access to some or all of this data but do not have access to Jenkins (for security purposes or otherwise)?

Some options to expose this data to other teams could be:

  • Sync your archived elements to a publicly available bucket in Amazon S3
  • Have a 3rd party monitoring service (Datadog, Sumo Logic, etc.) read and report test findings
  • Send out email notifications if some filtered stat within our results exceeds some threshold

There’s quite a few ideas to expand upon for this kind of testing and reporting.  We’ll cover some of these in a future blog post so stay tuned for that.

Private or Testing and Wrap-Up:

If you find this method of performance testing useful but are feeling a bit limited by Webpagetest.org’s restrictions on their public API (or the fact that all test results are made public), it is worth mentioning that if you’re needing something more robust or confidential, you can host your own, private instance of the free, open-source API (see Webpagetest.org’s github project for documentation).

Additionally, Webpagetest.org also has pre-built versions of their app available as Amazon EC2 AMIs for use by anyone with AWS access.  You can find more information on their free AMIs here.

Additionally, here’s the script we’ve put together through this post in its entirety:


npm install -g webpagetest
npm install -g jq-cli-wrapper
npm install -g wget-improved

webpagetest test $SITE -k $APIKEY -r $RUNS &gt; test_exec.json
&lt;span style="font-size: 1rem;"&gt;testId=$(cat test_exec.json | jq .data.testId)
&lt;/span&gt;testId="${testId%\"}"
testId="${testId#\"}"
echo $testId

webpagetest status $testId -o status.json
testStatus=$(cat status.json | jq .data.testsCompleted)

while [ "$testStatus" != "$RUNS" ]
do
sleep 10
webpagetest status $testId -o status.json
testStatus=$(cat status.json | jq .data.testsCompleted)
done

waterfallImage=$(cat results.json | jq '.data.runs."1".firstView'.images.waterfall)
waterfallImage="${waterfallImage%\"}"
waterfallImage="${waterfallImage#\"}"
nwget $waterfallImage -O waterfall.png

cat results.json | jq '.data.runs."1".firstView.requests[] | select(.url | contains(" <a unique identifying string for your web element> "))' > stats.json

We hope this article has been helpful in demonstrating how some free tools and services can be used to quickly stand up performance testing for your web applications.  Please check back with our R&D blog soon for future articles on this and other performance related topics.

 

 

Maintaining Test Data with the “someObject” Test Structure

Language: Scala
TestTool: Scalatest

How did we get here?

When systems become reasonably complex, tests must manage cumbersome amounts of data. A test case that may test a small bit of functionality may start to require large amounts of domain knowledge about the system being tested. This is often done through the mock data used to set up the test. Maintenance of this data becomes cumbersome, monotonous and can feel Sisyphean. To solve these problems we created “someObject”, a modular system that allows us to maintain data in only one location while providing the flexibility to create specific data for our tests.

Let’s Do A Code Time!

To start this post, we’re going to build a system without the “someObject” test structure to provide context for its use. (To skip to the “someObject” structure, jump to here!). Suppose we are building a service that reports on advertising campaigns. We may create a class that describes an advertising campaign and call it `Campaign`.

case class Campaign(id: String,
    name: String,
    client: String,
    startDate: UTCDate,
    endDate:UTCDate,
    deleted: Boolean)

Now we are going to store this campaign in a database, and we need to write some integration tests to make sure this operation is performed properly. A test that confirms a campaign is stored properly might look like this:

it should "properly store a single campaign" in {
    Given("we have a proper campaign")
    val campaignToStore = Campaign(
        id = "someId",
        name = "someName",
        client = "someClient",
        startDate = UTCDate(1955, 10, 6),
        endDate = UTCDate(1956, 10, 6),
        deleted = false)

    And("we store this campaign in the database")
    database.storeCampaign()

    When("we fetch the given campaign")
    val fetchedCampaign:Campaign = database.fetchCampaign()

    Then("All the campaign values were stored properly")
    fetchedCampaign.id shouldBe campaignToStore.id
    fetchedCampaign.name shouldBe campaignToStore.name
    fetchedCampaign.client shouldBe campaignToStore.client
    fetchedCampaign.startDate shouldBe campaignToStore.startDate
    fetchedCampaign.endDate shouldBe campaignToStore.endDate
    fetchedCampaign.deleted shouldBe campaignToStore.deleted
}

BlogIntro.scala

Add some functionality:

Now we add the ability to update some values for this campaign in the database, and we need to test this new piece of functionality. That test might look something like this:

it should "properly update a single campaign" in {
    Given("we have a proper campaign")
    val campaignToStore = Campaign(
        id = "someId",
        name = "someName",
        client = "someClient",
        startDate = UTCDate(1955, 10, 6),
        endDate = UTCDate(1956, 10, 6),
        deleted = false)

    And("some update parameters for our campaign")
    val updateParameters = UpdateParameters(name = Some("someNewName"))

    And("we store this campaign in the database")
    database.storeCampaign(campaignToStore)

    When("we update this campaign")
    database.updateCampaign("someId", updateParameters)

    val fetchedCampaign:Campaign = database.fetchCampaign("someId")

    Then("All the campaign values were stored properly")
    //Unchanged values
    fetchedCampaign.id shouldBe campaignToStore.id
    fetchedCampaign.client shouldBe campaignToStore.client
    fetchedCampaign.startDate shouldBe campaignToStore.startDate
    fetchedCampaign.endDate shouldBe campaignToStore.endDate
    fetchedCampaign.deleted shouldBe campaignToStore.deleted
    //Changed values
    fetchedCampaign.name shouldBe updateParameters.name
}

BlogUpdateFunction.scala

But here we see the duplication of test boilerplate code in `campaignToStore`. We don’t want to have to copy over `campaignToStore` into every test, so we might want to abstract that out to be used all over the suite.

object MyCampaignTestObjects {
    val campaignToStore = Campaign(
        id = "someId",
        name = "someName",
        client = "someClient",
        startDate = UTCDate(1955, 10, 6),
        endDate = UTCDate(1956, 10, 6),
        deleted = false)
}

BlogAbstractToSuite.scala

Now we can use the same data in every test!

Add a Test that requires unique data:

Suppose we now write a function to fetch all the campaigns that are stored in the database. We might need a test that involves uniqueness in the data we store, such as the following example:

it should "properly fetch all stored campaigns" in {
    Given("we store several unique campaigns")
    val anotherCampaignToStore = Campaign(
        id = "someSecondId",
        name = "someSecondName",
        client = "someSecondClient",
        startDate = UTCDate(1955, 10, 6),
        endDate = UTCDate(1956, 10, 6),
        deleted = false)
    database.storeCampaign(campaignToStore)
    database.storeCampaign(anotherCampaignToStore)

    When("we fetch all campaigns")
    val allCampaigns = database.fetchAllCampaigns()

    Then("all campaigns are returned")
    allCampaigns shouldBe List(campaignToStore, anotherCampaignToStore)
}

BlogTestWithUniqueTestData.scala

In the example, we re-used the campaign we abstracted out earlier for conciseness, but this makes this test unclear that `anotherCampaignToStore` is unique. What if someone else comes in and changes `campaignToStore` and it happens to match data from `anotherCampaignToStore`? This test would then become flakey and nobody likes flakey tests. We might decide to just make all data used in this test local to this test, but then we will need to maintain the test data in both this test, and `MyCampaignTestObjects`.

Add Some Arbitrary Data Constraints:

Suppose now that there is a new design constraint on how campaigns can be stored in the database. Now all client names must be lowercased in all campaigns:

object MyCampaignTestObjects {
    val campaignToStore = Campaign(
    id = "someId",
    name = "someName",
    //We change the client name to match our new requirement
    client = "some_client",
    startDate = UTCDate(1955, 10, 6),
    endDate = UTCDate(1956, 10, 6),
    deleted = false)
}

BlogNewConstraint.scala

Now we start to see the issue with maintaining test data across the whole suite that we’ve been constructing. We need to find every mock campaign that is used in our suite and ensure that its client field data is lowercased. Realistically, many of our tests, (specifically in this example, the `fetchAllCampaigns` test) don’t care about the client field of their campaign, and so we shouldn’t need to care about the client field value while setting up our mock test data. Because this example is small, it’s not cumbersome to directly update the value to satisfy the new constraint. Now let us Imagine a large set of suites, each containing hundreds of unique test cases. Suddenly this single suite requires a large amount of work to refactor one field across each test case. Nobody wants to do that monotonous maintenance. To address this, our team adopted the “someObject” structure to minimize this data maintenance within our tests.

someObject Test Structure:

When designing this test structure, we wanted to make our test data extendable for use anywhere it is needed. We used Scala’s `trait` to mix in necessary functions to provide test objects to the objects inside our tests, such as the `MyCampaignTestObjects` object above:

trait CampaignTestObjects {
    def someCampaign(id: String = "someId",
                     name: String = "someName",
                     client: String = "some_client",
                     startDate: UTCDate = UTCDate(1955, 10, 6),
                     endDate: UTCDate = UTCDate(1956, 10, 6),
                     deleted: Boolean = false): Campaign =
        Campaign(
            id = id,
            name = name,
            client = client,
            startDate = startDate,
            endDate = endDate,
            deleted = deleted)
}
 object MyCampaignTestObjects extends CampaignTestObjects {
    //Any other setup methods for test data
}

Now we can revisit our `fetchAllCampaigns` test example.

it should "properly fetch all stored campaigns" in {
    Given("we store several unique campaigns")
    val campaignToStore = someCampaign(id = "someId")
    val anotherCampaignToStore = someCampaign(id = "someNewId")
    database.storeCampaign(campaignToStore)
    database.storeCampaign(anotherCampaignToStore)

    When("we fetch all campaigns")
    val allCampaigns = database.fetchAllCampaigns()

    Then("all campaigns are returned")
    allCampaigns shouldBe List(campaignToStore, anotherCampaignToStore)
}

BlogBasicSomeObject.scala

Inside this test, we’ve set up two unique campaigns, by calling the `someCampaign` method from our test data trait. This populates the returned campaign with dummy data that we don’t care about. All we need out of this method is “some campaign” with “some data”. Now, instead of obscuring the intent of the test case by setting up cumbersome, overly-expressive mock data, we can simply override the implicitly available mock objects with only the necessary data. For the unique campaigns needed in the `fetchAllCampaigns` test, we only only really care about each campaign’s identifier. We don’t update the name, client, startDate, etc. because this test doesn’t care about any of that data. We only need to care that the campaigns are unique for our database. Under this test structure, when we receive the design change about the client names being lowercased, we don’t need to update our `fetchAllCampaigns`.

Another Example:

Let’s provide another example that our team encountered. Campaigns inside our database now need to also store the amount of money spent on each ad campaign. We’re adding a new column to our database, and changing the database schema.

case class Campaign(id: String,
    name: String,
    client: String,
    totalAdSpend: Int,
    startDate: UTCDate,
    endDate: UTCDate,
    deleted: Boolean)

Now, every test that has a campaign involved needs to be updated to include a new field; but under the “someObject” structure we only need to add two lines and all existing tests should be working fine again:

trait CampaignTestObjects {

    def someCampaign(id: String = "someId",
                     name: String = "someName",
                     client: String = "some_client",
                     totalAdSpend: Int = 123456,
                     startDate: UTCDate = UTCDate(1955, 10, 6),
                     endDate: UTCDate = UTCDate(1956, 10, 6),
                     deleted: Boolean = false): Campaign =
        Campaign(
            id = id,
            name = name,
            client = client,
            totalAdSpend = totalAdSpend,
            startDate = startDate,
            endDate = endDate,
            deleted = deleted)
}

BlogPostSchemaChange.scala

 

Behavior Driven Tests:

The purpose of the “someObject” structure is to minimize data maintenance within tests. We want to ensure that we’re disciplined about only setting data that the tests need to care about. There are cases where data might seem necessary for what we are testing, but we can abstract that data away to de-couple the test’s reliance on hard coded values. For example, suppose we have a function that returns the sum of all the `totalAdSpend` across our database.

def sumAllSpend(campaignsToSum: List[Campaign]):Int = 
    campaignsToSum.reduce(_.totalAdSpend + _.totalAdSpend)

To test this function, we might write a test like this:

it should "properly sum all totalAdSpend" in {
    Given("we store several unique campaigns")
    val campaignToStore = someCampaign(id = "someId", totalAdSpend = 123)
    val anotherCampaignToStore = someCampaign(id = "someNewId", totalAdSpend = 456)
    database.storeCampaign(campaignToStore)
    database.storeCampaign(anotherCampaignToStore)

    When("we sum all ad spend")
    val totalTotalAdSpend = sumAllSpend(database.fetchAllCampaigns())

    Then("the result is the sum of all ad spend")
    totalTotalAdSpend shouldBe 975
}

BlogPostNonBehaviorTest.scala

While this test does work, and it utilizes this “someObject” structure, it still forces data management at the test level.

`sumAllSpend` doesn’t really care about any one campaign’s `totalAdSpend` value. It only cares that we add all of the `totalAdSpend` values up correctly. We could instead write our test to assert on this behavior instead of doing the math ourselves and taking on the responsibility of managing more data.

it should "properly sum all totalAdSpend" in {
   Given("we store several unique campaigns")
   val campaignToStore = someCampaign(id = "someId")
   val anotherCampaignToStore = someCampaign(id = "someNewId")

   val allCampaignsToStore = List(campaignToStore,anotherCampaignToStore)
   allCampaignsToStore.foreach(campaign =&amp;gt; database.storeCampaign(campaign))

   When("we sum all ad spend")
   val totalTotalAdSpend = sumAllSpend(database.fetchAllCampaigns())

   Then("the result is the sum of all ad spend")
   totalTotalAdSpend shouldBe allCampaignsToStore.reduce(_.totalAdSpend + _.totalAdSpend)
}

BlogPostBehaviorTest.scala

With this test, we don’t care what campaigns sales were, we don’t care how many campaigns are stored, and we don’t care about any constant value. This test will return the sum of all campaign’s `totalAdSpend` value that we store in the database.

Conclusion:

In this introductory blog post, we explored the someObject testing structure in scala, but this concept is not intended to be language specific. Scala makes this concept easy to implement through the use of Default Parameter Values but in a future post I’ll show how it can be implemented through the Builder Pattern in a language like Java. Another unexplored “someObject” concept is the granularity of control in setting default data. This post introduces the “global” and test specific setting of default data, but doesn’t explore how to set test suite level data for our test objects, and the cases in which that might be useful. I’ll discuss that the future post as well.

Code snippets:

BlogIntro.scala
BlogUpdateFunction.scala
BlogAbstractToSuite.scala
BlogTestWithUniqueTestData.scala
BlogNewConstraint.scala
BlogBasicSomeObject.scala
BlogPostSchemaChange.scala
BlogPostNonBehaviorTest.scala
BlogPostBehaviorTest.scala

Publishing Load Test Results with Jmeter, Jenkins and S3

A while ago, I published a blog post that presented a tutorial overview of how to use Jmeter for load testing a typical RESTful API.

This post builds upon that original post with handy information on some updated reporting features of Jmeter as well a quick dive into how you can better propagate your load results in a continuous build environment (ala Jenkins).

if you’re completely new to Jmeter, please read my previous blog post, linked above before proceeding.

For those of you experienced with Jmeter (hopefully you already have some tests you’re wanting to deploy somewhere rather than let them live contentedly on your local machine), get ready because you’re gonna love this

67295777

It’s truly amazing the gifs you can find on the web.

Meet the New Jmeter, Same as the Old Jmeter (not really)

The previous load test tutorial covered setting up a load test from scratch using an early version of Jmeter.  One of the benefits of using Jmeter, as outlined in the original post was that it has been around for quite a long time, is well known, stable and free.  However, one of the major cons in using Jmeter is that its’ well, old – and has gone for a long while without major updates – especially when it comes to report generation.

Well, welcome to Jmeter 3!  It still comes with that same, great, vintage 90s UI but now with a hoard a bug fixes, stability upgrades and – wait for it – native HTML report generation, wooo!  Check out version 3’s documentation here:

Installation and Setup

So, let’s get down to it!  You can find the latest build of Jmeter 3 here: https://jmeter.apache.org/download_jmeter.cgi

Steps to setting up Jmeter 3 locally are similar to the installation steps covered in my original blog post.  Pre-installation requirements are the same:

  • Java JDK (version 7 or later)
  • Be sure your Java HOME local environment or path variables are set before execution.

Functionally, Jmeter 3 appears identical to older versions of the application but there are a lot of changes under the hood.  I recommend unarchiving and installing Jmeter 3 in its own directory, separate from any previous version of Jmeter you may have previously installed.

Migrating your tests

If you’re following along with this tutorial and wish to create a new load test from scratch, please follow the test setup steps in the original post’s tutorial however, be sure to SKIP the 3rd party report and listener portion of the test setup (guess what, you won’t be needing those plugins anymore).

If you have an existing test you wish to simply fire up in Jmeter 3, your mileage will vary in this case.  Some of the 3rd party listeners and other tools you may have used in your setup may be incompatible with the newer version of Jmeter.  If you have any custom listeners embedded in your test I recommend doing the following:

  • Open your original test in your older version of Jmeter
  • Remove 3rd party listeners, samplers or elements from your test (right-click and select remove)
  • Save the updated test as a new test
  • Open the newly created test in Jmeter 3

From this point, you won’t really need any of your original report listeners configured in your test, however, for debugging purposes, I do recommend keeping a simple response listener in your test – one for successful API responses, another separate for failed responses – just to help with debugging any issues your test might run into.

In a few cases, I’ve found despite removing any 3rd party plugins that some tests have compatibility issues with Jmeter 3.  Sadly, the easiest course of action is to rebuild the test from scratch.  Such is life (sigh).

Such is life...

Such is life…

Executing Tests with Report Generation

Given you have your initial test configured for Jmeter 3, let’s start up the app and view it in the new UI (though it looks like the old UI).  Once you start Jmeter froim the command line, you’ll see a message like this in your terminal:

================================================================================
Don't use GUI mode for load testing, only for Test creation and Test debugging !
For load testing, use NON GUI Mode & adapt Java Heap to your test requirements
================================================================================

You don’t say!?  If you’re a long-time user of earlier version of Jmeter, you’re probably thinking, “Well, no duh!” right?

Open your load test in the UI and give it a look-over to make sure its configured the way you intend.  It should be minimal simple thread group, your HTTP samplers, your results tree report and any CSV data resources you might need.

Your test may look something like this

Your test may look something like this

Next, save your test and close Jmeter (we’re now done with the Jmeter UI).

Now, we are going to use Jmeter’s command line interface to execute the load test and generate the report for the test.

Do the following:

Open your CLI terminal and navigate to your Jmeter 3 bin directory.  Execute the following command:

jmeter -n -t <test JMX file> -l <test log file> -e -o <Path to output folder>

Be sure to set the ‘test JMX file’ argument to the path where your load test’s .JMX file resides, relative to the Jmeter bin directory.

The test log file should be an arbitrary log file path/name that currently does not exist.

The output folder path is where your report will be written to.  This must be an empty directory.

A couple of notes:

If you configured your test to run for a set period, just sit back and wait.  The report will be generated in the output directory you provided once the test execution finishes.

If you configured it to run in perpetuity – you will need to formally stop the test before report generation will execute (simply killing your terminal session won’t help you here).

To stop your Jmeter test from the command line, simply open a new terminal window and enter one of the following commands:

Control + . – This stops Jmeter, halting any active listeners.

Control + ,  – This sends a shutdown command to Jmeter – it will wait until all active threads have finished their current tasks before closing shop (this is the recommended method to use unless you’re impatient

or there is an emergency – like, oh you forgot you pointed your test at production).

Are you really going to argue with Chuck?

Are you really going to argue with Chuck?

Once your test has gracefully exited, navigate to the output folder you specified and open your newly minted report in your web browser of choice.

Jmeter 3 HTML report

Jmeter 3 HTML report

Pretty sweet huh?
sxf8ww

If you haven’t seen Jmeter’s dashboard report before, you’re probably thinking, “golly, there’s a lot of stuff here” – and there is.  The dashboard report contains a pretty wide variety of data on your test execution.  Enumerating through all of that here would be exhausting however – if you’re interested in all the gory details of Jmeter’s dashboard metrics, check out this in-depth explanation here:

https://jmeter.apache.org/usermanual/generating-dashboard.html

Show and Tell (mostly show)

So now you have this fancy load test report but hoarding all this metric goodness on your local development workstation is no fun for the rest of your team.  But that’s OK because if you have access to a continuous integration build environment (ala – Jenkins), the rest of this blog post will walk you through taking your Jmeter test and throwing it into a Jenkins build that allows you to not only run your test remotely but also save and propagate your results to the rest of your agile team.

Make your test available

First things first, you’re going to need to take your Jmeter test, get it off your local machine and into some form of web-accessible delivery.  If you have AWS access, you could place your .JMX files and resources into a bucket and be done with it.  In our case, I recommend creating a private Github repository for your tests and committing your Jmeter working directory, tests and resource files to that repository.  There’s a couple of reasons for this:

  • It’s easy to access a given Github repository with Jenkins
  • You can manage updates and versions of your tests just like source code
  • You can manage variants of your load tests using Github’s branch management features

That said, before tossing your tests into Github’s gaping maw, here’s an important note regarding information security:

Depending on your type of load test, you may have some API or other service access keys stored in your test.  You don’t want to check in API or other access keys into your repository (even if it is marked private).  But don’t just take my word for it:

https://softwareengineering.stackexchange.com/questions/205606/strategy-for-keeping-secret-info-such-as-api-keys-out-of-source-control

If you’ve already committed your tests to your repository and you still have some software keys in there and are thinking, “Oh boy… What now!?”, here’s a link to some handy notes on how to extract some sensitive values from your test or code using some handy features of Git:

https://gist.github.com/derzorngottes/3b57edc1f996dddcab25

Getting Up and Running with Jenkins

Jenkins is one of the most widely available CI build tools in use today.  As such, we’re going to cover how to set up a Jenkins task to run your load tests.  If you’re using some other build system (e.g. Team City) this walk-through won’t be as pertinent but essentially, the tasks will be the same.

Step 1: Get access to Jenkins:

Assuming you already have a Jenkins build server of some sort available to your organization – let’s log into it.  If you don’t have access to Jenkins, talk to your friendly neighborhood dev ops engineer.  Note, that for this walkthrough, you’re going to need some level of administrator access to set up everything we need to execute our tests.

Step 2: Create and Setup your Job with Parameters:

Do the following:

  • From the Jenkins home page, choose a directory you wish to run your load test jobs from (or create one).
  • Click the “New Item” link and then choose “Freestyle Project”
  • Give your project a name and description
  • Choose the “This project is parameterized” option – we’ll use parameters to allow you to configure your load test If needed.
  • Select the “New string parameter” option and create a new parameter named, “thread”
  • Set the default value to whatever you want your test’s default number of threads to be (e.g. 40).
  • Create another string parameter and name it “ramp” – set its default value to the number of seconds you wish your test to take to ramp up to its maximum load.

If your test is going to loop, create another string parameter named, “loops” and set its default, numeric value.

If you have an API key (or keys) you need for your test to execute, as mentioned above, and have already removed said keys from your test (substituting them for a placeholder value) you can create additional string parameters here (naming them key1, key2, etc.).  You can then set your API key(s) to the default value field.  Granted, you are propagating your API key here but this is a much better practice in terms of security than leaving them in a potentially publicly accessible repository as your Jenkins instance should be private and protected.

5dbe72d3d6a5db0d23a62eea640d7c5a_1457280669769jpg-keymaster-gatekeeper-meme_250-250

Step 3: Git Jenkins to Talk to Git:

While still in the Jenkins job configuration screen, under the Source Code Management section, select “Git” and enter your repository’s URL alias into the Repository URL field.

If this is your first time doing this, Jenkins expects a specific format for your Github URL strings.  It should look something like this:

https://github.com:<your main repository name>/<your project name>

If you’re using a publicly accessible Github repository, skip the next step, enter your preferred branch name (or */master if you don’t need a specific branch).

Next, select the appropriate credentials option from the relative drop down menu to allow Jenkins to properly introduce itself to your Github repo (if you’re having trouble with this, ask your dev ops engineer or Jenkins server admin for help – and maybe do something nice for them while you’re at it like get them a thank you card or some cookies or something, because at this point, you’d be out of luck without them, right?).

bp4

Step 5: Build Settings:

You can kick off your tests manually however, it’s probably better to set up timer or other method to handle this for you.  Try the following:

Under the Build Triggers heading, check the Build periodically option

Enter your time range you wish the build to trigger at (this will be in Cron notation – which is tricky if you’re seeing this for the first time).

bp5

This article does a decent job explain Cron’s formatting: http://www.nncron.ru/help/EN/working/cron-format.htm

Now, click on the Add build step option and from the dropdown menu, select ‘Execute shell’.

Here, we will define the command we will use to execute the Jmeter test once Jenkins pulls the test resources from Github. This will be like the shell command we used to run Jmeter locally as described above.

bp7

Syncing to Amazon Web Services

Now we need to make sure we’re saving our generated report somewhere so the rest of our team can review it.  Jenkins has a Jmeter performance plugin that can archive our results.  Also, we could just have Jenkins archive the output directory Jmeter generates the report in (though handing all the recursive directories it will build might be a pain).

Given you have an AWS account, you have the options to sync your reports to an AWS bucket and host them like a static web page.  This method would be beneficial if you need to expose your load test results to team members or other parties that do not or should not have access to your build systems (e.g. product manager, support team members, etc.).

If you don’t have access to AWS, aren’t sure you have access to AWS or may have issues with your build server communicating to AWS, once again, now is a fine time to make friends with (*cough* bribe *cough*) your friendly neighborhood dev ops engineer.

Let’s say you’ve taken care of that and you and your dev ops guru are best buddies now.  Next, you’ll need to further configure your Jenkins job by adding another Execute shell build step.  Make sure this one is ordered sequentially after your Jmeter test execution step.

In the Execution shell, simply add a command to change to the output directory your report will be in (defined in your test execution command) and use the AWS sync command to send your HTML report to AWS:

cd bin/target
 aws s3 sync . s3://mybucketroot/mybucketname

Save your Jenkins job.  You’re done (with Jenkins at least).

Now, in the above step, we are assuming you have a bucket already created in AWS where you can sync your data to.  You’re also going to need access to modify how the bucket works in terms of its permissions.

Making Your Test Results Readable from the Web:

By default, your AWS bucket is out there on the web, you can read from it, write from it, but your other team members will have trouble reading the HTML report from that location.  We need to configure the bucket so that it will display its contents like a static web page.  Here’s how:

Log into your AWS account and access your specified bucket.

Click on the bucket and select the Properties option

Click the Static website hosting panel

bp9

Enable the static hosting option

Specify your index.html file location (this should be the relative path to the main HTML file for your uploaded report).

bp10

Save your changes and boom – now you’re done.

Save and run your job.  If your account permissions are all set up correctly, your test should execute and the test results will be uploaded to the specified AWS bucket.

From here, simply access your AWS bucket using the following styled web link:

https://s3.amazonaws.com/myBucketRoot/myBucketName

995f09e2e0291e1c61e3d22a62438b95f23395d966984d472eb5b43ea04f76beYour test results should now be viewable just as if they were a static web page by anyone on your team.

Wrapping it Up:

Now that you have your load tests residing in a Github repository, feel free to leverage Github’s branching features to allow you to manage multiple iterations of your load tests:

Commit different variations of the same branch to new branches

Use Jenkin’s copy feature to copy your existing job, and just point it to your different branches to allow you to execute multiple test types.

Use Jmeter’s UI locally to copy/edit your existing tests to make additional iterations.  Save these to your repo and commit them just like you would any other coding project.

Again, use Jenkin’s copy feature to create variations of your test run jobs – this time, just edit your shell command to execute your desired .JMX test file per a given execution.

Thanks for reading and I hope this give you further ideas and insight into how you can leverage Jmeter to further improve your software projects through load testing.

Database Migration

MalcolmInTheMiddleGif

(Always One More Thing…)

Who Are We?

The Ad Management team here at Bazaarvoice grew out of an incubator team. The goal of our incubator is to quickly iterate on ideas, producing prototypes and “proof of concept” projects to be iterated on if they validate a customer need. The project of interest here generates reports based on aggregations of event data gathered from several other teams at the company. As our project gained traction, it grew in size and scope, eventually leading to the need to revisit some of the design decisions made in the prototyping phase. Specifically, we found the original database system we chose, EmoDB, to not meet our needs as our requirements evolved.

Why Migrate?

When this project was started, it began as a prototype designed to get the project rolling as quickly and easily as possible. The initial team chose EmoDB since they were familiar with the in-house technology from their other projects and it fit our initial needs. As the project gained traction, and we had more data to operate against, we encountered scalability issues, initially resolved with caching and some refactoring. We found that we were querying EmoDB as if it were a typical relational database, when it’s not actually designed for that use case. (Emodb is an eventually consistent json blob store with a change notification databus that spans multiple AWS AZs and Regions. EmoDB powers many of our solutions at Bazaarvoice and is now open-source and available at:
https://github.com/bazaarvoice/emodb

We chose to switch to MySql to leverage the Relational Data Model for rolling up aggregations of data we collect and calculate. We ran into problems previously when we retrieved whole documents to perform aggregations on our data, leading us to decide that a technology that is optimized for relational models would suit the project much better.

How to Migrate?

Since our project already had trained users by the time we wanted to migrate database systems, we needed to design our migration with a no-down-time approach; “seamlessly” changing out the back-end implementation for our users. We also made these transitions configurable, such that we wouldn’t need to make one large switch to master from the new system, but we could choose which services were ready to be cut over to the new data back-end.
The following image is our design document that describes how we planned our migration. On the left is our origin code base named “legacy”. On the right was the proposed design for our new service stack for the migration. Inserted into the middle is the “Service Facade” where we intended to run our quality assurance against live data between the legacy technology stack and our new technology stack.

Copy of ad-management migration to MySql based stack - Page 1

How to Maintain Data Consistency?

Depending on the size of the data that is being diffed and migrated between databases, it can be expensive to run the necessary migrations. Our solution was both writing specific tasks that backfilled data or directly migrating data sets to the new data source. This allowed us to smoke test that our services are working, without expending large amounts of time or money finding bugs along the way. As our confidence grew in our custom tooling and services, we would backfill and migrate larger chunks of data, until we had migrated everything necessary to master from our new service.

What is a Service Facade?

The service facade layer is responsible for executing the respective operations out of the legacy and new stacks. This is where we placed our diffing logic to compare the results returned from Emo and Mysql for the same operation. The facade returns data from the pre-defined configured stack. This meant that certain areas of the application could be sourcing from Mysql, while other areas, that we weren’t confident in, continued to source from Emo. For example, our CampaignRoiReportBuilderServiceFacade written in Scala looked something like this:


class CampaignRoiReportBuilderServiceFacade @Inject()(
private val campaignRoiReportBuilderServiceLegacy:CampaignRoiReportBuilderServiceLegacy,
private val campaignRoiReportService:CampaignRoiReportService,
private val campaignConfig: CampaignConfiguration,
private val facadeDiffTool: FacadeDiffTool) {
...
  def buildReport(...): Future[Option[CampaignRoiReport]] = {
    val roiReportFromEmoDbFuture: Future[Option[CampaignRoiReport]] = campaignRoiReportBuilderServiceLegacy.buildReport(...)
    val roiReportFromMySqlFuture: Future[Option[CampaignRoiReport]] = campaignRoiReportService.buildReport(...)

    // Extract data from scala futures
    for {
      roiReportFromEmoDbMaybe &lt;- roiReportFromEmoDbFuture
      roiReportFromMySqlMaybe &lt;- roiReportFromMySqlFuture } { // Pattern Matching to extract values from scala Option (roiReportFromEmoDbMaybe, roiReportFromMySqlMaybe) match { case (None, None) =&gt; //this is an impossible case, but listed to avoid compilation warning
        case (Some(_), None) =&gt; LOG.warn("/*Report missing/mismatched data*/")
        case (None, Some(_)) =&gt; LOG.warn("/*Report missing/mismatched data*/")
        case (Some(roiReportFromEmo),Some(roiReportFromMySql)) =&gt;
          val mismatches = facadeDiffTool.campaignROIReportLegacyDiff(roiReportFromEmo,roiReportFromMySql)
          if(mismatches.nonEmpty)
            LOG.warn("/*Report missing/mismatched data*/")
      }
    }

//This is how we configure what source we return to the resource.
    campaignConfig.masteringFrom match {
      case EmoDb =&gt;
        roiReportFromMySqlFuture.onFailure{
          case e:Throwable =&gt; LOG.warn("Failed to build an ROI report on the MySql side", e)
        }
        roiReportFromEmoDbFuture
      case MySql =&gt;
        roiReportFromEmoDbFuture.onFailure{
          case e:Throwable =&gt; LOG.warn("Failed to build an ROI report on the EmoDb side", e)
        }
        roiReportFromMySqlFuture
    }
  }
}

The original resource classes will be modified to call from the new facade layer, but no other functionality should change. If constructed properly, the facade layer will act in the same way as the original service because the facade mimics the public functions available in the original service class. These duplicated functions will call to the methods from the legacy service class as well as the new service. With the responses from both the legacy and new services, the facade layer can make an assessment on the differences between the two service stacks. To report our differences such that we could be notified during API usage, we would log them out to our log management and monitoring system.

How Did We Capture Mismatches?

Logging was a large concern of ours. We knew that there would be many differences per call while we were debugging our new service stack. As an example, on one call, we reported 2000+ differences. We wanted to compose all differences into one log per call in a meaningful way. For this, we wrote custom diff tooling that would return differences in the data as sets of MismatchedField classes.


case class MismatchedField[T](name: String,
                              legacyValue: T,
                              newValue: T)


This templated class will hold the values returned from both the legacy service (legacyValue), and the new stack’s service (newValue), as well as some meaningful tag with which to identify where this mismatch came from (name). We would then compose all mismatches for any given call into a single log through our custom diff tool. Every function within our custom diff tool returns Set[MismatchedField[Any]]. We can then compose each set into a single set of differences such that we can use only one log call to write out the whole set of differences in one log entry.

An Interesting Finding:

One of the most interesting findings we had through this migration weren’t bugs that came in constructing our new service stack for the new database, but that we found bugs in our original database stack. One take-away from this was to make sure to investigate any mismatches found down to the source data. We found during the code migration process that some of our legacy functionality was written incorrectly. As an example, in our legacy code we were storing some aggregated data in sets, unintentionally masking duplicate data. When re-implementing these same aggregations for our new service stack, they were correctly implemented as a list, producing a mismatch in the data. Through our investigation, instead of simply matching the data to how our legacy service worked, we went back to our origin data, and ran the calculations manually through the Scala REPL. In doing so, we found that the new service was correct, where our legacy code was wrong. Fortunately, the bug within our legacy code was a simple fix. We implemented the fix within the legacy code and our mismatch disappeared.

Other Take-aways:

An important team take-away was to be very upfront and declarative about the work that the migration would require. Our investigation into the migration not only involved setting up a new technology stack for MySql, but also changing our build tool from Maven to SBT, introducing a Flyway + Jooq plugin to enforce type safety throughout the migration, designing a new data model (which was ultimately the driving factor for doing the migration in the first place), as well as upgrading our code up to the newest scala version to leverage all of the previous changes. Ultimately, we severely underestimated, and under-ticketed the work necessary to start our migration.

It is also important to keep in mind that every team is different and has different needs. When having conversations about database migrations, take the time to do a proper risk assessment for the work ahead. Keep these conversations going during the migration as well. As a team, we ended up prioritizing new feature requests and non-migration related bugs because the migration felt orthogonal to our production environment.

A further takeaway is that we could have saved ourselves a lot of time if we had more realistically assessed our users. In retrospect, the users of these reports were internal and would have been more lenient with smaller service outages, which would have allowed us to leverage our configurable services to migrate much sooner. At the expense of stability, we believe that we could have had a quicker migration by forcing ourselves to fix problems forward, instead of maintaining our legacy code for as long as we did. Still, most scenarios don’t have this luxury and we hope the façade based approach is of help to you.

Why the value of hackathons goes beyond free pizza Categories: Culture

bv_hackathon_logo_blue_background-01-copy-1-1024x644

About six months ago, we shared Why We Hackathon. At Bazaarvoice, we host a company-wide hackathon twice a year, and our next one kicks off this week. My previous post primarily focused on the people and company culture aspects of running a hackathon. In lieu of writing “Synergizing Innovation With Disruptive Hackathons”, this time around I’d like to  to share the real value we’ve seen from our hackathons, as well as some ideas on how to realize that value with your own hackathon.

Leveling-up our offerings

If you participate in a hackathon at Bazaarvoice, it is likely that what you create will somehow touch our clients. Many of the products and features we offer today got their start from a hackathon project.

Curations, our social media curation platform, started from one of our public hackathons. Bazaarvoice engineers working with engineers from FeedMagnet, one of the companies invited to participate, created the initial prototype of displaying reviews and social content together. Since then, we’ve put in years of work to build a system that can collect, manage, and display content at tremendous scale, but it all started with that original hackathon project.

Recently, we’ve seen many hackathon projects focused on helping consumers find the perfect product. Most of these took shape as various forms of personalization, like product recommendations, which we’re now offering as a limited-availability product. A different version of this technology is a new capability we’ve been experimenting with called Post Interaction Notification. In short, instead of showing products it thinks someone is interested in, it asks for reviews about products they’ve purchased. You guessed it — also a past hackathon project.

Our successful projects aren’t limited to display-specific projects. Our last hackathon saw projects that used different machine learning technologies to improve our ability to identify relationships across products, shopper profiles, and related consumer-generated content (CGC). For example, one project solved the problem of widely varied product titles and descriptions in our product catalog. If searching for ‘60” Flatscreen TV’, our product matching services would previously only return results that exactly matched those search terms. Now, thanks to the hackathon project, our system understands what 60”, flatscreen,  and TV actually mean. It can then find similar, but semantically different, items like “60 inch Flatscreen Television” or “60in. LCD TV”. This capability dramatically improves our personalization and syndication services by finding the same products across thousands of varied product catalogs.

While not every hackathon project grows into a full-fledged solution, they continually temper and sharpen the direction of our offerings and often improve our clients’ CGC programs.

Improving our client experience

Projects of this variety range from seemingly simple improvements to complex, long term initiatives. One project adored by both our clients and employees (we use these tools too!) was the ability to hop between different client accounts available to the logged-in user in the administration portal; simple, but a big time and frustration saver.

Speaking of administration tools, we’ve been building a new platform to build and deliver consistent tooling across our capabilities. Many of the last hackathon’s projects focused on providing actionable insight into the health of implemented Bazaarvoice capabilities on client’s websites, as well as supporting aspects like product-catalog imports and email-based review solicitation. These projects have shaped the solutions we’re building to proactively communicate the technical health of a client’s CGC program.

hackathon

How we hackathon

Normally, you’ll read “just give people pizza, beer, and games for a few days” as the recipe for a successful hackathon. In my opinion, those things are fantastic additions to any hackathon. However, it’s the people that make a hackathon great. To create success, empower people to find creative solutions to real problems.

We treat product development as a company-wide exercise. Having this sort of inclusive culture means individuals building, supporting,marketing, and selling our solutions are actively involved with client feedback, product feedback, and market opportunity across all of our functions. Providing a few days to create solutions to the problems seen over past months unleashes incredible ideas and creativity.

Secondly, this isn’t a top-down driven event. We involve a handful of hackathon participants and their peers to help plan and execute our hackathons. They make sure every activity, piece of swag, and communication is planned and ready to grow and sustain a big list of happy attendees. After all, the primary focus of the event to create space and time to be creative, have fun, and build stronger relationships.

Logistically, we made a change last hackathon that greatly improved the involvement from employees who weren’t participating on a hackathon team.It can be a bit challenging to stay focused with 50 teams demoing projects on stage one after the other, let alone remembering what you liked most in the final voting. Instead, we tried a science-fair style demo to promote walking around, engaging with various teams, and “investing” in the most promising projects using tickets. This boosted participation from those that aren’t interested in hacking, but who are interested in seeing and voting in the end results. This month’s hackathon we’ve got trifold panels, glue sticks, and markers to really get into the science fair spirit.

Synergizing innovation with disruptive hackathons

It’s not about buzzwords and forced team bonding. If you’re looking to run a hackathon, focus on the people — both those on hackathon teams and those who just want to spectate. If you’re looking to participate in a hackathon, try to find a real problem to solve. Regardless of your role, you’ll be pleasantly surprised at the creative, brilliant output. And at the very least, you’ll get some free pizza. 

Context and Higher Order Components: Two Immediately Applicable Topics from the Advanced React Workshop

Thanks to Bazaarvoice I recently attended an “Advanced React Workshop” put on by React Training and taught by Ryan Florence, one of the creators of React Router.

Ryan was an engaging teacher and the workshop was filled with memorable quotes. Here are some highlights:

  • The great conundrum of accessibility is that learning it is not accessible.
  • Like forceUpdate, never use context alone. Always bring a friend with you.
  • Good abstractions can always recreate other abstractions.
  • Is that a Justin Beiber song?

The training took place over two days in East Austin and I learned a ton of stuff that I was able to immediately put to use in my day-to-day projects. Let’s dive right in.

Context

In addition to the familiar props and state, React components also have access to a magical thing called context.

Context is for when your application has a hierarchical structure and needs to pass data from a parent component down to another level of child component (child, grandchild, great grandchild, etc.). As the React documentation puts it,

“In some cases, you want to pass data through the component tree without having to pass the props down manually at every level. You can do this directly in React with the powerful “context” API.”

Before we move on to the context API itself, let’s make it a little more clear why you would want to use it.

Consider the previous example of a parent node with a child, grandchild, and great grandchild. In order for the parent to pass a property (via props) down to the great-grandchild, it would first need to pass the property through child and grandchild. Note that there is no other reason for child and grandchild to care about that property.

As you can imagine, the desire to eliminate this kind of cruft increases as your project grows in size and complexity.

Context to the Rescue

context helps by passing properties through the tree automatically and any component in the subtree can access them.

When Your Component is Providing Context

Components (like parent in the example above) define a static childContextTypes object that defines the properties that they will put on context. childContextTypes is very similar to the propTypes object that we’re already familiar with.

Additionally, components that are putting properties on context must actually provide their values, and do so via the getChildContext component member function. This function should return an object with the same structure as defined in childContextTypes.

For example:

// Parent.js

static childContextTypes = {
  rating: React.PropTypes.string
}

// Just like a React lifecycle method.
// Don't worry about when this actually gets called.
getChildContext () {
  return {
    rating: '4.5'
  }
}

When Your Component is Consuming Context

To continue the example, your great-grandchild component now must consume this property. It can do so by defining a static contextTypes object that describes the context properties it expects, and can then pull the properties off of this.context directly.

// GreatGrandchild.js

static contextTypes {
  rating: React.PropTypes.string
}

render () {
  const { rating } = this.context;
  // Do stuff with rating.
}

Common Use Cases

Great use cases for context include localization, theming, etc. Basically any time you have things you don’t want to have to pass down through a bunch of components that don’t need to know about them. In fact, React Redux uses context to make its store available to all of your project’s components – imagine how cumbersome it would be to pass the store to every single component!

Common Pitfalls

Name collisions is a common pitfall. Consider the case where a component that has both a parent and grand parent component that provide a rating property on context. The suggested workaround in this case is to name your context after your component:

class ParentComponent extends React.Component {
  static childContextTypes = {
    parentComponent: React.PropTypes.shape({
      rating: React.PropTypes.string
    })
  }
}

class GrandParentComponent extends React.Component {
  static childContextTypes = {
    grandParentComponent: React.PropTypes.shape({
      rating: React.PropTypes.string
    })
  }
}

Then your child component could differentiate the two: this.context.parentComponent.rating versus this.context.grandParentComponent.rating.

User Beware!

A post about context would not be complete without pointing out that the very first section in React’s documentation about context is titled “Why Not To Use Context”, and it includes dire warnings like:
“If you want your application to be stable, don’t use context.”, and “It [Context] is an experimental API and it is likely to break in future releases of React.”

Ryan did point out to us however, that even though the documentation warns that the Context API is likely to change, it is one of the only React APIs that has stayed constant 😛

Higher Order Components

In software, a higher order function is a function that returns another function. Consider:

function add (x) {
  return function (y) {
    return x + y;
  }
}

const addBy2 = add(2);

addBy2(10); // 12

Similarly, a Higher Order Component is a function that returns a (React) Component.

/**
 * A higher order component that facilitates the creation of
 * container components that simply wrap divs around their children.
 *
 * @param  {String} displayName - The displayName of the returned component.
 * @param  {String} className   - The class name to use for the wrapper div.
 * @return {Component}          - A React component.
 */
export function build(displayName, className) {
  const ContainerComponent = (props) => (
    <div className={className}>{ props.children }</div>
  )

  ContainerComponent.displayName = displayName;
  ContainerComponent.propTypes = {
    children: React.PropTypes.node
  };

  return ContainerComponent; 
}

// Usage:
export const WidgetHeader = build('WidgetHeader', 'bv_header');
export const WidgetBody = build('WidgetBody', 'bv_body');
export const WidgetFooter = build('WidgetFooter', 'bv_footer');

Higher Order Components are great for reducing redundancy across components, and can also be used to “wrap up” a lot of complexity and encapsulate it, which allows the rest of your components to stay simple and straightforward. If you were familiar with the old React mixins, Higher Order Components replace a lot of what mixins did.

That’s All, Folks!

Context and Higher Order Components are just two topics I was able to immediately integrate into my React projects here at Bazaarvoice. Hopefully you will find them useful too!