GitHub GitHub Satellite India 2021 - DevOps: Day 2, 27 Mar 2021

Previous Meeting

⏯

youtube image

►

From YouTube: GitHub Satellite India 2021 - DevOps Day 2

Description

A community connected by code. March 26-27, India

Join us virtually for two days dedicated to celebrating India's developer community. Expect announcements from GitHub leaders, hands-on workshops, and inspiring performances by artists who code.

Increase efficiency, automate workflows, and create secure software all at once.

https://githubsatellite.com

A

A

A

A

A

A

A

A

A

A

B

B

C

C

B

B

B

B

B

B

B

B

D

D

D

Hello and welcome to day two of satellite india, I'm shonku- and I want to thank you for joining us today. We've got another packed day of programming and to start off, I want to take you, through a tour of github's products, to show off some of our latest features and to show you how you can use them together to build software across the world.

D

Now, yesterday, erica was here to celebrate all the amazing things that are going on in the github community in india. Growth of the community here has been incredible. In the last year, 1.8 million developers have joined github from india, and over a million developers have created their first repo on github.

D

India has now become the fastest growing country in the world in terms of developers contributing to open source. That is amazing.

D

Erica has also announced a new set of github programs in india that we're very excited about the github grants for open source program, which will provide 10 million rupees of grants to open source projects based here in india.

D

The github accelerate program, which will partner with startup accelerators and the indian government, to expand support to over a thousand startups, with our complete platform and finally, the github externship program for students, which will provide participating students a chance to get real-world experience working with software companies and we're really thrilled to see what you are accomplishing with github here in india.

D

This community is poised to double in the next two years and we're excited to see what you will build with github next now github is the world's largest developer community and it is for a purpose, and that purpose is to build together. We think of software development as the world's largest team sport, and we believe that it is a sport without borders.

D

Now I was born in india and I moved to canada when I was young to study computer science, and I had to move again to the united states to take advantage of the best opportunities in tech. That was the reality back then right. If you wanted an opportunity, you had to move to go, find it remote learning remote work, global companies.

D

These were all concepts that were unheard of, but today it is possible for a team of strangers to come together from across the world and build incredible things together, and this is what we see on github every single day collaborations and contributions from developers all over the world in real time, and it's also increasingly the reality at work.

D

Teams of all shapes and sizes companies from all over the world are going remote and they're going worldwide to tap into the best developers, whether they come from mumbai or mombasa or miami, and those developers need to be able to work together on projects from wherever they are. That's what I love about, github github was built for this kind of collaboration.

D

So today I want to take you on a tour through github's products and show you how you can use them to build from wherever you are now. Do you remember your first time joining a software team? I know I do that thrill of writing code and committing it for the first time and seeing it in a product right, but first there were all those things you had to go through right. Setting up your dev box, setting up your tools, getting your code and looking through it to learn going through docs and style guides.

D

All of those things that you had to learn before you can contribute something without messing it up, even with the best mentor to help you that could take days and the reality is today. You have to do that. All the time developers almost never work on just one thing: you're, always having to switch context, work on different projects, fixing live site incidents, contributing to other teams code or even just learning something new. That is not easy.

D

On average, it takes over 23 minutes to switch context and get into the flow, and that is if you've got everything set up for you. Often it can take a lot longer so github we care about this a lot we want developers to be able to get into the flow get into the code and contribute whether they're joining a project for the first time or just getting back into it after a while to show you how github helps you get in the flow. Here's neha.

E

Hi, I'm neha, I'm the director of engineering here at github for the team that works on discussions, I'm pleased to announce that we've now enabled discussions on private repos, and that means for this private repo. We can have our team conversations here and leave our issues for tracking our work. Here we have some internal chat about skylines. These cool 3d printed contribution graphs that people have created using a website we built, but we also use discussions to talk about ideas in the team and with others in the company.

E

My teammate diana had an idea here the other day about adding a year selector to the page, and I really like that. So I want us to work on this. um So let's create an issue from this discussion thread and track the work. So I'm just going to go in here reference the issue give it a title and then I can just scroll down and I can create the issue. So this shouldn't be too hard and I'm thinking I might be the one to pick this up, and so we can do it together.

E

Let's just take a quick look, so I'm going to go over I'm going to assign the issue to myself, and you can see that this isn't the code base that I work in every day. This is just for fun, so, rather than downloading all the dependencies to my machine and getting a full development environment set up locally, I can open a code space right here in github and I can get my own personal dev box in the cloud right away, and this isn't just running the visual code editor in the cloud.

E

This is a full-blown personal development environment in the cloud. So, for example, if I like go over to the terminal- and I type in npm- run dev here in the console, I can get the full local development and experience as if I'd had the code locally on my machine codespaces even sees when my code is running a server on localhost and so it'll. Let me connect to it across the internet, taking care of all of the htgps tunneling and authentication to ensure only I can see what is running on my dev box in the cloud.

E

So let me type in my github id nerd neha and let me check out this awesome graph, so this site's pretty awesome. It's got like a nice soundtrack and the coolest part is it's running on my dev box in the cloud, but yeah 2020 isn't really a year. I want to remember so. Let's get started on that issue that diana mentioned by adding a year selector here codespaces is already authenticated with github. So actually I can actually just go to the github extension in vs code and it has my issues already listed.

E

So if I click on it, you can see the issues listed here. I'm going to find the issue that I created and start working on my issue, and you can see that it's already created a new feature branch for me to add my changes.

E

So all I have to do now is take a look at the files, so I'm going to go over I'm going to find the form part which I happen to know is called hero.view, and I know that the code that I want to add in is between the at symbol and the submit button. So I'm going to quickly add the code for a year selector and then I'm going to hit save now. This is a full development environment.

E

So if I just jump back into my application, we can see it's refreshing just as if I was running code locally and it'll start to bring up the selector there. It is so I'm going to type in my username again and I'm going to test it with the 2018 year. I know that I had like a pretty good year that year um and also I had this like great break for summer and thanksgiving.

E

So I want to see if those show up- uh and so here we go okay, so you can see. Oh okay, so you can see my skyline and you can see that, like I even at the end, there's like some nice gaps, it's when I took my end of your vacation super seriously, um and so this is all running in my private development machine in the cloud um I can come in, I can mess with the graph I can scroll around, and so I know that this works and I'm ready.

E

Let's commit this and let's share the changes with the team, so what I want to do now is. I want to go back to my editor and I want to first commit that change locally. So I'm going to go into the get view um and you can notice. It already includes a link to the issue that we're working on and so um then I can go ahead.

E

I can commit it and now I'm going to create a pull request, so I can do that by going straight to the extension um and then I'm going to go into my github view, I'm going to create a new pull request. It's going to figure out which branch I want to compare to, and it's going to even publish the branch for me.

E

So I'm just going to scroll down here hit create and then I can create the pull request and it's going to do all of the work of publishing the branch, pushing the changes in the server and even creating the pull request in the server. So you can see my pr it's like right here. So if I go back into my repo on github, I can navigate over to the pull request tab.

E

So let's do that really quickly and there you can see my pr is available uh your selector for graph, and you can see that it's ready for my team to review and I can see that the ci builds have already kicked off. So I can to verify my change. So let me go in I'm going to tag this with the enhancement label and then I'm going to you make use of this super cool new feature.

E

It's called auto, merge and github, and so what I'm doing here is I'm having the pr get merged automatically as soon as all the checks are complete and the code has been reviewed by the team, and so I don't have to do anything here, and this leaves me free to get on with my next change. In the meantime, let me hand you back to shinku.

D

Thanks neha now, once you've written your code, it's time to get it reviewed and committed now. Code reviews have also changed in this new world used to be that you could invite someone over to look over your shoulder, review your changes and give you a quick thumbs up, but now that you're working on a distributed or remote team, you still need that review.

D

You just need new ways to do it now, on github, everything about code revolves around the pull request and the code review experience, pull requests were designed to let strangers review and collaborate and commit code together now, during the last year, as teams have become more remote, we've seen something amazing on github developers are actually submitting pull requests for code reviews more often than ever before, as much as 30 percent more frequently and when they do, those pull requests are getting reviewed faster than ever before, and that's good news.

D

If your prs are more frequent and smaller they're easier to review, and that leads to higher quality code reviews and higher quality code and when they're easier to review, they also get reviewed faster, which means that they get committed faster and shipped faster.

D

Better prs also means not just having human helpers, but the bots github actions and our amazing ecosystem provides an integrated ci platform with thousands of custom actions written by our community that can automate all sorts of reviews running your unit tests, applying linters and style guides. Checking your dependencies and just helping you commit better code to show us more here is a tool.

F

I'm atul and I work on the databases team here in hazard bar this is the pr from neha where she has added a new ear selector to the site that we are building for every change we make to this site. We build and test it automatically using data actions. Let me show you the ci build for neha's latest changes. If I go into the summary view, you can see the brand new visualizer showing us running the build and test steps on both mac and linux machines all at the same time.

F

This takes advantage of data actions and our hosted build runners which provide you with secure, build servers on demand. All of this infrastructure is managed for you by our team at github so that you can add new build machines whenever you need them for either linux windows or the mac. Let's go back. Let's look at how the build is configured. Our workflows are defined in the dark data folder under workflows. We are going to look at the ci bill.

F

First, we can see that it is triggered by new changes in the main branch, but also by any pull requests that are targeting the main branch we also define which operating system and node versions. We want to run our build on, and then we build our code on the linter and do a full test shoot pass.

F

The actions community is fantastic with over 7000 open source actions in the marketplace in the ci build, we have added an awesome community action from tim lukas and joseph peterson. That will update draft release. Note for you, based on your issues and full requests. So that's that's the power of community for you that you can use to change your ci. We also create and deploy a test environment in the cloud for every single pull request. This makes it easy for us to review the changes in a working environment.

F

So let's take a look at the dev test and development workflow, which is run in parallel alongside our ci bit. So, as you saw, this is triggered by pushes to main, along with new pull request. We come in get the code and then deploy it to our cloud provider using a secret which we store securely inside github's built-in secret store, we've also added a job to this workflow that will come back and destroy the environment when the pull request is closed, to keep our costs down.

F

So let's go back to the pr from uh the pr from neha and review it. um Let's scroll down. I can see the link a link that has been added by the deployment, so I can click on that and quickly test the changes without having to install anything myself.

F

I can see that the new here selector is there. So let's just enter a name and select right here, 2018 and then take a look at the graph. The graph is uh loading.

F

And yeah, I think the contribution block looks good. It has the ear as well. So I think the ear selector is working you're able to select the ear as well as sheet on the contribution graph. So let's go back to the pr and let's review the code, um the code, the code looks good, so, let's just say, lgtm and quickly approve the code now because they has enabled auto merge.

F

As soon as I approve the change, the code is deployed, the code is merged automatically and data actions take care of destroying the test environment for me, so everything just works. End-To-End.

D

Thank you now, by using github pull requests and actions, we are seeing our customers ship code to production way more often, but it's not just enough to ship faster. You also have to ship safely and securely.

D

One leaked secret, one vulnerable piece of code, one insecure dependency can cost you time, cost you money and cost your customers trust now we're humans and humans make mistakes. It turns out. The majority of security vulnerabilities are just human mistakes and once those mistakes get into code, they're hard to get out.

D

So if you can fix those mistakes before you ship or even before, you commit code, that's great, and if you can't, then the quicker you can fix them once they're found the better so with github security tool, security where it makes the most sense right into the developer workflow.

D

So you can find and fix these issues more quickly. So to show us more here is niroshan.

G

Hello: everyone, my name, is niroshan rajadheri and I'm part of the application security field team here at github, and I'm going to talk to you about the work we've been doing in this space through a quick demonstration.

G

In our example. Here we have advanced security enabled in our organization, and this allows us to easily switch on code scanning and secret scanning for our private repos, but everyone can still use these features on your public repos for free just go to the security section of your repo settings and switch it on there.

G

I can get an overview of the security situation for my project from the security tab in the repo beginning with supply chain issues. Here we have dependebot enabled it is automatically looking for vulnerabilities with my dependencies and it will create a pull request for me if one of them needs updating.

G

Looking here I see, dependebot is telling me about a new release of one of the actions that atoll was using. It's a minor revision. So I'm going to go ahead and approve this change to keep us up to date.

G

All the tests are passing, so I'm going to go ahead and merge this change now, even our build processes can be kept up to date with dependerbot back to the security tab. I see that secret scanning is also active.

G

Let's take a look to see if any of my team accidentally checked in secrets or api tokens, so we can stop them from being externalized.

G

If we look at the closed issues, we can see one of my developers accidentally checked in code that included a token in the past, so we've actually revoked that and removed it from the code.

G

So that's known, vulnerabilities and sequence taken care of, but what about our unknown vulnerabilities with codeql from github we're able to scan our code base looking for code which could be introducing potential security vulnerabilities into the code that we're writing? As you can see, it's identified one here for me. I can show the code part that was found with code cures, analysis tools, and I can see that this is definitely an issue.

G

However, it doesn't look like it'll lead to a major exploit in our case, but let's create a quick issue from the code, so we don't forget about it. This should be a simple fix to explicitly tostring the value from the url I'll. Also mark this as a good first issue in case someone wants to pick this one up once I'm happy we'll save the issue to make sure we don't lose track of the work.

G

We want to do excellent, we'll get back to working on that right away, while we've had the security tab on your project for a little while now today, I'm pleased to announce that enterprises with github advanced security now have a security tab added to their organizations where you can get a quick summary of all the projects.

G

You also have an overview of all the projects in your organization and what the status of their security features. I'm also pleased to announce that this is now available to teams who can see the security status of the repositories that they administer with that back to you, shanku.

D

Thank you nirushan. Now, we've got our code written reviewed and committed where our code really belongs, is in our users hands, and that means shipping doing the functional testing sharing with stakeholders and getting them to sign off and releasing the production and celebrating together. Now, github's new tools were designed to make it easy to test and release together as a team to show us more. Here's katie.

H

Thanks chanku, hey everyone: now we have a release: that's ready to head out the door thanks to a tool. We already have a draft of our release, notes, let's edit a new release based on the draft notes.

H

First, I'm going to set the title then I'll check the draft notes that were automatically.

H

Created to be honest, these look pretty good already, so I'm going to go ahead and publish this release. Now we have a workflow set up in actions to be triggered by the publication of a release with our new workflow visualization ui. You can see the release is now deploying to staging.

H

Then it'll run some end to end smoke tests before finally deploying to production. While all of that is running, let me show you how I've set this up. Let's take a look at the code in another tab in the github workflows, folder we'll find the releases.yaml file in the release. Yaml, you can see I've set up a couple of different triggers a manual trigger with workflow dispatch as well as triggers based on when a release is published in the deploy staging job. I specify the environment name and url, where this will be available.

H

You can see I'm also requesting the published token secret from github's built-in encrypted secrets fault. After that job is completed. We run our smoke test, then a deployed production again specifying the environment name and url before calling the same action with published token as a secret.

H

Now these steps are identical, yet they publish to different places. That's because we define a different publish token for each environment. Let me show you how we do this. I head into project settings then, in the new environment section I have staging and production on the staging environment. You can see we have a published token defined in github's encrypted secret vault, even as admin. I only have access to update this. I can't view it in the ui going back to production.

H

I have the production publish token defined in our secrets here as well, but what's more interesting is that you can see I've set this up to require manual approval before code is deployed to production. Only people from this team can permit code to go, live and there's a full audit trail to see who approved it.

H

Great staging has been deployed and our smoke tests have run. Let's take a quick look at the staging environment. Awesome looks good. I'm going to enter my id to test out my own graph and make sure everything's working well.

H

Everything seems to be working well, nate has code, is included, so I'm going to approve the release to production.

H

Back in the github ui, I can click on review deployments, select the environment and leave a quick comment approve the deploy there we go deploying to production is underway again with a different publish token from staging.

H

And with that we're deployed I'm going to click the url just to validate.

H

Great head over to skyline.github.com, if you want to check out your own contribution graph in 3d back to you, shanku.

D

Thank you katie, so there we are a complete way to build end to end join a team get into the code. Make your first contribution.

D

Get code reviewed with your peers with the help of automated ci, to keep your code and your secrets and your dependencies secure and test and ship and celebrate together now all asynchronous all remote friendly all available from anywhere in the world on github's planet scale cloud. Now software teams all across the world are using github to do incredible things. One set story comes from india's own infosys.

I

Like our most visionary clients partners, our goal is to become a completely knowledge and data driven organization with agility built into our core, so that we can sense all the business changes happening around us and continuously evolve in response at infosys. We are envisioning ourselves to be a live enterprise powered by our digital platforms.

I

Our developers, architects and practitioners are at the forefront of this, leading it for us, and our endeavor is to enhance their experience, make them highly effective and efficient in everything that they are doing.

I

Our digital platforms for productivity and the engineering tools ecosystem allows us to do this at scale and drive any change at the organization level in an accelerated manner. Our engineering tools and the continuously evolving ecosystem of home, grown, open source and off. The shelf, tooling forms an important cog in the wheel to drive agility and scale in a secured manner.

I

For instance, infosys devsecops platform, polycloud meridian leap, modernization suit are some of the examples of assets that power this for us. Github is a critical element of this ecosystem and is central to driving developer productivity, secure sdlc, inner sourcing, reusability and augmenting. The developer analytics for us, it is unifying the experience of of our developers by a variety of integrations around the robust devsecops ecosystem.

I

It is also a key enabler for a live enterprise store and marketplace where we plan to bring together the community of cloud builders on cobalt from amongst our partners, clients, open source, academia and the startups to drive contributions and consumption of various assets, as well as enhance the ecosystem with an open source. First thinking something, I believe that any forward-looking enterprise, engineering teams and developers would appreciate in a big big way, as the github team continues to drive the roadmap.

I

I look forward to seeing several of the new capabilities around security code scanning telemetry, observability, analytics and github actions that can keep on taking us ahead in our journey as a live enterprise.

D

It is exciting to see what companies like infosys and you, the developer community here in india, how you're using github every day to create amazing things. Thank you. You inspire us to keep making github better in just the last six months. We have shipped over 100 new features and improvements, including many of the features you just saw.

D

All you need to do to get started with. Github is available for free on github.com for developers and teams of any size. So if you're just learning to code or contributing to open source or starting a new team or or starting your own startup, you can do that on github for free today, all right, thank you for joining us on this tour today this was fun to do. I hope you enjoy using github as much as we love building it as it turns out we're not just building developer tools for global teams.

D

We are one ourselves. Github has always been a remote first company and we now have developers from over 15 countries working together as one team and a big part of our team is the github team in india, they're building github on github with the rest of us. So in closing it's my pleasure to introduce you to the github team here in india.

J

I am sanjay malpani. I work from hyderabad as vp of software engineering. I love our dedication and focus on building one of the greatest developer experiences.

K

Country manager of github in india, based in delhi, I love working at github because I get an opportunity to support developers, students and customers that are driving india's growth. I'm.

F

Atul malvia director product management at github, I love working in the top across catalyst for the world, build software hi.

L

This is divya vaishnavi director of product for github education, I love being at github, as I can empower the lives of students and teachers across the globe. I.

M

Am mohit enterprise solutions, specialist for asia region? I have the opportunity to do the best work of my career so far, I'm.

D

Namas and I'm the regional.

N

Sales director for github india and I just love the fact that I'm able to empower.

O

The developers with the right set of tools to become.

E

O

Best in them hi, this is pulkit agarwal, I'm product manager here at github, india github to me, is an open platform, a free marketplace.

P

Hi, I'm richard senior director for software engineering. I get to empower developers of today and tomorrow. This is satya.

Q

What I do and my teams do- impacts millions of developers across the world. I'm.

N

Shaishan gunsit I'm product manager for the github marketplace. Together we make the best software solutions come to life. My.

Q

Name is the head of finance for github india. I love working for github because of the amazing people around.

P

Me I'm usha product manager at github. I love working at github, because everything that I do can impact millions of developers who love to share their work and build some great software products.

P

M

Good morning, india and I'm super excited to reconnect with you all again today on day, two of github satellite india. A warm warm welcome to you all, and this is mohit enterprise solutions specialist at github joining you live from delhi and let's make some noise for the bestest co-host, divya vaishnavi.

L

Yay, namaste and hello, and welcome everyone to github satellite india 2021. I am vivia vashnavi director of product and live streaming live from hyderabad day. One was so much fun and day two has started at such a high note. I still can't get over the big programs launched yesterday. What was the highlight for you? But before we start, can I get a cup of coffee too? Please.

M

Uh-Huh diva, you will have to win it, so let me ask you something: okay, yesterday we announced a new program for the future software developers in india. What was that.

L

Oh, I got that that's the github externships program and what you know, students don't get the opportunities to get mentorship to learn on real problems, and this program actually connects the dots. It connects open source companies and students all together. So what was the big highlight for you.

M

Before that just take it away and make this coffee yours, oh.

L

Yes, thank you.

M

Yes, the big highlight for me was actually celebrating the colors, the cultures and the developer tribe here in india, but now today, it's time for day two and what an electrifying keynote yet again.

L

Oh yes, such an amazing keynote by chanco showcase of products and 100 plus features released in the last six months. So next time you go on github and find a new goodie, don't be surprised. We are active at work here code spaces to reduce your time and develop from any box anywhere on the world releases for continuous deployment. Obviously, discussions issues, you name it and we have it all. Folks today is actually a day of github. We share under the hood what is happening in github?

L

How github is built using github, how github uses collaboration with github and ship using github ml ops with github everything, github github, github, github, github.

M

Oh, yes, it's going to be amazing, so david, I have a small challenge for you and I am giving you a chance to win these amazing github goodies and make them all yours. But only if you answer my questions and if I answer yours, I will empty your shelves out there and will get all the goodies.

L

Let's see who does that who empties the shelves I'm up for it.

M

Okay, okay, so here is my first question for you: we announced a lovely yoga cat yesterday and we shared it with everybody. I'm sure you noticed it. You got to tell me what is the yoga pose in which yoga cat is sitting.

L

Oh, my god, is it the nindra mudra.

M

No, no, no, no, it's the lotus pose the padmasana, and with that I'm gonna get a goodie from you.

L

Okay, here you go, you've been here.

M

And I'm so excited to just kick off the day. Now I wish that we are. You were right, but audience tune in, and I'm so much excited to hear from you again on twitter.

M

So hashtag satellite tell us what you thought about the keynote, what you like the most from day, one and most importantly, what are you looking forward to and excited about in day two in day two again, we have two channels for you, one on devops, second, on open source and if you're watching us on getupsatellite.com, you can change the channels on the top go check out current and anisha, who did an amazing job yesterday and we won't mind that at all come back whenever you want and if you miss any of the talks here, you can get that on demand later on 29th march, that's karen anisha, for you.

L

And here is the attendee survey just go. We want to hear your feedback click on the link below the video and share your feedback as well as discuss with us on discussions on any topic you want to discuss during the session post. The session use discussions to discuss with us so come to discussions.

L

Talk to us with our subject matter: experts, the speakers and mohit, and I will bring questions from there live with the speakers. Yesterday's workshop were super awesome. We have like so many folks asking to register more, so folks will come back with more workshops like this, but remember today again the workshop starts at 3 p.m. If you have not signed up, go ahead and sign up so what's coming up today, mohit.

M

We got so much on both the channels at devops. We are gonna getting into mlaps, workflow, automation and github actions, and we have got speakers from github as well as some of our favorite customers. Who will be talking to you and answering all your queries on open source we'll be covering about the ci workflows ai for the web and how to do github like a boss.

M

So without further ado, let's get into it.

L

Okay, so the first session is overview of how we build our platform or sneak pew into our engineering culture and development in india, and it gives me immense pleasure to invite and introduce the person who's doing the talk. That's sanjay malphani. I've worked closely with him for the last five six years. No he's a developer at heart and a minor for data. Sanjay malpani is the vp of engineering software engineering at github.

L

He loves to engineer solutions to real customer problems, has worked in multiple startups and large enterprises, both in india and us and, like I said, leads the engineering team in india.

M

Remember to engage with us on discussions with that sanjay. Take it away.

J

Hi, thank you, deviant, mohit, hello. Everyone thanks for tuning in to this session and staying on to this track.

J

Today, um I'm going to talk to you about how we build github, I'm going to give you the insider's view, but before I start I am vp of software engineering. I work uh from hyderabad.

J

uh We have a large presence of you, know, engineering team here and we'll talk about it a little bit more. So in today's talk, what I'm going to share is our engineering practices, our tooling and our automation.

J

This will probably give you some hints and some of the practices that you can use, building your own software using github. Apart from that, if you really like what we do and the exciting work that we do, you can always reach out to me on my handle later on.

J

But before we get started about how we build github, let me actually give you a sense of the scale at which these software have to operate. The best way to start about the scale is to sort of know how many developers are using it.

J

So you do know that the github is the home for developer across the world. What does it really translate to? What it translates to is that we have 56 million developers that work and use github?

J

They have created hundreds of 100 million repos and by the way this number is, uh you know, continues to grow now, when you look at it from these numbers perspective, what is the infrastructure that we use to sort of support these many developers?

J

We are running this stuff, our infrastructure, which is just I'm talking about github.com our public site. It uses close to 50 000 cores of cpu. It hosts about uh 26 000 cubitis pods.

J

So what it really means is that whatever you do, whatever work that we do and it needs to run at the scale- and it goes out to so many developers- isn't that cool by the way. This numbers that I'm talking about is expected to grow from 56 million developer to 100 million by 2025.

J

Isn't it cool the kind of scale that we need to deal with? Now? That's just the scale of how it is being used, but let me dive deeper. Now and tell you that you know from an engineer who is at github, what does it really mean? So we are about thousand plus engineers who are working on this product simultaneously, ashtangku talked about across the world.

J

We, this translates to about 600 000 builds in a month, and that takes about half an hour to build in 98 percentile of the time we care about how that experience is for the developer. So we also monitor what is the reliability of this build system that we have once you've done the build?

J

You know we also watch. How long does it take to deploy these onto our production site? So in the last month we made 535 deployments into our production site.

J

These 535 deployments were the results of 4 000 plus prs that were generated to be deployed onto the production. So you can see that you know look not we sort of seem to do some magic to make it into file 35 deployment. We actually create a train which combines multiple pr's and then deploy them together.

J

This entire deployment finishes in two and a half hours to 2.8 hours in the 90th percentile, so you can see that a pr from build to deploy for 90th percentile of a time takes about three and a half hours. Amazing, isn't it as a developer? I start working on something and I'm ready and that's available on production to all all of my users in three and a half hours. That's really is the is the kind of then we develop we deploy under production and then we operate. This live site and this loop continues.

J

The biggest focus in this loop is that we want to make sure that it's reliable, it is secure and it is fast, which means that I can repeat, you know quicker, so you saw that right that you know from develop to deploy. It only take three and a half hours, and this is something we are not happy. We want to continue to improve on it.

J

The other thing about this devops cycle is: it's focused on developer on team, and so we have made sure that this cycle can be run independently with each developer and each team that empowers them right. That empowers them to sort of say that whenever they are ready, they can deploy- and this is this just creates a very different kind of energy in the team.

J

We also make sure that everything is written down, so that increase the trust in the people in the developer, right and- and it really creates a kind of collaboration built on a trust, and we have made sure that we also have tools that we are using processes that we're using that encourages this.

J

This collaboration, we use slack, we use github to sort of document, our you know discussions, and you will see this in the examples as I go forward.

J

So let me start with planning first, so our planning, the first thing that we start off with, is a public roadmap.

J

So you can go to our github public website and you will find that we have put out a public roadmap. This public roadmap is our intent to let our customers know what we are planning to release in in coming quarter or few quarters down the line.

J

It's very important because our customers depend depend on us on our features that we are providing. So I have only picked up a one quarter worth of you know small slice, but you can go to site and look at others. This is backed by the okr, which is an internal process.

J

The objective and key results is what every team does every quarter so every quarter we come together as a company and decide. What would we do for that quarter? What are the objectives and the key results that we want to achieve when all the teams come together? It's also easier to make sure that the dependencies with each other is also sorted out. So in this example, the objective is that we want to make sure that student developers make github global campus to home.

J

This objective have a clear key results defined and, as you can see, each of these objectives has a particular person who's responsible for it, and then we continue to track this over the quarter every month. We score ourselves course correct. Wherever is needed, this okr then translates into initiative. This is really the kind of work that we will end up doing to make sure that those key results are uh something that we can achieve from these initiatives.

J

We translate them down to epics. Epics are the the smallest piece of work that we bring it down to the customer, so you can see when you look at it that we're being very intentional.

J

We want to make sure that we tell customers what we want to do and that intentionality is backed up by a process in okr which is written down, complete transparency. Each team knows what the other team is doing and as well. It creates a shared aligned goals and in initiative in epics, we have made sure that this world is divided into small enough chunk. That can be validated on its own and course corrected, and you also notice that we create a lot of high prioritizations.

J

We are focused on only fewer things that really matters now that I talked about planning there is one thing that happens is that at the end of the day, end of the end of the deployment we also have every month something called the change log. I didn't show that, because look most of the times, what we start off is not necessarily all the time it gets to the chain block just like any other software.

J

Well, let me talk about the develop phase now now that we have a plan now the developer knows what needs to be done, those initiated, so those epics get translated into the features. So in this example, the epic is to sort of uh take our ms team's integration with github generally available.

J

So in that epic, we need to do one feature called reminders uh for both personal and the team, so you can see that very clearly in the future we identify the engineer, who's responsible, the product manager, who's responsible and the designer was responsible to work on this feature, these three form a team to deliver a feature there are such there are success, criteria that are very clearly defined and in most cases you'll also see that you know what are the features or user stories that we need to do to complete this feature.

J

In some cases when it is not clear, we start things with the spike, which really helps us identify everything that is required for this feature.

J

So, in this case, once the spikes were done, all other work and the user stories were identified very clearly, but as the spikes were going on the way the theme would work is that continue to document their finding and this documentation really happens either on the issue or it happens in the repo itself, and this provides a great means for everybody in the team to participate, in fact outside the team as well, because everything is visible to everybody, and you can see that some of the conversation here that is going on and so not only that everybody got to participate.

J

We also have now historical documentation that anybody can come in and understand the context in which the decision was made, and then they can actually take the next sort of decision uh appropriately.

J

Once these things are sorted out, a user story is then converted into tasks very clear end goals, testing everything identified and eventually that results into a pr where team can collaborate further to provide the feedback to make sure that the developer is not not missing out something important once you've done that the pr is ready.

J

Now we are trying to sort of take this pr to the deployment, so the first thing that we do is to sort of uh test it right and we have lots of tests and which continues to grow, but I mean the number you know roughly today is about five seven thousand test suite in the 5000 files, and this thing keeps growing so now. This is the very important thing when you see this pr process is a very comprehensive pr process.

J

The reason we have this such a comprehensive pr process is to make sure that we validate, but at the same time keep the time for validation to be short, and so that's why we turn off multiple parallel ci jobs. We have also divided ci jobs between what is must to do before the pr is related to deployment to what can be done in a deferred way. So, as you can see, we start required ci job which needs to complete and then long-running ci job once the required ci jobs are complete.

J

This pull request is ready for the deployment and I'll talk about deployment in the next slide.

J

But, as I said, we want to make sure that we continue to validate all other scenarios and that's where our long-running ci job continues and the way it happens is the developers need to continue to monitor and see if there are any issues and they quickly get to either fixing the test or a code that gets deployed immediately.

J

But if there is some issue with the deployment and the and the in in the ci- and uh you know, developer has not been able to sort of address it in a stipulated time we get into what we call is the compliance or a corrective path. Where you know we stop the entire pipeline, stop all the deployments and all hands are on deck to make sure that we fix the problem uh immediately.

J

It could be rolling forward or it could be rolling back. The change which has caused the issue now that I've talked to you about what the pr process is and the pr is ready for the deployment. Now, let me take you through how does that really work?

J

So the deployment process at the github is a is a safe, rollout process. What we really do is once the pr is ready. We only deploy it on the two percent users, which means that the change can only be seen by two percent users from there we move to exposing that to 20 of the user, and once we find everything is fine, we really deploy it for the every user and then the pr is ready for the merge. But as we are going through this phases, we, the developer, gets a time to validate.

J

If the changes that they put out has a has expected impact, and that expected impact would be on two types. One is that you know you're looking at your business metrics and trying to see that you know they have the right right impact. Does it is it meeting our business requirement? The second one is that we look for exceptions or a quality related thing to make sure that the change is not really causing any issue in the production for our customer.

J

Are you know time based, which means that you know we go from one stage to the other stage, but the developer has a full control if they want to sort of, uh you know hold for some more time so that they can do their manual testing if they were not able to complete within the time or if they find some issue, they can roll back from any point of view over the year over the time we actually introduced this middle uh stage with the 20 calorie, because we found that two percent uh canary does find issues, but there are times that you know the corner cases.

J

Those issues may not be. You know visible with two percent and we found that 20 actually works better, and hence we introduced this uh another stage. So, combined with this two stages really helps us make sure that uh we are. We are deploying it safely that our customers get to use the features soon, but we ensure that it's not disturbing, or there are no quality issues in our deployment.

J

Now that I've told you about how we plan develop and deploy and roll out. Let me now talk about what do we do when every with our production site, like any other production site, we monitor it? We use data dock for metrics. We use sentry for our exceptions in splunk for our logs.

J

We have made sure that we have built good amount of alerts around these, so that you know we get to know the any issue on the live site early on and these alerts um are routed to either page of duty, which you know uh which, which wakes up the person on call and the engineer on the call takes on to that issue that has come in. We make sure that it is, uh you know, done such a way that the appropriate team is actually get involved.

J

Now there are other issues that may not be of that severity. We continue to track them on the slack and in our day, time address those issues as and when they come. But these.

E

Are all about when.

J

The actually the problem happens on the production side, but we also continue to monitor our services what we call a service level objective.

J

In this case, the classroom microservice have defined two service level objective, that is exception in the health, and they have a clear objective goal that they want to achieve, and it is get monitored every week, every quarter and every month, and this slo review really happens on every weekly basis at the highest level of engineering leadership.

J

The reason for doing that is to make sure that everybody is focused on making sure our live site. Our production site is always reliable and available. We also do it at the highest level to make sure that we identify these patterns uh from from any other service, but apply to everywhere else, so that customers don't find the same issues again.

J

So I hope I've, given you a good sense of you know how we build github using github, but with any other engineering system or any other system. We continue to focus on what needs to be done further. We continuously improve this. We monitor two things very closely: the cycle time that we talked about the developer, raising the pr to the time it takes the deployment we keep identifying any opportunity there to optimize and reduce the time.

J

The second part of it that we continue to monitor is the life side, reliability, security and availability, and we continue to make sure that, as our users are increasing, that the service is holding with that scale and we make those changes ahead of time. As we start to see any issues with them.

J

We continue to use more of the github using uh building the github. uh You saw some of the things that we did, how we use issues, how we, how we use the you know the pr process to do our design reviews, uh but you know we'll continue to. We continue to evolve to use code, spaces or actions, and things like that, and eventually we have multiple skus off. uh You know, github. uh What I talked to you about is predominantly github.com, but we also have ghes.

J

We have github enterprise cloud and uh you know ghae, and uh today the developer actually has a different developer life cycle uh for each of these, and we continue to work to make sure that the same developer life cycle works for all of this queue, making sure that it's easier for developer and as well as for a customer. There is a feature parity with this. I want you to remember that you know we have built a great automation processes which empowers our engineers. It is their experiences, their productivity.

J

It really helps us get the best out of the work best best work out of them. We've created a lot of transparency, everything is documented and that transparency leads to the trust and community building and inner sourcing within our company.

J

And, finally, we measure what matters and continue to improve, and hopefully you will find some other things from here to apply to your own uh application building.

J

We continue to talk about these uh on our blog getup.blog and the categories engineering, where you can see most of these things that we continue to come back and talk to our uh you know our developers. Thank you. Thanks for uh staying with me uh in this session, we'll move to q a.

M

What an insightful talk sanjay! Thank you very much for that, especially the bits on empowering engineers and building processes around it, transparency and measuring for impact in what matters wonderful learnings for all of us here we got some amazing questions from you from our viewers and, first up is you got so much of diverse experiences across different organizations, startups enterprises?

M

What is that one thing that really stands out and you like about being at github.

J

Yeah I mean I like so many things about being at github, you're, really making me choose the one, and only one thing um you know the one thing when I joined github that was really unique was how much love you get when you wear the github branded clothing or you care about the github branded goodies. um You know it's so so I'm so proud that you know look.

J

This is what really happens, um and when you look at it you know, sometimes you wonder, is it accidental, but really it's not accidental, as I have sort of learned about it, and this is the one thing that I like about it a lot of the times I go in and talk about hey. If we do this, this is what it means to us. This is what means to our business and I get corrected whatever we do. It has to be centered around the developer. What does it mean to the developer?

J

What are we doing to improve their experience, their productivity and make them work in a global environment where there is a community? There is a trust, and so that is the biggest one thing that is different. We are totally and 100 focused only on developers.

L

I'm sure everybody hearing here would be super delighted to learn that every day we think about them so much sanjay, also from the whole journey and the workflow that you share.

L

uh What's one talent or like a cultural or a technical challenge that you would want to share with folks hearing on the devops channel as an inspiration of what they can look at and how can they overcome that.

J

Yeah, so I I github is, I mean a good one: there's lots of challenges right, I mean a company which is growing. We talked about 56 million users going to go up to 100 million in 2025, or there is no depth of challenges that we see day in day out right um and but if you think about it, what has been for the last six to 12 months? What has been a big challenge?

J

The one thing that our company is going through is from a small company to a you know, a medium-sized large company, where a lot of a lot of enterprises lot of communities depend on their day-to-day work, and what we found was that as a team, while we were doing the right set of things, each one was following their own. You know process practices, and we have now become intentional about making sure that there is a uniformity without missing out the empowerment for each team to do what they want to do.

J

So you saw the public roadmap right. This was not something that we used to do. We started doing it because you know that's really tells our customers that what they can depend on and the same thing internally you saw slu. You know we still provide each team to sort of say that what they want to measure but at the same time it is a process that we have developed now across the teams and we meet at the highest level of leadership to make sure what is happening on one side.

J

The other side learns about it and quickly reacts to it. So that's really has been the one of the biggest. uh You know, cultural changes so to say that we are going through um but, as I said, there's no earth of challenges at github.

M

Amazing to hear that and yes absolutely, uh we got one more question for you and.

J

Am I also allowed to ask you a question.

M

J

All right, you talked about how good great you felt about the talk. I mean that one number that I care the most. I want to ask you if you actually, you know paid attention to it. How long does it take for a developer when the pr is ready to deploying it to the production site.

L

I got this three and a half hours. You had said.

J

L

ah No, not every time.

J

Well, you got my question.

L

J

E

I didn't ask you about the reliability, but yes, it's three.

J

And a half hours in the 90 reliability so yeah the way you want you may want to take something from uh you know. Mohit now.

L

Yeah, I'm gonna, I'm gonna, take a goodie. Thank you so much sanjay again it was such an insightful session. I just loved having you here and sharing the whole workflow with everyone. How github is built using github folks, remember: sanjay is going to stay with us in discussions so head over to discussions and ask all your other questions from sanjay. Thank you, sanjay again. Bye. Thank.

J

You and look forward to engaging with you on the discussion.

J

L

So I just want a goodie where's, my goodie, hey.

M

You winner, I'm gonna, miss my greedy. There. You go.

L

Yay, I got it awesome. Thank you.

M

Okay, let's jump into the next session. For today, uh we are going to talk about how you can create your own github apps and your own github actions and we're going to share examples of resources to help you with those so that you can make your own action or your own app and we have got shashank bunsen in the house. The product manager for github shashank has 15 plus years of building and managing software and having lived through the transformation of multi-year releases to on-demand deployment.

M

Shashank shares a deep sense of value in sharing and collaborating a very good friend of mine. I'm really looking forward to hearing from you shashank.

L

I am also very excited about this session. It actually extends the capabilities that github gives and take it forward to make it suit your needs and is also one of the biggest ways folks come like contribute to the community.

L

So shashank over to you folks remember, engage with us on discussions mohit and I will bring the questions for you from discussions over to you.

N

Sasha, thank you. Thank you mohamed, and welcome everybody to satellite india 2021 as as mohit and shared. What we're going to talk today is how you can create and share your own github, apps and github actions before we get to it. My name is shashank bansil. I am a director of product management here at github, taking care of the github ecosystem and that's really the platform that enables you to contribute your solutions and help the community at large um to start a journey with.

N

Let's, let's probably ask what's in it for us, what's in it for you, why do we have the ecosystem and a core part of github's philosophy is is to enable choice. So we provide choice to end users, provide choice to you as a consumer so that you can customize your workflows. You can customize.

B

Your software development journey.

N

With the tools and services you love, it provides you a bunch of them built in, but you can bring in your own complement it enhance the workflow, make the best out of it and that choice. The is part of that choice is offered by you, so you can contribute. You can come in and help us build that choice for planet skill, developer, community and that's what the ecosystem team focuses on, enabling you to do that, and that is where apps and actions come in here.

N

While you do that, you build a brand for yourself. The github community knows you as the developer as the github handle.

N

Add value to software development across the board and build your own business plan implies business, that's what's in it for you and that's why it's important for github as well to complete win-win making each other successful um before I before I get further into apps and actions. How about we? We try and build something. What better way to start up talk to start our session than by doing something hands-on.

N

So I picked up a.

R

N

Here the problem I picked up that we want to solve together is we saw github release in the keynote we saw github release being the platform to release what we are building. What I want to solve today is whenever a github release is created. I am in the product, I'm in the business. My software is a library, so I want to compile that library build it and upload the built artifacts automatically to my release.

N

So when I really publish the release, the built artifacts go out, that's what I want to do and I want to enable an actions. Workflow do this automatically. I don't want to do this manually.

N

I start researching about it. I come.

O

Across the cli.

N

Extremely powerful command line utility that uh works with all the github github resources, including github release. So using the gh release, create or gh release upload.

N

I can modify existing github releases and using a command line, so I decide to use that what better way than doing it than running just a command, and I sketch out my workflow, it's a high level workflow actions handle that I've sketched out, which does the build that tests, my library automatically and after it does I'm gonna, add a youtube uh added command to upload the release using ghg support.

N

But I want to focus on this part of that workflow before I can really use the github cli. I need to set that up on my runner. I need to download it install it make sure it's ready so that I can use the command, and this is what we are going to build today. We are going to build an action that does the setup that that creates this and prepares my runner to be able to run the command a quick look into the into the constituents of this command. So it consists of a repository.

N

It's a public repository in which I will share my actions code, followed by the version I want to use represented as a tag and the inputs that actually needs without further ado. Let me jump and let's start trying to build this together.

N

So I go to my browser here.

N

Just give me a maybe I need to share things again.

N

Okay, so I am in my browser and to start building my action. uh The first thing I do is there is this javascript action, template repository and and the actions workflow github has already shared this template repository to simplify you to simplify everybody, start building your own actions. I'm going to rely on this, I'm going to use this template and clone it into my own repository that that I'll modify into the action. I need make sure you mark it public. The actions repositories need to be public, so the action learners can access that.

N

I have already done that and created this repository uh in that process note. I call that example, setup gh and it's generated from the javascript action once I have this. There are really three files that I go modify to prepare my action.

N

The first file, which is important, is what is called actions.yaml. So it's it's. It's really the declaration of my action, so you know I I provide a name to my action and then I just give an input to what are the versions that I will expect any users to provide me I've, given a default value. In case the.

F

Version is not provided. I will install.

N

1.2.0, and so that's that's really the action gamma. So I update the yaml file for name description, branding.

B

N

So that's the first change that I make.

N

The second file I need to change is what is the javascript that runs my action, so basically the core heart of my actions and and update this code to really do that download and install the actions. Sdk provides me a bunch of packages, in this case I'm using core and tool cache as the two packages that I'm gonna rely on.

N

To give me a utilities to give me ready to use functions, to simplify my actions code like, for example, in this case, when I want to download from a url, I want to do girl the tool cache gives me download tool as a function, so I don't have to write my own code to actually download.

N

Similarly, I have the extract function available to do the unzip to do unzip or extract it from a card file, and then the core module gives me operating system manipulations like I finally added to the path. So when I say gh release after that, the runner finds the github client path and just by these small uh lines of code, it practically look like pseudocode written in here. I think I have my end-to-end action running. It's completely functional this node.js script. I can run outside in an action and do the job last, but not the least.

N

I want to give a good readme file to my depository so that anybody who lands on this repository understands what the action does and what it's about.

N

That's it so I I have my action right up and running. I have it built now is the time to share it further to share it. I have a hint here, but I can really share it on the marketplace. I will start creating the release. You see the github automatically passes. The actions repository understand, shows you what it will look like, and I need to give a category.

N

Let's call it utilities for now for this one and give it a release title and I can publish this release once I do that my action would be live on the marketplace like, for example, this entry action is, and the readme file that we created is automatically shown here to help end users understand my action better.

N

So that's it, we did it. We have an action build and we use the cross-platform it's cross-platform. It works on linux and windows and mac. We use community tooling. We use the libraries available. We didn't have to write a lot of code ourselves and we contributed it back. We are enhancing the entire makeup community while using what they have.

N

Why is that simple? That's a complete action built right from scratch in in 10 15 minutes that we spend um let's change tracks so now that we have built something. Let's look zoom out, look at our integration journey with github overall and, as you would uh understand, and like any software, it consists of plan built and release in terms of planning. Github provides three ways to integrate with github there is github or app and github action. We just page the github action in to understand the github, app better, a simple way to think about it.

N

It's a bond and completely autonomous execution entity that can independently uh communicate with the github resources, even driven actors that are responding to events happening.

N

You maintain and operate it on your compute and it just responds to even coming in it gets its own fine, grained permissions and and that's what helps it be very effective and powerful.

N

You probably don't have time today to go deep into building github apps, uh but I wanna point you to probot. So it's a framework available that helps you build github apps.

N

There are a set of example, github app shared with you with code available, so you head to robot.github.io and it will give you all the utilities, all the help.

C

N

To build your own github app, if you want to know more, there is a complete demo day series that we did earlier this year, it's available on youtube and on the partner center. So you can look at that. A step-by-step video, step-by-step guide that helps you build big web apps further.

N

uh Second up, let's talk about what apps, so what apps? I call them as authorizations for third-party services and clients anywhere that you need to say login as github, and then the client and the um and the services can operate as the user. That's where we use github app or sorry. Excuse me, uh examples being you want to plug in an ide and say hey. I want my code to be synced up.

N

So whenever there's github desktop kit of cli vs code, all of those use what apps behind the scenes they are not as powerful as github apps. They don't get fine grade permissions, it's really the user permission, that's getting limit and given to the client but very effective when it's just about logging, in, like a user operating and and last up.

F

Github actions.

N

Essentially, I call them as independent units of work that can run effectlessly, execute on computes not owned by you, so they will run it on github runners that could be owned by github or the customer to do an explicit amount of work like we set up the github cli, and then we had another action to upload the release. So there are two good examples of actions.

N

Well, how end users use it? Is they chain together a series of these actions for an end-to-end purpose? So that's that's really good actions for you, anecdotally.

N

This is a view of volume between these three we see most github actions being billed in terms of numbers on the marketplace are close to 8, 000, guitar actions and 400 gigabytes, and that's not a testimony of their power. That's just a testimony of the amount of effort it takes to build them the amount of operational cost it takes to run them.

N

Arguably, github apps are way more powerful way more. uh You can do more amount of things you can bring in your own compute. You can bring in your own logic that can help developers, but github actions are those simple things that that help you get started quickly and interact with an external system via apis or cli.

C

N

Yeah, that's that's really the smallest number of the lord uh but they're very effective for the purpose they have um you. May you may ask me okay great, I understand all three but but which is the one that I should use. So I'm going to leave you with this small cheat sheet of sorts, comparing github actions with github apps right. Both of them are event driven actors, pickup action, workflows, triggering when events happen. It have apps respond to events on github, uh but how do they compare is essentially on the screen?

N

Let me just quickly walk through.

B

N

Github actions run on github runners, compute not owned by you, so whenever your solution can practically run independently on.

E

N

You would use that they interact with other systems outside github via api of cli. It's not a lot of heavyweight code.

J

N

Within the action, whereas github apps are fully functional, web apps, fully functional solutions that you can host on the cloud and respond to events from github. So for long.

E

N

For more heavyweight solutions, database or simple lightweight interactions can have actions. Github actions run headlessly, they don't have a ui, whereas in most cases what we have seen is github apps bring their own ui. Once you post a web app, you also bring in a web page with it, which can have user interaction which can show you results. We can add value.

B

N

The processing that you get most github action workflows are stateless. You don't store a lot of data. You don't have data storage beyond what you can interact over api and cli, but with github apps. There.

B

Are databases at play.

N

There's a lot of storage. So that's! Basically your solution ends up having persistent data. The most critical part of the difference. Github apps works with a fixed set of permissions decided by github. So we pass a github token to the action and that token has a fixed set of permissions. It can run on any execution workflow. So any elector which includes your action, will be able to run that github apps have a fine grade permission model. You can really say what permission does my app.

B

N

And build that confidence within consumers. You need not have all.

B

The permissions.

N

That you need to have and there's an administrative workflow enterprises love that to say what I it can only run on repositories after it's installed by an item, so more control for administrators there and last, but not the least github apps are standalone. They do one thing at a time. Actions can be changed together to fulfill a bigger purpose.

N

The next part of our journey, the build journey so github gives you a set of resources out there in terms of docs sdks apis. uh I want to particularly highlight the sdk, so there's opticate.github.io and you land there. You will find a bunch of sdks managed by the community, so the go sdk to the python sdk that help that simplify how you can interact with github and empower your solutions, not just that. So there is partner center at github. So it's available on partner.github.com.

N

It is a collation of resources, a collection of all the uh examples, all the presentations that you have to help you build integrity and extent. So it's built together by partners like you, so you can go, join the partner program and get access to all the resources collated across the world and interact with other partners as well.

N

Let's move to the last part of the journey, so we have planned what we want to build. We have built it using the resources that we have now is the time to release. So we looked at it earlier with the actions. That's where the marketplace so github provides what we call the github marketplace.

N

I call it the window to the world for your tools whenever you have a solution, whenever you have a tool and you want to share it, the marketplace is the means to do that. There are 400 plus apps and close to 7600 actions on marketplace already and growing over time.

N

So you list your listing and listing.

B

On marketplace,.

N

Is pretty simple? All you have to do is fill up the description for your app in action and it goes live on the marketplace like you saw not just that you can also do a sense of verification for your organization.

N

It's a means for you to really get credit, really a test on your apps and actions being built by you.

N

You see it's easy on github for somebody to say I publish this, and the name of the handle might look very similar to your new handle.

N

You might get confused, differentiating that so by.

B

N

Publisher verification, you get the credit you make sure your organizational identity is verified by github and that tracks transfer to your apps and actions as well. Giving you the extra level of credibility that extra level of proof of things.

N

uh And that is what we had today, so thank you for joining satellite. Thank you for coming and listening to me. I hope I was able to help you with understanding the layout of apps and actions and and hope to see your apps actions come on the marketplace. Thank you.

L

Thank you sashank. This was great layoff plan of apps and actions. Let's start with questions, so the first question that I think folks would want to know is any github integrations that you really recommend.

L

N

See as github my our platform is really to enable all of you get the best out of it right. So when you write a tool or solution, it's as important as the one github root itself, so my recommendation would not be based on use a solution, a versus b head to the marketplace, search for what you need. If you need a ci to just look for ci solutions on the marketplace and choose for yourself, we are an open source company.

N

We enable everyone to get a fair uh lay of the land, and I want you to make your own choices based on what works for you. So I'm not going to name integration, but I'm going to preach what you should do.

M

Wow, that's pretty smart and we got another one on the same lines for you shank. How do I know about quality of an app.

N

Yeah great question, so that's actually something that comes to me very often. Right like there are like I said there are 400 apps, 7 000 actions, which one is better than others, and and you see I going by the open source principles again, it's really about you rating them. You waiting them to say which one works better, which one gives you the best support.

N

At the end of the day, you want to use an app or an action, that's functional and supported that we get an update in case you hit an issue on which is what we try for. So if you hit the marketplace, you will.

N

First of all, like I see the publisher verification badge that I talked about, which tells you that the publisher is genuine, that publisher's identity is verified by data if you drop them in email, if you reach out to them, you will get a response, and, besides that, there are indicators like stars indicators like update time that you can see and just like you would use an open source library. You should look at apps and actions as open source libraries to choose from, and and that's what I mean by you- make your own decisions.

N

J

With information.

N

But decision is really ours.

L

So we spoke about uh from a consumer point of view when I'm using or wanting to use, uh apps and actions. If I think, from a contributor point of view or as a publisher, uh what are the ways that I can ensure my app or my action gets the maximum footfall or more people get to use it.

N

Yeah, so um pretty much the first okay I'll, maybe be preachy here on this one.

N

If you are writing an app or an action, my I would advocate you to build a good listing page write, a good readme for your action, write a good listing page with screenshots, so that end users can understand what you do can give you feedback on what you do maintain a repository on github, where consumers can open issues against your app or action can give you feedback through that means and last, but not the least, get that publisher verification back.

N

So there is a minimum level of curation that github can help you with by doing those three things. Besides the regular ones of more front doors for yours so blogging about it, writing about it. You get more eyeballs on it and by giving the right level of support by helping the community, you will be the popular actor.

M

Thanks shashank, you were very loud and clear that we need to stick to the fundamentals and thanks a lot for the insightful answers to all the questions. Viewers shashank is not going anywhere. He's gonna, stick with you for another 30 minutes, but in get up discussions, so you know where to catch him and get all your queries answered. Shashank, we'll see you next time. Thank you.

M

L

So much good stuff so far, but there is more to come. Coming up on devops channel is a session on mlogs and how github uses github to collaborate later on the oss channel is how to get up like a boss, as well as the state of ui and ux in open source. Both channels will have a great closing sessions from our very own manish.

M

Sharma and we are loving, all the active all the comments, all the activity that is happening on twitter right now, so keep tweeting to us at github, india, hashtag, github, satellite and tell us what you are most excited about. Seeing today tell us your favorite moments. We are all ears to hear what you gotta say, and we are gonna comment on everything that you're talking about.

L

Yeah, or also maybe share how you're watching us so share snippets of where you are where you're watching us from so moving to the next session is a session, a lightning talk about accelerating inner, develop with kubernetes applications with scaffold and how scaffold can help you create a production ready, ci cd pipeline sharing. All about that is ashish chaudhry. A senior developer from city he's an active blogger, a technical writer who obviously loves java spring devops and the cloud, and he is a big advocate of open source technologies.

M

Remember to engage with us on discussions with that ashish take it.

B

B

S

So uh hi hi everyone uh welcome to get up satellite. uh My name is ashish chaudhary and thanks for such a great introduction, and so since this is a lightning talk I will be. I have to be a little bit quick, but I will also make sure that you understand the concept that I'm trying to convey the message in this lightning talk, so the topic of this lighting top is excel. How you can use a tool called as a scaffold. This is an open source tool developed by google.

S

So you know you can accelerate, accelerate your inner development loop and, as uh so as uh as it was explained in in the introduction uh that I was for a financial services company city group. I worked there as a senior developer and, let's just uh you know quickly, go through the agenda for this talk. So this is what we are going to cover in this presentation.

S

uh So first we will talk about application, development, inner and outer loop, and then we will talk about related to one of the concept is: how is the current developer experience with kubernetes and how scaffold can help that you know improving that developer.

S

Experience and then we will definitely talk about github actions and there will be a demo on how you can build and deploy a java application using github actions and scaffold, and then there will be some references which you can after this talk, you can go through for your uh to get some understanding on the topic, and so this is pretty much we are going to cover today. So let's get started.

S

So let me uh talk about the application, development, inner and outer loop. So this is not a new concept. The term may look like a bit new, but the way we have been developing application over the years as a developer. It's pretty much the same so this so when you start, you know working on an application or you work on a story item, so you you are in your inner development tool. So basically, this is where you are. You know you do. You are doing coding.

S

You are building testing and verifying everything in your local if possible. In your case, and so most of the time we do. We we as a developer, spend time in the inner development group and then obviously, when you know, uh you feel comfortable that your changes are fine. You can share your changes with other you.

S

Do a git get push and merge your pull request and then your entire cicd pipeline comes with a picture. So this is where your you know: outer development comes into the picture and obviously you are doing bunch of things there. You are running your integration, functional test and your sonar, sonar checks, code, quality checks and then finally, the application gets deployed so with this being in the we've been doing this for all the years. So how does it changes uh when kubernetes come come into the picture? So uh so this is.

S

uh These are some of the steps uh that uh we as a developer, you know typically have to go through once we start to. You know, thinking about making application cloud native so so like, for instance, if you are, um you have to build that, build your application image.

S

You have to tag it and then push it, and obviously you have to create your kubernetes manifest and if there is any change- and so you and you have to manually- update in most of the cases your kubernetes manifest, so that it points to the correct image version and then obviously you have to deploy and test, and you know verify your changes.

S

So if anything goes wrong in any of these steps, and if you see your changes are not correct or if you you have to so, you have to repeat that entire lifecycle, so there is uh so kubernetes introduce lot of complexity into the development lifecycle and related to that is like. So, if, if we as a developer like how do we feel about the developer, experience that we have with kubernetes in general?

S

So let's talk about the kubernetes experience that we as a developer feel whether how how we feel while interacting with the communities in general as a developer.

S

So as a developer, I'm the primary thing that I am interested in you know I'm I want to build, build the product or add futures ahead and exciting new features to the product, or something like that. So most of this is related or revolves around coding and as a developer, I might not be interested in knowing kubernetes.

S

So even though it will definitely be going a good skill skill set to have but do as a developer, do I really need to know how my application is going to be containerized or where it is going to be deployed. So kubernetes is definitely a great tool, but there are some hardest that we as a developer, need to understand. We also need to understand that developers are not kubernetes experts, and this is you know, one of you know fun way of looking at it.

S

So if some, if someone say that developing application with kubernetes is easy, so uh I I will respectfully disagree. It is not easy, so no uh so as we have seen in the previous slide, so there are a lot of you know additional steps added. So with that being in picture, um it is definitely not easy, so we as a developer, uh we need better tooling, so that can help us to you know, handle this lot of boilerplate.

S

Part of you know, building pushing and deploying so there has to be some bad, better, tooling, around instant. Instead of we going, you know into the uh reinvent wheel, sort of mindset and develop things of of our own. So this is where scaffold comes into the picture and it it helps developer in automating. A lot of those stuffs, and so scaffold is really a powerful tool. It's an open source tool.

S

You can use it if you are facing such challenges and the primary primarily the problem that it solves is like you know how you can uh handle the workflow of your.

S

So typically, your build push deploy cycle gets automated with help of scaffold and as you as a developer, you can focus on the most important part, which is writing code. So you can go back and focus on. You know solving real uh real life problems and you know all those things, and so, let's get into little bit into scaffold like uh party the architecture of scaffold.

S

So, as you can see, scaffold and the main. uh The thing that I like most about scaffold is that it has a pluggable architecture. You can pick a build tool of build or deploy tool of your choice. You may be using uh docker in your local and then you might be using something build back or jab, or maybe one plugin for your deployment and in your local. You might be using cube cutter for deployment and then for your production scenario. You are using helm, so it has all those features.

S

You know you, you can pretty much play with any build tool of your choice and it has its own tagging strategy to tag the images and so let's now, let's talk about like how scaffold works like what are the things that you know how it the two as a tool it works. So if you look at this diagram, uh so at the bottom, if you see there are two commands, scaffold, dev and scaffold run. So these are the most uh used command with scaffold.

S

So when you run scaffold, what actually happens is that it looks it what it has its own file watcher mechanism. So it looks for any changes in your source code and if it feels that there is a need to compile, then it will do the whole compilation of the application. Otherwise it has the you know capability or it's it has that intelligence so that it can not only just sync the file and restart your code and it can skip the build and push part.

S

So this is, and this can accelerate your inner development loop and obviously you can not only use for local development. You can do it for remote and remote deployment like for. If you are using gke, for instance, then you can definitely use carbon run command.

S

So scaffold run is a sort of you know: end-to-end city pipeline sort of command, so it runs your entire workflow at once, while scaffold dev is a continuous mode command in which it also stream logs, which is a really cool feature, and once your ports are started, you can in your local. You can definitely see their application is get started and up and running.

S

So this is how scaffold uh typical workflow with scaffold is, and- and this is the scaffold configuration files which, which is how stuff would understand like what it need to do. When you run the scaffold, dab or stop hold run command.

S

So, for instance, in this image you can see that I'm using jib as a maven plugin, so jib allows you to containerize your application. uh So you input is your source code and the output is your uh containerized image and then with cube catal we are deploying uh we have. So this is how scaffold configuration works so light camera and github actions.

S

So github actions is, I think, one of the cool feature that, within the comfort of your github repository, you can create your ci cd pipeline and there is a there is a great market place available for you know to choose any date of action available there based upon your need. So before you know I jump to my demo. I just wanted to give a quick uh understanding about the how data vaccines are. What are the main components in that? So it's mainly work on the event mechanism, uh it whenever any.

S

For example, whenever someone there is any pull request or push changes done on your branch, then it triggers your workflow so workflows. Typically, you have a ml syntax file in which you define your jobs, and then you have some steps and actions which gets executed. Actions are basically such as whether you wanted to check out the repository or install some tool like, for example, java and then.

S

Finally, there is a runner which is your server instance, where everything gets run, uh this entire workflow get executed and it and you can see the it you can see the logs in the data repository itself, whether the actions are successful or not, and so this is a really cool way of you know creating cicd pipeline from your comfort of your id. So next I wanted to give a quick demo about how you can you know, create ci cd pipe pipeline with the comfort of your id comfort of your github repository.

S

So so this is my data repository and this repository uh is a pringle trust application and uh it exposes states endpoint and once you hit the end point, it display indian states and their capitals nothing fancy here. So let me just give you a quick walkthrough of the yaml file, so so this ml file is typically in your git dot, github uh root directory, and then these are some of the prerequisite required like you need a since.

S

We are deploying 2gke, you need a google cloud project and container registry and kubernetes engine api needs to be enabled, and obviously you need to have a gta cluster which I already have, and then you need some specific roles for you for you to be enabled. You know so that you can push or deploy to using your service account key service account key is going to be used as a id for this deployment. Think of it like as a functional id or something so- and this is my gta cluster.

S

Yes, it's up and running. There are two nodes and uh yeah, so this is the clip catal command. We will. This is the cloud cell environment, and this is my container registry. There is no image available there, but after this deployment there will be and these in the environment section you see that we have defined some secrets and these secrets are available in the github defined itself in the github repository in the in the secret section and in the job section.

S

As you can see, it runs on ubuntu latest, which means it's on on the ubuntu machine and the job. There are steps like checking out uh repository for master branch, then I'm installing java, so all the dependencies that are required to make this action work is listed over here. Obviously, I need google cloud sdk so that I can communicate with my gpa cluster and it requires some service account key and project id which are injected as secret at the runtime and then obviously, I need uh scaffold to be run.

S

The final command- and if you see at the bottom there is a scaffold run command. So this is this will eventually going to kick our entire workflow and then I will just verify everything with the get all command.

S

So um so what I will try to do now is, I will, you know, create a pull request, and so in this uh I'm just going. This pull request will eventually create an event and I'm just merging my changes from develop to master brand. Actually it should be main branch. uh Sorry about that.

S

So I have just created the pull request, and uh now uh I I so these are the changes that I have done. These are mostly the commands. I have added nothing fancy here and yeah, so I can now safely merge the pull request.

S

So once I confirm the merge, we will soon we will head to the actions tab and you can see that action that the workflow has kicked in. It is based upon the event and it has started the deployment. So it will step by step, do all the steps that we have defined in the workflow ml file. So it has already checked out the code installed java, and uh so it will do everything.

S

What we have you know described in the workflow ml file, and uh so it's trying to download all the dependencies required to make this work and so yeah just taking its own tweet time. So it's able to connect cluster yeah. So this is the main, uh so this scaffold run what it it will do now. It will build the image containerize our application and also push it to a container registry to our gcr registry. So it has started the build right now.

S

So it is downloading the maven dependency and soon after that it will uh create a containerized image of our application, and once that is done, it will push and deploy to gk, using everything using scaffold, so scaffold, run, command c is doing everything here. So, as you can see, the image is created and, as you can see, parts are also started, so we will quickly check whether the image is pushed through gcr or not yeah, so image is already pushed, so we are good.

S

I think the workflow is completed and everything looks green, which is really great, and now uh one of the verification will do in cloud shell. We will just run the same queue cuttle command, and this should give us the correct output, and so what I'm looking for is the external ip, because the service type that I have used is load balancer, so that it is exposed for internal uh sorry, external access, so I will just copy the ip address and just hit the url and yeah. So this is the expected output.

S

This is pretty much we were expecting and this actually completes our demo. So what we have learned is in this demo is that we have deployed the application to uh gke cluster and- and this is the entire csd workflow we have created just now. So these are some of the references that you can go through after this lightning talk. The source code is available and one article that you can go through if you wanted to know more about scaffold and some scaffold, documentation and scaffold data repository.

S

If you want to go through- and these are some other way you can connect with me- I'm available on twitter.

S

If you want to follow me for some awesome java tech related stuff- and um this is my github username and my blog url, and also available on social, I'm in the net professional networking site uh linkedin- so thank you. Thank you very much. uh I'm really grateful for this opportunity and I wanted to thank the entire uh github team.

S

So thank you. So I will see you uh in the discussions now so still then yeah thank.

S

L

Wow, that was a great lightning talk, isn't it connecting inner loop with the outer loop, as well as how scaffold can help with your complete ci cd pipeline? I hope folks, a lot of you got insights on how you can do that too. Remember. Ashish is with us in discussions, so go ahead and connect with him connect with others who are using kubernetes and scaffold and learn from there. On discussions mohit, I had one question for you. I know you are a solutions specialist at github.

L

Why not share an insight and if you do it well, I'll give you a goodie share. An insight. Tell us about a specific industry which has been able to leverage github like for unique use cases absolutely.

M

Absolutely okay, so I'm going to tell you a specific use case for financial services industry, uh so, typically the the banks and financial bfsi customers in india.

M

They lean on a lot of isb solutions right who build specific ip for them, and what happens is that the banks and fsi customers have a requirement to collaborate with not just the developers within their own organization, but also with the developers in the isvs organizations right and hence you can guess now that the requirement for them is to really provide these two set of developers a secure platform to collaborate on the piece of code that is getting developed and for that github enterprise is the perfect fitting solution.

L

Awesome you get a huge for that.

L

I don't want to get rid of the humor.

M

Yay we're gonna get into our next session.

M

A second rapid fire, lightning talk on workflow dispatch and how the automation has been in the past present and future story about discovering a faster way to develop internal tools with github actions using workflow dispatch. We got vishnu bharti in the house. He is a software engineer at hasura. He loves, building developer tools and have spent a lot of years of his career, doing that at scale. He regularly blogs at vishnu, bharti dot codes and guess what he munches chocolates.

M

While coding that's the secret to the wonderful software that he is writing and we're gonna hear right from vishnu. Vishnu super excited to hear what you have got in store for us here. It is take it away.

T

Hello hi, I hope you are having a great day at satellite. Today we are going to speak about workflow dispatch in github actions and how it's helping to do automation for the past present and future a little bit about myself. I am vishnu bharati. I currently work as a engineer at kasura building, developer tools for developers who are building developer tools, and you can find me on the internet as cripnal and here's a really fancy picture of me having a having power rangers, wallpaper and bsnl modem back in the 2000s all right.

T

So today's agenda is we'll first see what workflow dispatch in github action says, we'll see some of the internals and devtools at asura, uh then we'll talk about what I call the past present and future problem and uh solution to it.

T

What is workflow dispatch uh workflow dispatch is a special kind of workflow in github actions, which you can write that will take a set of inputs and run it manually whenever you like it from the github ui. So if you are familiar with datum actions, you might have seen this ui, where you can see the workflow runs for any event that triggers a workflow but workflow dispatch. Has this special uh ui, where you can provide a set of inputs that you like and press the run workflow button to run the workflows whenever you like cool.

T

So it's kind of like a manual thing that you trigger right. uh You might ask what it has to do with automation, but yeah. We will kind of see what the kind of power it gives okay. So how do we write the workflow dispatch?

T

First, it's like mentioning it's like any other uh workflow workflows that you write in your github actions. You just create a vml file and have a name and in the on section you will say workflow dispatch. uh Usually if it's like a pull request or something some other event, you will mention the event here right. Instead of that, we will mention workflow dispatch and give a set of inputs.

T

The input section here will, whatever you are giving in this input sections, will exactly translate to that in the ui, uh where your users can actually uh give their inputs and after that, it's the normal thing where you can give a bunch of logic that, based on your inputs, that your data workflow can execute so here. So this this piece of yaml, I took it from our protection code, which we use to trigger manually, trigger.

T

Manually trigger previews for pull requests, uh so we, the inputs that it takes, is a link to the pull request and the commission from which, which should be taken for the preview app to deploy. If you leave it empty, it will just take the latest commission, so you can kind of mention all these kind of conditions there and the workflow takes it and deploys a review app for that pull, request and comments on the pull request like this, you might be wondering what to do with these kind of things.

T

Right, like you might you you can just trigger the job whenever you want, but the thing is, once you have this core logic uh that you can run whenever you like, you can build on top of it. For example, uh we we built a slash command kind of interface on top of products patch. Basically, whenever someone comments, slash hiroko, deploy on a pull request, we would actually trigger the workflow dispatch via the github api and the workflow dispatch, runs and comments like this.

T

That's why we had the commands there, so this kind of workflow is usually normal in azure, where we kind of comment stuff on the pull request itself to get like immediate feedback.

T

Okay, so here are some lists of internal tools such as we have a review app for pull requests. We have a project for shadow over which we'll be seeing. Next, uh we have some more bots. uh If you see most of them are written as workflow dispatchers.

T

uh You might be wondering why right that's uh to understand that we will need to understand uh how we were writing a bots and internal tools before before workflow dispatches what we used to do is we used to have a how github web books configured and whenever we want a internal tool, for example, let's say we want a tool that deploys preview app software request. We will write a node.js server or any other server and host it on somewhere.

T

For example, we had a we have on here: node.js server hosted on versal and we had another tool uh written in python and hosted on kubernetes.

T

So if you want to add something new, you you either need to go and edit them and or create a new server and host it on something and configure your books to be received from there. The thing with this is all these environments might be limiting, for example, it might be hard to do. Docker builds on node.js server right or running on version, uh so because it's it's stateless, so the the whole the it was causing kind of friction where you need to do all this hosting and creating new servers every time.

T

So this is the current state of internal tools. uh We, whenever we, whenever someone wants to write an internal tool, we just write a workflow dispatch first and manually like give it to some some developers who are interested in running it, and they will run it. They can trigger it from the browser and, at the same time, uh once we are okay with the automation.

T

What we will do is we just run one server on version that kind of handles all the books and kind of delegates. The requests to workflow dispatchers.

T

So this kind of workflow, if you want to add a new automation, you don't need to write a new server. Instead, you just write a yaml file and just run a workflow dispatch within an instant like you can start triggering it manually or whenever you are feeling confident you can adjust merge apr in your delegation, server that triggers them for your web books.

T

Okay up next is the what I call the past present in the future problem. uh We'll try to understand this by uh seeing a scenario that happens at hashtag every day. uh So asura is a has a graphql engine open source project which is being like used by a lot of people, uh and also we, we have a commercial product which uses the uh open source graphql engine code inside that in order to make sure that uh changes made in open source does not break anything in our commercial product. uh We we have to like.

T

We have a monorepo which actually syncs code from uh the open source project and kind of syncs it back. We will see the workflow here whenever open source contributor tries to contribute to graphql engine, they just open a pr first.

T

After that, what we do is we have a workflow dispatch that shadows their pull request. Shadowing essentially means uh transforms their code so that it's in the right places uh in our mono report directory structure uh and opens pr uh on the moon. Repo. This monopoly is internal and private. uh So what happens? Next is our hashrate member uh reviews, the code and the ci runs, uh and if and once, and if the change seem to break our commercial version, uh we'll fix those things the asura.

T

Basically, the team member will be working on the shadow appear once the uh that's merged back to the monorepo. It mean we already. We have a push workflow in github actions that uh sync back sync back the commit only to the graphql engine, oss repo, and we have a bot that says thank you to the open source contributor. So this is the whole workflow where the code comes in from the oss and that gets into our private repo and goes back into our graphql engine voices right.

T

So if you take this tool, uh the past is like you know, whenever whenever I want to create this tool, the past is like I want to work, make this tool work for already open, pull requests, and I want to open uh make this tool work for uh the early adopters. uh Whenever I'm building that tool, uh I might be getting feedback instantly and the ultimate goal is to automate it whenever some new pull requests come in uh before we had that automation in place the product or shadow in place. We want to do that.

T

uh First, we naively went with a pull request approach and figured out that it kind of uh restricts us from some things. The thing is, if I'm doing on pull request, you can just shadow pull requests that are coming in after the uh automation is in place before you can't trigger it for already existing pull request, or uh you know, whenever you're working on it and you, if you want to just release it to some people in your org, we can't do it without onput request. It's completely, futuristic and even rerunning on hotfixes.

T

You need to merge your automation code to main branch and uh kind of get feedback from there. But if you take workload dispatches since it's manual, you can just put your link to the already existing pars and you can run the automation uh for early adopters.

T

You can share the links before sharing it to everyone uh and kind of do it and more I'll tell more about the early adoption and the hotfixes part after the slide uh kind of and of course, the shadow pr, as I told will han, will have that delegation server, that that you can kind of use to get the web books and uh and like automatically uh shadow new pr's.

T

If you see this right, the hot fixes on the other part, the core. The core uh thing that I love about uh workflow dispatches is that it has this branch-based workflows.

T

You can trigger workflow dispatches from any branch you like, which means that if I have had that automation in place- and if I want to do a bug fix, I don't want to like wait for it to be merged. I can just quickly write a branch and have that bug fix in that and give it even for rolling out new automation. We can have this.

T

We can have the code in a separate branch and uh we can kind of give that to the early adopters and ask them to trigger from this branch so that you get this like they get to use it before anyone else.

T

This branch-based workflow had been like very uh helpful for us, especially uh you know to like test out and stuff. So in summary, I would say: workflow dispatch is awesome. Definitely give it a try, it's. It might be kind of seem like a manual thing, but I would say that the future of automation is the ability to trigger it manually. It gives a whole lot of flexibilities, so yep give it a try cool. uh Thank you.

T

I hope you had a good session.

M

Bye, I loved how he mentioned that future of automation is that will be triggered manually. What an incremental approach to automation, great talk and uh viewers uh you can head over to github discussions and catch him live for next 30 minutes get your queries answered. It's a very interesting uh talk out there, which just happened so just get him on discussions.

L

And have you checked the voices.

L

Have started holy early, this they're playing holy and having a blast there. So remember folks, you can go to the oss channel as well and check the content out there, as well as attendee survey.

L

We want to hear your feedback, so click on a link below and share all about your feedback and how you liking- github, india, satellite, okay, so coming to the next session- and this is a really interesting topic uh of a favorite of for many bringing machine learning and operations together, ml ops and how data scientists and engineers can start communicating, collaborating better how to create that workflow. So bringing in is the next session is all about mlaps using github, as ecosystem gets more complex, deploying mlaps gets harder.

L

Let's understand how can we do that whole mlops lifecycle in github sharing all about that is shruti chaturvedi, a backend ai developer at corexa, ai, a howard v code lead technology, fellow a certified microsoft trainer are developing ai solutions and leading and part of the, as well as a stanford ambassador. Really looking forward to this talk by shruti on ml of using github over to you should.

L

U

Hello and good morning, well we're clocking in at noon, but I hope you had a wonderful morning. Listening to the amazing sessions we had lined up for get up satellite india thanks for tuning in to this session. My name is shruti chaturvedi and I am a backend engineer: building scalable solutions at connexadi.

U

Imagine a world where teams would not have to spend months trying to recreate experiments. Their colleagues did and the good news is, in this session, we'll take a sneak peek right into that world, we'll be looking at github as an amazing tool to build test, deploy, monitor and version ml models and assets to allow faster and iterative development.

U

So it's pretty amazing and let's get started so emma lops is essentially an engineering approach to accelerate the adoption of machine learning into your business solutions and when you're, looking at continuous x's in mlaps, there are quite a few so continuously training your models to achieve better performance and higher accuracy over time. So, whether that's through retuning, your hyper parameters or changing the base, algorithms itself, all of that counts as continuous training continuously delivering your updated models to the end users, because that's a crucial step.

U

We then have a continuous integration, which is an important concept in devops and mlaps alike, which means you're continuously, integrating your local changes across to production code, which then gets continuously delivered to the end users and continuous monitoring to make sure that your data assets and pipelines work. The way you want so by using github at the center of ml ops enables building ml models, versioning data and assets, tracking and testing ml experiments and deploying machine learning solutions super easy.

U

So these are some of the tools and services we will be using to build a binary classification pipeline, so we are using something called a dvc or a data version control which is a open source tool available for free to use, and it sits right on top of git and we'll be using it to build and version our pipeline and we'll use github to orchestrate the pipeline and also to enable us to share those versioned assets between different team members.

U

All right. So, let's move into the coding. Now we will be building a binary classification pipeline, as I said earlier, and we are using a cancer survival. Data set all of the code is being hosted by github and dvc will very beautifully push all of the data pointers onto github on all of the tracked assets as well.

U

Also we'll be looking at some super cool automation, tool, called github workflows to do a lot of the heavy lifting for us and also solve the problem of comparing performance between different models with that, let's head over to github, and we will begin by cloning, this github repository onto our local system. So this is going to act as our source of truth and whatever is into the master branch.

U

Right now is what is onto production, so whatever is here is being supplied to my consumers and my job as a data scientist who is joining this team is going to be taking this accuracy and making it more accurate right and like it seems, pretty feasible. So, let's see if we are capable of doing that, uh let's switch over and see what do we have.

U

So here we have three different scripts and we'll begin by looking at getdata.pi.

U

So this is where we are getting our data from and storing it in our local system as haberman dot csv. So let's switch and see how our data looks like.

U

All right, so we have three different independent variables and one dependent variables, and this is what we are trying to predict: whether a person survived after cancer treatment or not. um So all of these are numerical. This is the age of a person the year in which the operation was conducted so 1964 here and the third column is an indication of how spread out the cancer was and again one means that a person survived and two means that it did not.

U

um Let's switch back over to we as code and look into our process data script. So this is where we are taking that dependent variable values of one and two and encoding that into a more standard format by using a label encoder and transforming into zero and one and storing that into our local system as haberman process.csv.

U

After that's done, we're gonna move into training script, um and this is where we are loading our haberman process.csv, transforming that into a data frame and then we're running a normalizer just to ensure that all of our the independent variables are normalized across their values.

U

So, what's into production right now and what my users are receiving is a simple logistic regression model right, like makes sense why the accuracy wasn't so great um and we are pulling in parameters from params.eml files. So that's a pretty interesting file. Make note of that we'll be looking at in a second.

U

We also have a bunch of metrics. We have accuracy, specificity and sensitivity, and what we are doing with those metrics is dumping them into a metrics.json file, and what this is going to do in the future is help the code reviewers and the mergers to look at metrics.json at a master branch and metrics.json at a different branch which we are working on and be able to tell if this is something that they want to merge, based on the accuracy and all the metrics a little bit of continuous delivery aspect.

U

Here we are using a github workflow. The super interesting tool that I talked about, we'll be looking at how we have configured that so as soon as a model is pushed onto master. So as soon as maybe we have a pull request into master, we are going to put the final model into uh the inference service. So what I am using is aws, but it can really be anything switch over to params.yaml, and this is where we have defined all of our hyperparameters. You know we just have two, but you can define.

U

However many you like- and this is where my training data script or this script right here is calling my parameters from so I'll only have to tune them and change them here, rather than you know, switching back and forth between train data and different scripts.

U

Now how I want to design my pipeline is, for all of these steps to run sequentially, getting the data processing, the data training, the data and also I want a way so that my as soon as my parameters change the training script should run again and produce a new model so how we have configured that is through dvc pipeline, and this is the pipeline config file, so dvc.yaml.

U

So we have three different stages and all of them are going to run. Sequentially and dbc is tracking each of these stages and it's going to track the outputs and the dependencies of each of these stages together.

U

So as soon as a person runs um their pipeline using dvc repro, what dbc is going to do is look into its local cache, folder, which we do not have right now, since we have not run the pipeline and it's going to see if a particular script has changed or if a particular data is missing and dvc is only going to run that stage so consider that you're gone on a vacation for two months and then you return back and params.yaml has changed like significantly.

U

So you pull all of your code from github and all you'll have to do is run dvc repro and it's going to see what has changed going to run the train data script going to produce the final model and put that in your local system to ensure that you have the exact same experiment and model that your colleagues do.

U

Let's just take a look at um what do we have? Okay, not the command. I wanted to run but never mind. We are going to try again, okay, let's clear this out, run dbc status, just to see where we are in this whole process.

U

Okay, so none of these files are in my cache, so when I run dvc, repro dvc is going to run all of these stages, pull in a new cache folder over here in the dbc folder and as it goes through the stages, it's going to put the outputs in the cache. So as it's going through a different stage, it's going to get the output and put that into cache.

U

Let's see if that happens, and if you're lucky enough to see that okay well, it seems that we are kind of lucky: okay, it seems to be working and as soon as this is completed, what happens is if I was to turn over to my git ignore so github is not tracking.

U

All of these um data files, instead of what we are pushing on to github, is actually pointers to where my data is is stored and that's honestly, such a smart way, because you have pointers to each of the versions of the assets that you're producing so now. What I I want to do is check out a new branch and, let's see, let's switch over to train data. Where is it okay and what we want to do is use a multinomial naive base model.

U

You know, since our job is to make this whole system more accurate.

U

Let's try out uh okay, perfect, let's try, multinomial naive bayes and here is our classifier, so I'm just going to change it to using multinomial and b and we're going to comment this line out so now what I want to do is run dbc repro again, so the training data is the only script that has changed. It's going to reproduce a new uh model and it will put that into my cache and I am going to track that particular new model with github.

U

So soon as that's done, let's take a screensh um start off where we at or you know just just to make sure that everything remains intact, doesn't break all right. So, let's push this on to origin and as we push it onto github, we'll see such a cool thing. That's gonna trigger there. So should we switch over to github and see what's going on all right head over to the actions tab.

U

uh Let's see awesome so something just got queued and it seems to be running a workflow right. It's it's like a whole bunch of steps which are running so where did we configure it? And how is this running all super good questions which we are going to answer?

U

So, let's look at the github, slash, workflows and workflow.yaml file? Okay, so what workflow is essentially defining on high level? What needs to be done and what I want is as soon as I push on to uh any branch onto github, I want to run this particular script. I want to run this job called run pipeline, which is going to run on an ubuntu latest machine and in ubuntu.

U

All of this machine is actually configured with this container, which has pip, install dvc, etc, and all of this is running on github provision machines called github runners and github runner is going to be configured with something called so we're running these as steps and when we have these steps you're using something called an action and that's like a a predefined sort of term or predefined uh tasks that we want to do so here. What we want to do is check out action, slash checkout.

U

We want to check out all of our code onto that github runner, so so that github runner will be able to look at my code and then do whatever it wants to do. So it has these secrets on github token, which is getting it access to read and writing from my repository, which we'll look at where the secrets are stored. So as soon as it has all the code onto the runner, we are going to install the requirements.

U

We are have we're going to reproduce the entire same pipeline that we did in our local system by using dvc repro we're going to fetch the like the how the tree is looking like on github and then there are two things that we're doing, which are super cool, is we're looking at the metrics.json and params.yaml at the master branch and then we're looking at uh metrics.js and params or yamaha different branch and comparing them and then we're sending that as a comment to that particular comment and what that's going to do is again help the code reviewer, see and understand.

U

If this is something they want to push into production, so this action still seems to be running. Meanwhile, let's switch back and check out. Oh wait. Did it okay? We will come back to that in a bit, let's also check out uh another branch and let's try lda here.

U

So what I'm going to do is from let's have a discriminate analysis right. I I I'm just going to try out all of my luck and see which model works best and take credit for all of that work.

U

So let's also comment this line out. So again we are going to do the exact same step. We are going to run dvc repro to ensure that our new model is on us and it's also with our teammates. You know, like team working with the team, is really important to ensure that everything runs smoothly.

U

Okay, so once that's done, let's add our files and make a git comment.

U

Let's say lda, not a super descriptive comment, but that works git push origin and let's do let's push it to lda. So from previous experience we know an action should have been triggered right.

U

Okay: well not yet it's so it just got pushed and it's running so let it run. We can look to our push that we made onto the multinomial knife base model and see what do we have on there. We know that we were looking for a comment right and let's see what this comment is going to look like okay, so it seems that we kind of did worse by using a multinomial live based model and right now, I'm a code reviewer right.

U

I don't want to push this into master because um the sensitivity, the accuracy really not what I like. So I'm just going to do nothing with that and I'm going to wait on lda, which my team member has pushed okay. So switching back, let's look over to the other um deployment file that we have, which is called deploy to aws, and this is another github workflow. What I want to do here is directly deploy to aws.

U

So as soon as I push on to master or the code reviewers push on to master, what this is going to do is take all of the code package into a zip folder and then put it directly into aws. I will not have to do anything will not have to switch over to aws. You know will not have to configure um steps over there. Very simple, very nice. uh So here are certain environment variables which we are using in this workflow.

U

So again, like last time, we were checking out this repository onto the github runner. Then we are creating a deployment package using um like a standard command line tool, um and then we are configuring aws using something called aws actions, and these are actions made available by aws themselves.

U

So these are open source you can check them out and what this particular action is going to do is check out, um or it's going to configure the aws cli onto the github runner with secrets and access key id just to ensure that that github runner has access to all of my resources. So it's able to pull or push or you know, read and write onto different resources here, it's just s3 and where these secrets are stored is back here onto settings switching over to secrets.

U

This is where we have secrets defined and you can create a new repository secret and that's how your workflow.yml file is using these secrets from so that is for the second script. Let's see what happened with our actions?

U

Okay, so that's good that it got completed and okay, so this was an lda branch. Let's see what's going on with this branch, we are hoping to see something better like something positive, at least so. Okay, that's a good sign. We have a quite significant change in accuracy and sensitivity.

U

um I don't really care about maybe specificity here. So what I'm going to do is I'm just going to create a pull request onto master, and we know that, as I compare my pull request with master, our second workflow yaml file, which was to deploy to aws, is going to get triggered so switching back here, let's refresh this okay, not yet.

U

Okay, let's see if we actually um merged changes. Okay still.

U

Okay, never mind we never merged it. Let's confirm the merge so now we should go back to actions and see something. So that's good. So we have two of our um workflows running. We have a deploy to aws workflow and we have a the one that we were using previously to compare metrics. So as soon as these get completed, we can switch over to aws and see what's going to happen right, um let's see so this is okay. This got completed so that's good.

U

Also, one thing that I wanted to show is what would happen if there was an edit in your script right. So if I made an intentional letter earlier just for demonstration- and I wanted to show- as there is an error, you can go and check into your laws. Where exactly is that error and you will be able to resolve it and run it again?

U

So I absolutely love that continuous testing and monitoring part just to make sure that everything works so switch over to um s3 and let's refresh this, and we should have a new, zip deployment file if everything if everything works, including my internet, okay, ooh, awesome, I'm so glad. So this is our zip deployment file and this contains all the updated code. So anyone I can you know if I have some different setup, which is using deployment from here. We can switch back and use that.

U

So that was for me using a very you know, a big way and a very sort of complete way of doing an end-to-end machine learning, uh pipeline you're versioning, the system, you're testing, it you're configuring, it you're, trying and making sure that different algorithms are getting configured and your branch managers are able to see and um sort of tell if this is something that they want to go into production, and I really do think that, with um with aws and with github actions and workflows, every development of any sort is really sky. Is the limit.

U

Essentially, so that's a wrap for this session and it's time for q a you can also head over to github satellite.com discussions. We're going to be there and answering all your questions about um getting workflows and actions and how we're using it and how you can use it all right. So, thank you so much.

U

L

Back, thank you, shruti. That was an awesome session and love the demo gods as well as internet gods, like everybody played along the first question that I have for you is: how has data version controls, like dvc's functionality, been amplified by.

U

Github yeah, so I I really like this question because imagine if you have like a team of people, you're working with right and like you, have data coming in every day and like you're, trying to build you're trying to version and it gets so complicated that there's no way for you. You're versioning it. But there's no way for you to share those versions between your team. So maybe you're emailing one person, maybe you're whatsapping your code and your data to the other person and like just imagine that one person gets one data.

U

The other person gets the other data which is not going to work for your team. So get up a very central role that it's playing is making sure that every team member can switch between um different versions and it acts as a source of truth and everything works for the entire team. Rather than just the person who um is working on that particular assets and those questions.

M

Great great thanks for that insight. We also learned in your introduction about you being at connexa.ai, which is all women-led startup would love to hear more about. How is it working like at that place and what is the story behind it? So tell us more about connexar rdi.

U

Yeah, absolutely I absolutely love working there. It's all women, it's all women of color who are running the team, all super powerful and how it was. This actually started during these code months and during the lockdown, where the ceo thought that there's not a lot of women who are leading in the space of a.I and leading in the space of startups, so maria that's the person who's, leading it sheath high.

U

You know, let's, let's get together a team of amazing women developers who are really excited about getting into this space and like building software um and within like two months. There was so much positive comments and like positive feedback, we were receiving from people and we are just going to launch our new mvp product and it's so so fun working with dynamic women from you know even different countries. We have people from portugal, from brazil, from canada, all amazing and um and over there as well.

U

You know I'm always sort of uh pushing for using github actions. I'm like let's go. This is such a great. This is such a great system because I am managing like several less infrastructure deployments there and I'm always like you know what let's use github action. That's so cool, absolutely love. It.

L

Thank you and a big shout out to all your leaders there for thinking about it and starting something which, like all of you, feel so passionate about. Yes, big, shout out to all of them there. The question I have for you is in the roadmap for actions like what are the: what are the big pieces? You're looking forward to.

U

Yeah um well whatever I'm thinking of actions or like github actions, uh there's such a great space and like first of all, an amazing developer community who is like building the actions to make sure that all of the platforms and the systems you know you have aws. You have azure in gcp, which was the the session we had earlier, which was using actions and gcp.

U

uh It's such a wholesome environment of building and making sure that uh all of the developer community, whether they're building with python or whether they're building with java javascript they're, all um they'll, feel comfortable and that's such a big thing um and especially how it gets configured and built together with the github workflow. I also the openness that you have. You can like choose between different options and just the dogs themselves, they're, so cool. uh Absolutely.

U

You know, I'm always looking at the marketplace and always gets so excited. I'm like whoa, a new action in the marketplace. That's so cool uh so yeah that I would always um support and push for github actions as much as I can.

M

Wow that was music to my ears, and you talk about a lot of github features and capabilities. Here we got this interesting question to almost every speaker, and here it is for you as well. If you go to pick one feature, this is which is your favorite in github? What would that be?.

L

And now this time you can see, I would say it's probably.

U

It's probably discussions, that's what I'm gonna say because it's bringing that developer community together, you know like we are able to talk. We are able to discuss things and it's such always. Such a good thing to receive feedback right and just like talk with the community and github has such an amazing developer community, and it's always growing yes, and that's so cool that this feature was implemented.

U

You can go in, you can talk. You can talk about what you want to build. You can talk about what the problems are. I I, I would always say, get up discussions. All the actions will come. Second.

L

Wow, I love talking to you you're, like a super fan of github, so I thank you so much there.

L

Another one that I see coming up is how, especially again coming to ml ops, bringing data scientists and engineers together like what's one one thing you would want to share of: how can they start collaborating much more better.

U

Yeah, no, that's a really wonderful question um and I would say that you know it's such a broad infrastructure like ml and ml ops is such a broad infrastructure um and with their different aspects. You know you have the etl with data. Then you have like the training of it. Then you have the deploying of it, um but it's such a central system, especially with like tools and services like github.

U

It's like such a central system that, whatever everyone is working on, they can collaborate very easily and also they can um sort of see the team movement as they move through different processes, so the the whole ecosystem of developer engineers and ml ops.

U

I think it's uh it's a great infrastructure for especially for me, I love being in this infrastructure and I love being able to contribute to not just the ml part of deploying things or not just the ml part of training, the data itself, but also the whole infrastructure of how or where the data is coming in from and how we can design it and how we can make it more suitable to the users that continuous feedback from the team and users super essential.

M

Wow you packed a lot of goodness in this talk and we're gonna tell our viewers that you are not leaving them for another 30 minutes and you will be with them at github discussions. So we'll take your leave and uh bye, bye, shruti, but viewers do throw all your queries. All your questions to her at github.

L

L

ah So we're getting close to the end, I don't know I have a happy and sad feeling it's been so good, uh but yes, uh we also coming to an end to it.

L

Folks, remember to share with us on twitter, github, satellite and share with us your favorite moments, your favorite session, as well as what you've been loving. Where are you watching us from so share everything at github, satellite.

M

Okay, divya, I got one last chance for you to win this awesome. Octopus. Are you up for it? All you got to do is tell me: why should a student subscribe for the developer pack.

L

Okay, students: do you have money? Can you buy all the tools that mohit and I or everybody in github can no right. The best way to get access to top industry tools is by applying to the student developer pack, you get tools hundred plus tools worth of 200 000 dollars, and you get all of that free. If you apply for the developer pack and you can then develop code, it just democratizes software development. Isn't it you'll have the same tools that everybody in github in the company in the industry has so go, go to education.github.com.

M

You said it the best way possible, and there you go emptying my shells, one more.

L

Time, oh, I love this. Thank you. So much and folks go get the student developer back.

M

Absolutely and with that, let's jump into our next session, which is about how github uses github for collaboration right, we just don't make the tools, but we use them ourselves as well. We're gonna hear a lot about how internally kedup is getting used across the organization we developers beat the sales teams meet the marketing teams.

L

I want to steal this opportunity from you mohith to introduce satish. He is one of my favorite engineering counterparts, someone who I go to for advice as well as our in-house rapper, so sharing all about github for collaboration is shatish kumar he's a staff. Engineering manager at github leads the chat, ops area and is passionate about leveraging technology to create an impact at scale for developer products and platforms. To achieve more so over to you, satish and.

R

R

Namaste: everyone welcome to github satellite.

R

Github is a place where github is a place where people build software every day. Millions of developers use github to create things that touch billions of people to be at github is an awesome responsibility and I lead the chat, ops, integration team in github building products like slack and teams github with a goal to amplify engineering, productivity.

R

Let's start with numbers chat, ops is a team of nine engineers and the team has done 80 deployments in last four weeks, making the average to four deployments per working day. It is truly awesome from a devops metrics point of view.

R

I let you all guess what the other phone numbers are and post your answers in the github discussion.

R

Or twitter and use hashtag, github satellite.

R

There is a story behind these numbers that makes them magical and absolutely motivating for my team. The story of emotions started last year in march 2020 and I still remember, sitting in a conference room with leadership team, pitching the idea to build the vision of building a unified chat, ops platform.

R

My team was super excited to build this vision. Project execution barely started.

R

And everyone in the company was asked to work from home due to pandemic major blow to team collaboration, resulted in reduced team velocity.

R

Keeping the team velocity aside every engineer was going through varying emotions, making them feel anxious, isolated, annoyed, worried and sometimes frozen and stuck, but this is fast coming back to present, we are just a couple of weeks away from shipping, our fourth deliverable and making the vision a reality in april.

R

In next couple of weeks, I'm super excited to share the journey of unified chat. I'm super excited my team's journey and give a walkthrough of how github uses github for collaboration.

R

I mean inclusive collaboration and, along the way, share practices and processes in engineering that help us shift faster, whether we work from home or office in simple words, collaboration means working with others. By end of this session, we will learn about eight key elements to enable an effective and inclusive collaboration, collaborative culture in organization using github.

R

The code that powers github.com mostly resides in a monorepo- it's surprising, but that's the facts, and to enable team academy, every team own repository setup, specifically for the issues and team documentation allowing them to tune their day-to-day process. There are also teams which maintain their code in dedicated repo.

R

I have got a picture of slack repository, which stores the chat, ops code and the team has enabled almost all the features of github. Structurally. Everything in github organization starts. With the repository there is a repository for everything literally for onboarding job posting benefits, values even entitlements are maintained in a repository.

R

The best part of the setup is it helps everyone in the company to find the information quickly and easily planning and tracking is standardized across teams in github.

R

The goals translate into small consistently sized issues classified as initiatives, epics features and tasks to track work across all of engineering, and this facilitates us to speak a consistent language in terms of scoping and prioritization, while collaborating with teams.

R

Project boards are used across teams to organize and prioritize the work project. Boards are made up of issues, pull requests and notes that are categorized as cards in different columns.

R

Team project boards are accessible across organization and anybody in github can discover this project board easily just by searching and they can start collaborating either by subscribing to an issue of their interest or maybe share their thoughts or comments to enhance the future.

R

If something is a priority. Yes, it's a priority for the whole team and we limit the number of in-progress epics to two for a team of size. Eight, and this has been a great way to have a focused execution on delivering high customer value faster.

R

We have seen increased team bonding as there is developer engagement among people as there is deeper engagement among team members to work together closely towards a common and shared goal.

R

Issue templates issue templates play a wider thinking, consistency by customization and standardization of the information captured and github issues when they are created.

R

Let's take an example of a github issue which is created using epics, which is created using templates. The labels are automatically included when an issue is opened using templates.

R

Labels are used to classify different types of issues, epic feature and others, and issues can have as many labels as we want.

R

Epic is focused on a specific problem statement and epic acts as a shared goal for a group of engineers working on it.

R

An epic is a group of features captured as task list inscription and task list has clickable check boxes. We just go and select and unselect this check boxes to track the progress of if a feature is complete or not, and these help us to track the progress easily with minimal follow-ups, and if you see there are milestones set, which sets the right expectation with the state stakeholders on when this epic is going to be shipped and the same. Epic shows up in multiple project dashboards for tracking at different management levels.

R

If you see the c chat of project dashboard is the dashboard owned by my team. Apart from that, this epic is also getting tracked into two other dashboards moving to feature a feature, acts as a unit of work for around two weeks and anything greater than two weeks is moved to different feature, thereby enabling us to do work parallelization.

R

So in this way we can have multiple engineers working on the same epic and different features.

R

A feature is usually assigned to one engineer: giving them an opportunity to take complete accountability to drive the feature to completion and feature is a group of tasks and similarly, the way we use in epics. We make use of task list to track the progress of all these tasks.

R

The best part of using task this year is we have something called definition of done and capturing a list of tasks to be completed as task list sets. The right expectation with the engineer working on the feature in terms of what needs to be completed and features only show up in the team dashboard.

R

But no one is stopping them to monitor in our dashboards too and features are shipped as and when they are complete. Empowering the team to ship, often iterate fast and encourage calculated risk taking, unlike epic feature, is only tracked in team level. Project dashboard, as I mentioned.

R

Is the discussion happening all the action happens in issues github issues? Let's take an example of a issue created in a bug bash done by the team.

R

The issue was created by a product manager, putting a title in the details in the description and then a senior engineer, trages the issue and the engineer questionnaire, makes use of the ad mention functionality to start the collaboration from github issues.

R

Ed mentioned results in sending a notification to the engineer who was mentioned, and it mentions provide a flexible way to get the right people involved in the issue. Once the engineer gets the notification, they click.

R

The link available in the notification come back to get a issue and the collaboration starts from there and if you can see, the bug was closed in the same day and primarily because the entire collaboration started from github issue and, apart from getting things done, capturing everything at one place in github issues helps us to maintain the context and the history of actions which enables information, discovery and the decisions taken can be referred in future by the way. If you are wondering what all goes into the issue, then literally everything is captured.

R

Every detail from spec clarification, design, proposals, design, chatter, meeting notes, decision taken literally everything. You also use discussions. Discussions are used for brainstorming when there are new initiatives and when we know it's going to take longer, they're also used for team building like how was the weekend. What were you doing? What's your plan for the coming weekend and all these gets into discussion discussions? Are we use not only for technical but also for fun elements? I'm increasing the collaboration in the team.

R

How do we keep everybody in the entire organization to be aligned on the status? Then we have a concept of status, update comment. It is just like a comment on the epic which covers the status, whether it is red, green or yellow towards the projected ship date.

R

Apart from that, we encourage having video demo video links and these demo, video links can be from anywhere. It could be a small working prototype coming from a developer environment or production or staging they don't need to be only collected or recorded in production, and we believe in demos and not memos, so the other part of standard status update is risks and mitigations.

R

This helps us to set right expectation with our stakeholders and we call out dependencies dependencies and risks that we have, which can delay our shift date, and we also share the mitigation plan.

R

Moving to next once the design is closed, the core part of the engineer who wants to go and develop and ship the product as soon as possible in the whole exercise of shipping. This after coding, I want to talk about the best chat, ops, chat, ops is about conversation, driven development by bringing the tools into conversation.

R

Every organization or most of the organization makes use of slack and teams for day-to-day interactions in chat and chat. Ops is one which brings all the tools and processes into the conversation where the engineers are.

R

This is an example of a pull request created by an engineer, and in this we encourage engineers to capture design changes and also user experience, and if the change results in a user experience, change then putting before and after experience helps the reviewers to get the context and review faster.

R

This pr is also linked to an issue and as soon as this pr is merged, that issue will get automatically closed.

R

The pr is assigned to reviewers and to make sure none of the prs are getting into production without review. We have branch protection rules set on our repository, which helps us to ensure that every line of code getting into production is reviewed by the engineer.

R

Apart from this, the quality is also baked into the whole pr workflow process. We leverage, actions and actions, help us to automate customize and execute our software development workflows.

R

These are the list of actions available in my team's repository and to give an example, we use code coverage as part of our pr workflow and as soon as the pr is raised, the code coverage tool runs and post the outcome of that in a tabular format, and if there is a dip of even 0.5 in unit test coverage, then the pr, even if it's approved it, can't be merged.

R

Okay, now, let's go to the pr workflow, the author creates the pr and ads reviewers and then the reviewer receives the request notification on slack github, app slack github app is the chat ops offering from github. Apart from slack, we also have teams github app available in the teams marketplace.

R

So once the notification is received by the reviewer, they navigate to the pr from the link in the notification reviews the pr and leaves the comment for changes, and once the comments are left in this case, the reviewer is asking for a couple of changes from the author and then the author receives the notification and modifies comes addresses the pr command. Then ed mentions the reviewer. Since the comments are addressed, the reviewer gets a notification goes to the pr reviews, the changes and approves the pr.

R

That's it, and once the pr is approved, the author of the pr gets a notification on the github app now. What next? The pr is approved code is ready to get birds deployed in production and deployment in one way is a complex process. If even there is a new engineer in the team right, they need to understand a lot, but in github deployments are simplified right now I am going to show you the real power of chat, ops.

R

The engineer in this case heads to chat, ops of slack channel and delegates the work of deployment by running a simple dot deploy command. The work is delegated to none other than hubot. Huboid is github's.

R

Huboid is github's most friendly, efficient and collaborative employee. Almost every github developer interacts with youbot literally every day multiple times and hubot is a robot. We just ask the hue bot to do the deployment with a single command and the job is taken care. The deployment is done. What next time for a party.

R

The engineer is super happy and the engineer goes to the channel and shares the completion of the epic. The team, the team starts to sparkle. The engineers who contributed towards team goal by shipping, the epic, the dot dot sparkle command that you are seeing on the screen is a hubot command used to recognize team members effort cue. What not only does complex things. It plays an instrumental role in bringing fun and spread joy within the organization.

R

This brings me back to how chat ops team used github for collaboration.

R

As I said earlier was an emotional journey when I think about the last year when we started in march, and the team used github for collaboration and the product helped us to navigate a transition from our emotions of isolation, nervousness helplessness out of control to be creative, confident, inquisitive, amazed and joyful.

R

Now, let's do a quick recap here is the glimpse of github features showcased in this session, embracing the right practices and processes along with these github features, act as a catalyst in boosting the key elements of collaboration and creates a culture of effective and inclusive collaboration in an organization.

R

Moving to the puzzle, I am hoping everyone was able to crack this and post the answers, either in github discussion or on twitter.

R

Let me decode this for you. As I said, 80 deployments by team in last four weeks, making an average of four deployments per working day. The rest of the numbers are and you're able to achieve this 80 deployments by working on 66 issues which were closed in last four weeks. 351 issue comments, which shows the amount of collaboration we do on these issues.

R

72 pr's were.

A

Raised and almost.

R

Every pr lands into production 253 pr comments to ensure the code is sent with high quality.

R

That's it from my side. Thank you.

M

Oh, my god, satish what an amazing rendition of that one year journey hats off to you for that, and what those numbers really blew my mind up right before we jump into q a we heard that you are in-house rapper at github. So will you just do a small bit for us and all our viewers right away.

R

Are you sure you want me to do that.

L

100 waiting for it, okay,.

R

I have to get in the mood first, then.

R

Code issues, pr discussions are action, code issues. Pr discussions are actions. Every feature is built with passion it mentions and notifications enables inclusive collaboration. Oh my future generation do check out github education code. Spaces is the new workstation.

R

Github is developers ultimate destination ultimate destination.

M

R

L

Yes, like once more once more that was awesome satish, I love it.

L

R

Really yeah, by the way I don't wrap. I just try to rhyme lines.

M

Come on it was amazing. It was amazing, let's submit to q a now. We got a bunch for you.

L

Yes, I have to now get back in the zone to ask you questions.

L

B

L

About uh the emotional journey, what was the biggest turning point there for you, as well as your team.

R

Yeah, so if I recall it correctly to be very honest personally, I was also struggling to through that phase right.

R

This was the second time in my entire professional career, where I was feeling helpless and out of control to help my team.

R

And you have to think what help? Really it's not a big thing, but I think the weekly check-ins that we started after our scrum every monday.

R

We have scrum every day right and then we started this way of interacting with folks doing a weekly check-in on mondays, where we used to share with each other how everyone is go, what everyone is going through, just speak out and what is working for them? That's it right, and this whole concept made everyone realize everyone is going through the same phase.

R

We are not alone, the whole world is going through this and we need to work together as a team to find out how to come out of or how to handle this unprecedented, complicated and hard situation that the whole world started seeing last year.

R

So if I have to pick one thing, then it was that weekly check-ins that we started doing with the team.

M

Very insightful thanks for being so honest and transparently kind of bringing these and talking about these things. We got one more for you. What was the most challenging key element in leveling up on collaboration across the teams.

R

Interesting, so what I showed was eight key elements of collaboration, see in simplicity, collaboration means working with others, but that is not that easy. When we work with like 2000 people in the organization or even when we work with 10 people or when we have five people at phone rights, the most challenging one is always bringing in that inclusion. Part in the whole collaboration story.

R

That is the most difficult one. We talk about diversity, a lot right, but if there is no inclusion, it is very difficult to make the diversity successful in an environment, and this is something that we achieved by writing a lot in this async mode and we started sharing and putting those in discussions issues everywhere, where the team was able to see who is doing what and they had an opportunity to come back.

R

Read about the proposals, share their thoughts and contribute and share the diverse opinions in this way, as a team able to come to a inclusive and conclusive conclusion.

R

But it's a journey, we are fully not there, yet we are still learning it's not that we are perfect, but I'm extremely happy to see the graph the way it went up and you're still learning and there's a lot to learn and github is the way to make use of all these in the right way to make everything successful. I'm joyful too.

L

Thank you, sateesh. Thank you again for your vulnerability and sharing from your heart. uh This was really insightful and again things that everybody around us can learn as well. Again, thank you for a great session and remember folks. Satish will be with us in discussions, so you can have your heart to heart, chats with him there as well and ask him all the questions folks discuss with satish on discussions.

L

Thank you. Thank you again, satish bye-bye. Thank you.

R

R

M

Okay, here we are the final session of github satellite india 2021 hope you enjoyed it as much as we enjoyed hosting you all thanks a lot.

M

Thank you very, very much for sharing your comments talking to us on all the platforms and hanging out with us for these two days and a huge shout out to current and anisha on the oss channel, who equally engaged you and brought so much insightful conversations to you all and an extra special shout out for all the guests and all the speakers who came across engaged with you all and brought such knowledge of nuggets for all of us.

L

Last two days has been so much fun like so many months and weeks of work in in such awesome. Joy experience like mohit, said, hope you folks enjoyed as much as we did and to adding to the thank you notes, mohit, I want to say a big thank you to you for being the best co-host everybody backstage. Thank you so much lauren brian hayley tammy. Thank you. This was not been fun without you folks. So thank you so much for that.

M

L

M

About all the content.

L

And thank you for my awesome goodies as well.

M

Absolutely absolutely and don't worry all the content that you watched here. If you want to relive any of the movement, you can go to kitab satellite.com and all these beautiful movements are right here for you any whenever you want, you can consume them and remember we really want to hear from you and want to improve on areas do better than what we did this time so go fill out. The attendee survey- it's right here for you to share your thoughts with us.

L

And share with us on twitter, we are continuing.

L

And we'll come back to you now, let's go to the last session of the day and it's time for the final session by manish, sharma, general manager of github india, avid reader in 2005. He actually started the first online book library called bookmebook.com and to continue with the passion of reading books and sharing it with everybody in every walk of life. Manisha is based in delhi and enjoys learning guitar, as well as video games so manish over to you one second, mohit manish is your boss. Do you want to share a snippet about him.

M

Oh absolutely I'll share a small secret manish is a coffee connoisseur from what means to use how to crush them, how to brew the best coffee? You can ask him about anything under the sky, about coffees, so manish, looking forward to hearing from you take it away manish over to.

K

K

You've been excellent hosts, um I'm sure everybody enjoyed the last two days and all the fancy stuff that you've been doing. I've discovered two secrets: one mohit is a great bhangra dancer, so you know stay tuned for more meetups, where you will see him in action, and I also found out that karan loves to play holi.

K

So I'm sure it was fun for all of you, but most of all thank you to each and every one of you for being with us over the last two days as we celebrated the india developer community, it has been a pleasure uh serving all of you over the last one year and we are overwhelmed with all the love. The engagement that you've given the session started yesterday with erica sharing with all of us how we've grown to more than 5.8 million developers from india who call github their home.

K

It is super super exciting to see that we added 1.8 million over the last one year, but what's more exciting and current talked about how I always love anything open source has been seeing that indian developers have been driving the most engagement on open source over the last one year on github. This is just fascinating.

K

You know I come from a very uh different era. I I started coding in the last century. I always like to say this because I keep people guessing. You know what my real age is, um but it's so great to see the changes that have been happening in the ecosystem, with young developers like yourselves innovating and contributing back to the open source community. So we thank you for all the support.

K

The last two days have been great. uh I know it has been a power pack session with so many content coming your way, but, as karen mentioned, all of these 22 sessions will be available for you on demand for you to go and check whenever you feel comfortable or have time to do so once again, thank you so much.

K

I also want to talk about what shanku and the engineering team took you through in the morning. There is a lot of convergence that is happening across security across development collaboration, especially in today's world, where everybody is distributed.

K

We have so many customers from india who are really driving innovation and leveraging github for getting their teams to collaborate and build software, and these customers are all of all sizes, whether large or small, from any industry, whether they are companies that are transforming into software companies or bond digital companies and one such large organization is cognition, needs no introduction. They are one of the largest developer companies in the world, with teams and customers across the globe uh that they service.

K

Let's hear from hari, who is the avp enterprise architect at coppesant on how they are driving transformation and developer productivity using github.

V

Hello: everyone cognizant's internal I.t team serves one of the largest enterprises in the world with 300 000 employees worldwide.

V

It's a fortune, 192 company, 16.7 billion dollars in revenue companies, critical business operations, runs on internal systems and processes as part of the cognizant's digital insight strategy, which is about transforming the internal processes and internal systems. It is based on three pronged approach. First, is modernizing employee experience. Second, is building a strong data and integration foundation.

V

Third, is modernizing our application ecosystem on enterprises cloud first mobile first strategy: we have to be kind of very nimble and agile in our processes, and we also need a cultural shift in terms of how we deliver our products and services.

V

Our entire engineering processes and tools have undergone transformation using github and devops hundreds of software professionals now on github. The conversation is shifting towards. How do I engineer the code? How do I internalize the best practices available in the industry.

K

Thank you, hari. Let me now talk about some of the programs that erika mentioned and I'm sure all of you are as excited as I am when we announced it yesterday and these programs are really going to drive the engagement for developer communities across students startups and the open source community.

K

Many of you have already started applying for these programs. I am so overwhelmed by the mails that we've received from each and every one of you over the last 24 hours. Since we announced this, I do request you to leverage these programs and give feedback a lot of these programs. We have created over the last one year based on the feedback that each and every one of you gave on what you expect from github.

K

We are really looking forward to feedback coming our way for creating more such programs to drive the engagement with developers like yourself and the whole purpose of this. Is you know to get more and more developers on the platform I like it?

K

There is a huge generational shift, that's happening in the industry today, with more than 500 million applications that are slated to be written by 2023, which is more than the last 40 years combined.

K

I think it's a great opportunity for open source for students who are stepping into the software development world and also for enterprises to leverage github to really build software, and that's been our purpose ever since we started our operations last year to serve you better to be closer to what you would require from a platform like github as we push the boundaries of developing applications to move human progress, so really excited to be here and in a way it's been great um doing this virtual session with you. I can share that.

K

You know last year, when we were kicking off um our operations, we did our kickoff on valentine's day, 14 february 2020, it's so vivid in my mind. uh It feels like 10 years, but it's only been one year.

K

We were, you know, expecting 250 300 people to show up, and we had 2 2000 people show up at the event and it was a physical event and we were overwhelmed by the love that we got from developers and I'm sure by now doing this virtual, we were able to reach out to more and more developers across the length and breadth of india so really excited, and I'm thankful that you were able to spend that time with us as a next step. I would request that each of you follow our repository on github.

K

This is a public repository that we've created to share more details about three programs that we've announced and also on an ongoing basis. We will keep sharing meetups hackathons, more programs, and you can use the discussion boards to interact with the india team on how we can engage with you a lot better.

K

I think closing satellite would be incomplete without me talking about my favorite moment, which is you know, yoga cat which we've unveiled for india. I I just I just love this mona of car and I'm sure you can see the details in the lotus petals that show the git operations um for a community connected by code like india. This is, I think, a wonderful representation of you know what github is and what it means for india.

K

Thank you very much everybody. I wish you a very happy holi um for those who are going to be attending the workshop all the best. I hope you have a great time learning from our engineering team on how to use github in a more powerful thing. Thank you and have a good weekend.

K

K

K

K