Why do we need DevOps?

Engineering Manager, Eugene Dina, discusses dev-ops and how the company is leveraging it to streamline its process, and minimize its time to market.

By empowering its engineers at every step of the development and deployment pipeline, and utilizing automation, infrastructure as code, and automatic provisioning, teams are able to build and deploy software more efficiently, and with fewer defects.

Developer Interviews: Justin Podzimek

What do you do and how long have you been at bbcom?

I’m currently a Principal Software Engineer and have been working at Bodybuilding.com for about 8 years now. In addition to normal development, I’m responsible for helping my team of client-side engineers, software engineers, and QA in whatever project we’re currently working on. I also have a helping hand in guiding the technology we use at Bodybuilding.com along with the other technical leaders.

Linux or Mac?

Mac, but only slightly. There are some applications and utilities on my Mac that just aren’t there, or aren’t as good, on Linux. That being said, a good portion of my work is spent in the command line, so as long as it’s not Windows, I’m not super picky.

What is a typical day like for you?

The mornings are fairly consistent, with a daily scrum to discuss what we’ve been working on and what’s coming up. After that, my day is spent in a combination of working with the other members of my team, meetings, and heads down with my headphones on getting some work done.

What is in your developer tool box?  

#1 - An IDE. Depending on what I’m working on, it’s either IntelliJ or Sublime Text.  #2 - Slack. Getting push notifications on my phone when someone needs me and I’m not online is a huge help. #3 - Google Chrome as my preferred browser.  #4 - Spotify. Music is important to my development mojo.  #5 - Terminal.

What are your favorite pro tips?

alias yolo=‘git commit -am “Deal with it” && git push  -f origin master’

Kidding. 

Honestly, the best tip I can give is "stay in the flow”. It’s tough to get back in once it’s broken, so do everything you can to minimize distractions.

What did you code up today?  

We’re currently in the process of migrating upwards of 28,000 articles to a new CMS system living in Amazon Web Services (AWS). Today, I continued on with the migration script to download every image from an article and upload it to S3. In addition, tweaking and fine-tuning the process to make sure the cutover is as seamless as possible to our readers.

What has been the most challenging problem you’ve had to solve?  

Last year, we update the wrapper (header, footer, and sidebars) across our entire site. This was a HUGE project. Every technical team was involved in updating around 15 different Java and PHP applications, migrating our front-end code to use a new build system and AngularJS, site-wide feature enhancements like search and a new navigation, and countless other features. This was all completed in just a few months before the holidays. In all my years working at Bodybuilding.com, never have I seen collaboration between such a large group work as smoothly as it did.

Any crazy Bbcom stories?

The only one that comes to mind is the “poo-nami”. I think that one should be left up to the imagination.

What is the best thing about working at Bbcom?

You get to work on the coolest, most cutting edge products, with the coolest, most knowledge people around. There’s a lot to love about working at Bodybuilding.com.

Everything You Need to Know About Bower, Part 2

In Part 1 of this two-part rant on Bower, I discussed many of the arguably non-intuitive quirks of SemVer as implemented by NPM and Bower.  In Part 2, we round that out by sharing how we've dealt with those peculiarities and successfully collaborate using a private bower server.

How to Make it Work

At Bodybuilding.com, different teams are often making distinct changes in the same code base, and we need to be able differentiate between their releases.  Git branches are obviously going to keep the work separate, but in order to consume or deploy the changes, we need to pre-release the app to Bower, and then it's out there for anyone to consume.  Here's where the whole "stable release" restriction and shorthands can cause trouble.

We Need to Be Restrictive

Third-party apps are generally fine to leave open-ended, using the ^ or ~ shorthands.  More often than not, we want the latest bug fix.  But as we're developing our own interdependent apps, we need to lock down the version request.  It's verbose, but in most cases the only way to restrict the version to a set tuple, while allowing pre-releases within that tuple, while also rejecting any version outside of that tuple, is:

>=x.y.z-0 <x.y.z+1 (where -0 is literal text and z+1 is one more than the number represented by z)

Or, if using a pre-release tag:

>=x.y.z-tag.0 <x.y.z-tag.100

Ya gotta tell Bower what ya want.

For example, use >=1.2.3-0 <1.2.4. Note that we need to explicitly use the -0 part of the version; without it, Bower will only look for stable releases, and, in effect, we could have just said =1.2.3.  With the -0, though, we gain several things:

• Bower will consider pre-releases acceptable, and pick up the latest 1.2.3-x version.
• Bower will refuse 1.2.4 and higher, pre-release or otherwise, ensuring that the team gets the version they're working on.
• Bower will consider the 1.2.3 release valid for that range and prefer it over the pre-releases, meaning the app that is requesting the version doesn't necessarily need to update its bower.json in order to release.  Ideally, at release time, bower.json specifies exact version numbers, but if a version was missed then this notation won't break the desired behavior.

Similarly, if a different team uses >=1.2.4-0 <1.2.5, they will ensure that they don't accidentally pick up version 1.2.3 when it becomes an official release.

Examples

Note that you can try these with your own private Bower registry.  Create a new git repo, register it with your Bower server, and then iterate over changes to the readme file, while releasing the various tags.  Then you can test with bower info [your-bower-project]#[request version here].  

Further note that you need to either escape the >, <, and space characters – for example, bower info test-semver#\>=1.0.0-0\ \<1.0.1 – or else wrap the whole package#version part in quotes, like bower info "test-semver#>=1.0.0-0 <1.0.1".  (For reference: how the command line uses spaces and angle brackets)

Examples of Favoring Stable Releases

The examples below assume the following available versions:

• 0.0.0
• 0.0.1
• 0.0.2-0
• 0.0.2-1
• 0.0.2-2
• 0.0.2
• 0.0.3-0
• 0.0.3-1
• 0.0.3
• 0.0.4-0
• 0.0.4-1
• 0.0.5-0
• 0.0.5-1

Requested version	Resolved version	Notes
^0.0.0 or ~0.0.0	0.0.3	Bower assumes you want the latest stable release. 0.0.4 and 0.0.5 exist only in pre-release versions, so 0.0.3 is resolved.
>0.0.3	0.0.5-1	Only pre-release versions are greater than 0.0.3 (because we aren't including 0.0.3 itself), so even though Bower would assume you want the latest stable version, there is no latest stable version in this case, and the latest pre-release is returned.
~0.0.4	0.0.5-1	Essentially the same as the above example. We want the latest patch greater than or equal to 0.0.4, and since there are no stable releases in that range, we get the latest pre-release.
0.0.5 or =0.0.5	none	Declaring a stable version with no range will not return any version if the stable version doesn't exist; Bower assumes you want the exact stable version requested and will not return the pre-release versions.

Examples of Pre-release Tags

The following examples assume the following available versions (released in order):

• 1.0.0-alpha.0
• 1.0.0-alpha.1
• 1.0.0-beta.0
• 1.0.0-beta.1
• 1.0.1-foo.0
• 1.0.1-foo.1
• 1.0.1-bar.0
• 1.0.1-bar.1
• 1.0.1-z.0
• 1.0.1-z.1
• 1.0.1-a.0
• 1.0.1-a.1
  

Requested version	Resolved version	Notes
>=1.0.0-alpha.0	1.0.1-z.1	Apparently, Bower considers anything with a pre-release tag as simply a marker to get a pre-release. It considers the "latest" to be the highest in an alphabetically sorted list of tags. Therefore, pre-release z is chosen.
>=1.0.0-alpha.0 <1.0.1	1.0.0-beta.1	The range really only limits the numeric triad to 1.0.0, so we filter out all of the 1.0.1s. But the "latest" version is the highest alphabetically, so beta is returned.
>=1.0.0-made-up-stuff	1.0.1-z.1	Note that the pre-release tag doesn't even exist, and moreover there is no "dot number." Bower sees no foul in this. It muses, "Ah, you want a pre-release. Here's the latest pre-release!"
>=1.0.0-alpha.0 <1.0.0-alpha.100	1.0.0-alpha.1	This actually works! Cumbersome as heck, but it does enable separation of pre-release tags.

The End (v1.0.0-a.1)

Hopefully this guide to how Bower handles SemVers reduces your aches and pains.  If you'd like to read more, check out the section below.

References and Further Reading

Official SemVer spec: http://semver.org/

Official node-semver spec: https://github.com/npm/node-semver#versions

A decent introduction to SemVer ranges: http://developer.telerik.com/featured/mystical-magical-semver-ranges-used-npm-bower/

A short and easy-to-read attempt at making the SemVer spec more approachable: https://github.com/dominictarr/semver-ftw

Building an Upload Microservice using AWS Lambda

One of Bodybuilding.com’s core missions is to help people transform their lives through the adoption of a healthy lifestyle and exercise. One way we motivate people to begin such a journey is through our yearly “transformation challenge.” Over the course of 12 weeks, participants in the challenge are encouraged to adopt a training plan and set new habits that will lead to a fitter, healthier body. Prizes are awarded to the people who make the most profound transformation over the course of the contest. 

In order to participate, contestants are required to upload photos of themselves before and after the challenge. As one might imagine, the time period shortly before and shortly after each challenge tends to produce a spike in photo-upload traffic to our servers. Historically, these periods have been problematic for the scalability of our backend systems. In order to improve our end-user experience (not to mention our own), we decided to re-architect our photo-upload system in late 2015.

This tech talk describes how we created a new microservice-based architecture using Amazon Web Services technologies as well as a Java/Spring stack. Our goal was to greatly improve our ability to handle spikes in image-upload traffic without manual intervention. As a result, we created autoscaling RESTful services using Spring Boot and AWS Elastic Beanstalk. Image-processing logic has been encapsulated into highly scalable Amazon Lambda functions.

We recently launched the 2016 transformation challenge. With our new image-upload infrastructure in place, we have been able to handle peak loads without a hitch.

Video: What is RESTful? (2.0.0)

Roughly two years ago, the Bodybuilding.com Commerce Engineering Team set out to deliver native mobile phone applications for iOS and Android that offered the same rich user experience that hundreds of thousands of customers enjoy within our web applications each day. We started with a couple of UI/UX mockups, a firm knowledge that we needed to develop an API, and a strong desire to make that API RESTful. A large, dedicated team spent 18 months of their lives learning and growing to achieve a soft launch of the applications in May of 2015 in the Google Play Store and iTunes. We believe a project of such magnitude must, beyond its primary goals, create useful and reusable assets to be considered a success. By that definition, this project was a huge success. We have a few libraries we plan to open source, a bunch of best practices to document and present, and a ton of knowledge and opinions to share with the development community.

Other than the applications themselves, the most valuable assets developed during the project have been our opinions on what defines a RESTful architecture. Even months into the project, we found ourselves in long debates that led to redefining our views of RESTful architecture. Each of these cycles triggered new breakthroughs and pitfalls in our implementation. Even now, months after the launch, we still find ourselves re-evaluating past decisions and making significant changes in our implementation to align with the newest visions. The pieces of that newest vision were collected together into a presentation named "What Is RESTful?" that was used multiple times internally to achieve a common base understanding from which the more interesting conversations around implementation grew. Even this presentation has gone under multiple revisions and has recently achieved the SemVer status of 2.0.0. This 2.0.0 version was recently presented at the Bodybuilding.com Fall 2015 Tech Talk and is now available for online viewing. We expect to use this presentation in future posts and presentations as a launching point to dive deep into the finer aspects of our implementation and the software technology that grew from our vision. It is certain that we have not gotten everything right, and we encourage feedback, both constructive and critical, in order to develop the 3.0.0 version of this vision.