npm npmCamp Presentations (2016), 30 Aug 2016

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: npmCamp 2016 - Ember.js, DevOps, and You: Lessons Learned From... by Iheanyi Ekechukwu

Description

Ember.js, DevOps, and You: Lessons Learned From Building an Ember Addon by Iheanyi Ekechukwu

A

A

Sup, hello, everybody welcome to my talk. Enbridge is DevOps and you this type will really be sitting around about DevOps and what I learned about building a plug-in for ember CLI deploy.

A

So you're, probably wondering who the heck is this guy, and why is he talking about DevOps that we're talking about NPM? But my name is a Hanukkah cuckoo I'm, a software engineer at digital ocean in New, York City and you can see that's my Twitter handle right there. So question of the day. What is DevOps like aren't too specialized people for that? Why should I even really care about that? Whenever I really see the word, DevOps I kind of have this reaction like because you know right so yeah DevOps.

A

What DevOps is is actually a clip compound of the words development and operations. So, looking at these two words, what you can do is break down the definition into well-defined, get as DevOps being the income processes around how you build your products and also how you deliver them to your end, users, and at first I thought. Devops was just like deploying, you know, like karoku get push, but it's actually more nuanced than that.

A

In addition to deploying DevOps is also provisioning or the configuration of your servers to run the necessary software, so your application can run it's also versioning, so controlled releases of your application. You should be able to track each release of your application and roll back to an older version in case something goes wrong and the scent and it's complicated, because since it encompasses the processes around how you ship your products and build your products, that's going to differentiate from person to person company to company so from continuous integration.

A

The code reviews there's so many different, like processes that can fall under DevOps. So it's really hard given a concrete definition, but one in a real part of DevOps is automation. Its key and while automation is, is the normally. You have manual tasks that you do in order to deploy an application, but with automation you can translate those tasks to code, and normally you can do that.

A

We're like IT automation, software like chef or puppet or ansible, but the importance of automation is that since you can see what your code is actually going to do, it leaves less room for human error unless you mess like up the code, but you can test that, so that's nice. So what do ember.js and DevOps really have to do it one another.

A

So a little bit history before, like back in the day before ember CLI, you had to just build your ember application, just the old way put into your javascript file and hoped everything was okay, but after ember CLI was built, ember CLI deploy came along and it really revolutionized how you could build your applications.

A

Rather than is manually, deploying it in SC paying files onto a remote server. You could just type in ember deploy production and you're good to go and like most ember plugins AC adapter plat pattern and I originally started delving into DevOps with the Ember applications, because I wanted to build a number see a light employee plug-in to deploy fast, booted ember applications to digital ocean droplets and fastboot is the library that allows you to surrender your ember applications for that SEO juice.

A

So, let's talk about the internals of an ember CLI deploy plugin, so a number CLI, an employee plug-in, can implement up to 11 hooks here at the 11 hooks that you can implement with it. It's actually missing one and it's did deploy, which is activated at the very end, so technically ember CLI deploy doesn't only just deploy it can provision in a creative way. If you did it in the configure step, it can also build your application before you deploy it and, of course you can deploy due to activation flag.

A

You can manage your releases and so I really don't really like the name. Ember CLI deploy I kind of told Luke, like hey, you should probably name an ember CLI DevOps, but that didn't really go over so hot, but yeah. Now that we have an intro I'm gonna talk about my experience, actually building a plug-in.

A

So the first step for creating an ember CLI deploy plugin is you have to create the M Brad on and then install to ember CLI deploy base plugin so with ember CLI, you can run this command. Ember, add on ember CLI deployed digitalocean, and that will generate your folder for you.

A

Then you can CD into that newly created project. Then you can npm install the ember CLI deploy base plug in the beauty of this base. Plugin is like those lips I mentioned earlier. You would have to like reimplement them from hand if you wanted to build a plug-in, but actually through installing this base. Plugin, all you have to do is extend it and you get access to all those hooks, so it cuts out my boiler plate.

A

The next step was actually manually, deploying the application and taking note of every single step and I'm really serious. Whenever I say every single step, so let's actually run through what these steps were to get a fast food application running on a droplet. So yeah first step create the droplet.

A

Next you want to SSH into the droplet whenever you I'm, while association to the droplet, you want to install the relevant software that you needed so enter next node and mpm, and then I'm professor boot server on your local machine in your ember application. You want to build the application and install its dependencies, and then you need to SCP the bill. Temper application onto the droplet dandy guy go back into the machine, go to the upload app and run the fast food application.

A

And then you have to update engine eggs to proxy / fast food support and serve up the static assets as a fallback. Then he want reload engine X and then you're done.

A

Then next step was to automate this entire process and translate some just translate code and I saw had that realization and I was just sitting at my computer like.

A

So when it comes to automation, I kind of like to break you down into either like problems and then solve those problems, one by one, so a series of task, so the first problem was: how do I create a droplet for the user? um You can like using digital oceans API, you actually can programmatically create a droplet and I create an ember command, allowed you to create a droplet with the relevant software, or even go into an already created droplet to install the relevant software dependencies and all that.

A

So, in order to do this, you need to create an access token from the cloud control panel export to an environment variable. Then you can run the command ember do provision which shows you which walks you do is wizard about. What you want to do, and the first step actually is just create a new droplet or list one to re exist.

A

Environment would be like actually to deploy target that you wanted to do and yeah all the other stages are like um things that you'll see in the actual control panel online. I guess like you, should be a disclaimer actually start building this before I started working at do so. This is not like some viral marketing type gig, but yeah, and then the next problem was, but I guess, like the last command, was kind of a misnomer in the sense that didn't install any of the software dependencies with it.

A

So we actually need to provision the software to server properly. um On my team, we actually used chef for provisioning, um so we just like designate like the cookbooks and recipes of test run, so you have a primary server that has all the configuration and once that changes it alerts all the secondary server is connected to it like hey, I, changed so rerun your task again or whenever it's deployed so yeah, it's nice to have that with like chef, puppet ansible and all those are ones.

A

But you really don't have that what ember so yeah before fastboot, meaning a regular member application, that's rendered on the client side. You really just had one like software dependency and I was just like a web server like n genetics or Apache to serve up your static assets, but now with fastboot, you have more dependencies. It's not only engine X. We now need node mpm and also like the fast food application itself and another solution that I figured out from like just stuff that we had internally was our one click.

A

Install images I'll be able to use like the node image to help install the node in mpm dependencies and then for like actually installing other software. You can just SSH into the droplet and execute commands. 10 sod rest of the dependencies, so yeah like I, was mentioning yet a one-click install app of on the node for no DJ s and you can actually access all of these one-click install images from the API, so I just figured out what the idea of it was and used it in the provision command from earlier and then inside.

A

The ssh configuration for my inside my code whenever you do create a droplet, you can designate ssh keys. So with that you need to set the environment variable, say: okay, here's where my private key is stored and also here's the pass phrase in order to get into it and it's Frankenstein code, because I haven't refactored it yet but yeah. This was swollen current, like the will upload hook before. We even think about uploading the app and just connect to server update to dependencies and also or install relevant dependencies. It's kind of drinky, but it works.

A

I'll, actually talk more about how this could be improved later and so like with the provisioning finished. Now we need to figure out how to actually build my number application and get it on to the droplet and the beauty of ember CLI deploys that you don't really have to reinvent the wheel for certain functionality because of like the plug-in system. So you can use ember CLI deploy, build to actually build the application.

A

Then you can mpm saw the dependencies locally and as a PS CP, all the files onto the droplet and yeah just handles all the building for you. So I don't have to really worry about running that command to build the application and here's an ftp client for to actually uploaded dist directory of a built, ember application, and before you upload the files, we will need to install the NPM dependencies locally for fast boot.

A

um Best moods actually changed a lot, so the step may not be necessary anymore, but I'm still trying to like catch up and update my application or plugin for it, and then we will actually upload the built files onto the server and upload it to an ember, app directory on the remote server and remember how I mentioned ember CLI, deploy build earlier on. This variable context at distr gives any add-on access to the files that were built by the ember CLI deploy build plugin. So like that's, why?

A

It leaves us one less thing to do and you can actually use this to pull in other plugins like there's some that automatically tags or releases for you and like as you build an application will give it a new tag, and you can just use that to your advantage, so it's nice being able to piggyback off of the community in order to give back and contribute and then another prominent, yet ballast fear whenever you're shipping them any application production application.

A

What, if, like your server restart or something goes wrong like how do I make sure that app, the fastboot application will stay up and running so in order to keep server running, you can use boom tues up starting this system to configure applications on the server I really had no idea that Linux, this administration stuff was so difficult. I like actually reading our tutorials to figure out most of the stuff, but this like do creating this um upstart service.

A

Essentially, what this file is saying is like hey: whenever the server has started and the file system is ready, you should try and install ember fastboot server if it isn't already installed and then you should run the fastboot server application and make sure it's up and running and if the user, if the server stops or the user explicitly says service stop fast boot, then you should exit out the application and it's graceful select in case anything happens.

A

I'll always be up and running and then using that off, like SCP, SCP, client, again I just over I, just added it to the NIT folder and that put down automatically put that service and the server's event system. And then the last problem to solve is how do you make sure that the application is up and the user can see it and one like just like the upstart stuff that we just did? One can actually create a custom and your next template to upload to the server and in place of the default engine.

A

X engine X configuration and is like a snippet of the engine, X config for the Emperor application and then similar to the upstart service. We just directly upload it to the droplet as an engine next on cough file and then, after everything is said and done, we can restart or actually, in this case, probably reload to end your next config and theoretically. A step may not be necessary so long anymore, but I just did as a safeguard and we're finally done. Automating yeah.

A

That's me a big allah. Blank stares, I feel like a law. You are like yeah.

A

So yeah there's one small mpm trick that made building deploy, pluck like this deploy plugin way easier and it was hidden the documentation, but it's like em p.m. link so before I was like man how to actually test this and like mice, like application, I'd have to like publish it to NPM. Every single time wanted to test something.

A

But with this you can actually use NPM link in order to use NPM link on the plug-in directory from any amber project and you what will do is it will install the installed like package from your local file system into your other project, but you don't really have to keep on installing it every single time. You make a change. It keeps up with the changes in real time, so it's pretty nice to be able to see like the changes I'm making immediately, and it's really simple, like you can like run it this way.

A

Just by, like you know, do an MPN link the project name and then relink it in another directory or shorthand just go to the project you want and then using NPM linked to the absolute file path of the JavaScript package. You can also do that as well to save time and yeah, so we went from being a remember flames to like figure out the slizz, so it takes so much time on touching the addon.

A

Pretty clutch like gonna lie so yeah you can check out the repo here, it's with the source codes at try and do some more work on it right now and improve upon it and like water plug in the dope it kind of habit. It does have it short the shortcomings for one. This is really no way to deploy to multiple servers. At the moment. Like I said earlier, like chef, you just have one primary server and then the secondary server is communicate with that primary one.

A

But you know JavaScript right now, so not really having that right now and I. Think the importance of having multiple servers is mostly from, like you know, a load balancing perspective, so that would be kind of crucial, but may not be as big of a deal breaker for some individuals or small projects and a way to plugins coated right now there are provisioning tests. Are that get rerun unnecessarily?

A

What I'm trying to do right now is actually move that provisioning, like installation, step of like engine X and everything else from within the actual plug in into the provision command.

A

So after the draw play is successful, successfully created, it should just SSH and automatically install the dependencies that needs, and it should be good to go and also there's no means of controlling, like there's no controlled release management in order to activate a new version of the application did theoretically would have to go back like get check out an old version and then redeploy, but I kind of have an idea of how I'm going to solve this, because what you can do with fast food app server, which is a production server.

A

That's ready for fast poo is create Watchers. That alerts the server whenever a new version is created so using symlinks on a base directory and concatenated. Like each version on to the end, you just have to switch up the directory for the server and once those files change it will restart so yeah, that's cool and but yeah I can't do that until I get rid of the old, fast food serving logic.

A

So, like I said it's hard kind of keeping up with the community and how things move really fast in web Lin, and even the last like one year, but like so yeah I'm, hopefully going to get to use fastboot app server in place of the old, fast food logic serving logic. So yeah and I definitely have ideas for future enhancements.

A

Like I mentioned earlier, zero downtime deploys be pretty nice um and I think it's once a migrating fast food app server, since it's running a clustered it like spins up 16 processes or something like that and restart someone one process at a time it should have with zero downtime deploys if I can get that strategy like ironed out, get it more closely and get it working and ssl support.

A

You know I'd feel like it'd, be dope their automated support for SSL or TLS wet, let's encrypt and just automatic renewal of the certificate, because it'd be nice to just be able to ship your application. She's, like oh, hey, sorry got like ssl set up for me. Cuz. Nobody really wants to set that up.

A

Every single dime- and you know today before today, and never really knew how to define semantic versioning I just saw it a lot on the Internet I was like yeah cool, get makes sense, you know, but um really it's important to have a semantically versioned library, especially when it comes to something like as business critical as deploying because you don't want like have a breaking change like unexpectedly.

A

So if you get to break you up into major and minor releases to make sure like every nothing breaks and everything's all fine and dandy, because I really don't want to be held liable for losing somebody money and yeah I guess the question stands.

A

What exactly is the future of DevOps there's a lot of talk about docker and docker containers and, like you know, the lightweight, and they really allow for a rapid deployment of like applications and even marrying like a production environment that closely to your development, environment and I'm, like what's the future of dr. number, together I'm kind of like I'm, trying to explore how we could container eyes a number of fast food application to or how the two play together and I'm actually like I'm wondering? Is that really even possible and I had no idea?

A

So, oh yeah, so yeah, just a couple, Danks Tom Dale he's helped me allow a lot would like thinking through the logic of what I'm doing and just all his fastboot work. Luke Melia he's got a created, ember CLI, deploy. He answered a lot of my questions. I had around the deploy, plug-in and like contributing so I really appreciate him David pets in our amber developer. He was 11 the few individuals that tested out the application or the plug-in on a side project and getting real-time feedback and, of course, digitalocean for like hell.

A

Flying me out here and let me be here and speak and everybody else so yeah thanks.