Continuous Delivery Foundation cdCon 2020, 20 Oct 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Lightning Talk - CD, Continuous Doughnuts - What Can Creating Doughnuts teach us a... Ravi Lachhman

Description

Lightning Talk - CD, Continuous Doughnuts - What Can Creating Doughnuts teach us about Continuous Delivery - Ravi Lachhman, Harness

Doughnuts are delicious and the perfect container of happiness. Going through the process of developing a new doughnut flavor can teach us a lot about continuous delivery. There are no Yelp Reviews for software applications [there should be]. What are the typical steps in creating a doughnut flavor vs a Continuous Delivery pipeline? Feedback is important in both doughnuts and software platforms. If we get this wrong, we might have a lot of stale doughnuts and tech debt to pay off. Let’s call this Continuous Doughnuts and find out!

A

Hey rob your luck with harness and today we're going to be talking about something very delicious continuously nuts. You thought that cd stood for continuous delivery, well, actually you're correct, but let's take a look at how food influences technology a little bit about me, robbie, evangelist again at harness. I worked at a few firms as a distributed assistant, engineer that worked in app dynamics, mesosphere, red hat and ibm and throughout all these firms. My love of donuts has stayed true and also I had a lot of time.

A

Engineering distributed systems, but let's take a look at something that might be a little bit closer to my area, expertise, technology equals yum well, ironically, a lot of times when we're talking about technology, we're talking about feeds and speeds, but no one really uses taste. So today, let's talk about taste, and so what we have to do is supporting taste. Innovation uh here are some pictures of one of my favorite donut shops in the world.

A

It's actually in denver, colorado, called habitat donuts and if you're unfamiliar with habitat, donuts habitat donuts claim to fame, is that they actually infuse their donuts with alcohol. So on the left is a rose. Infused strawberry doughnut on the right is a whiskey, infused, bacon, glazed doughnut. Now I'm not sure who came up with this idea, but this type of innovation is crucial and now how would we go about? Let's say that it was our first time making these donuts.

A

What do we actually have to do to go through the innovation curve so like something I would like to call idea to production in the donut world. So let's take a look at a donut deployment pipeline. So when building a donut or making a donut, let's say it was her first time actually deciding you know what let's go ahead and make that rose, flavored, donut or that whiskey, flavored or infused doughnut. Now the first thing is: we have an idea right and the second thing is: we have to go, build the donuts.

A

We have to go gather the raw materials, you know, get the flour get get the sugar get, the eggs get any sort of extracts and we really have to gather our raw materials which is pretty manual and then after we make a few things and we actually build it. uh It's time to ask our co-workers if they taste good and unfortunately, taste is pretty subjective right. Ironically, sulfur could be subjective too. I ended going up and building hype, so hype is always important.

A

Like hey you know, what is our donut tasting right then, most likely your co-workers, who, like you, will say sure uh nah, I'm gonna be like you yuck, because they want to support you right and then we're gonna head and go ahead and deploy that donut we're going to go, deploy that delicious donut onto the countertop off, let's say of our store.

A

Let's say I was working at habitat or we were working at habitat, it's going to be out of the wild, and this is where it now changes from our control to the control of our customers. uh For example, uh how do we go about verifying if someone likes a donut? Well, don't actually made to be eaten right, so we need to make sure we have verification.

A

Are people putting it on instagram or people eating it or people generally happy uh and then, but we don't want to happen, is seeing that people turning around and just tossing in the trash and using your phone to say you know where is the closest krispy kreme, uh looking at yelp or google maps for that?

A

But let me take a look at the amount of time this actually takes usually building something could take days right like hey, like you know what, if it's just getting something right can take days, baking things trying different variations and then, as you get to going look towards the customer, uh the amount of time starts to decrease. For example, testing it be couple hours hey, you know, amanda or bob have you eaten my doughnut and then deploying it just taking it from the oven to the counter uh takes minutes and then verification.

A

If people are liking it I could take minutes or if it's rolling back, uh you know it could be, could be sad right, but again the new treat pipeline uh from the inception uh towards the end. uh You know the time decreases as we move along the steps, but we have one problem here: uh I'm assuming for most attendees at cdcon.

A

Our donuts actually look like this uh they're, actually we're not in the physical donut business. Now I wish we were, if you, if you actually aren't attendee and you're in the donut business, do hit me up. I would love to try one of your donuts, but our donuts typically look like this, like a pie chart right, and so let's take a look at what the software development pipeline is.

A

So we can actually break that out in the same number of steps, for example, building it right so building it we're gathering raw materials, it's going to be from third-party open source to purpose-built or artifacts that we have uh testing it sure it's you know, kind of automated right, so this is going to be a large push for today's delivery uh that we have some sort of test coverage uh and then deploying it um hey.

A

You know what without continuously delivering your first deployment, uh you're deploying with custom scripts and it's something that we're here to solve at cdcon and then verification right, so we're trying to tie all this together. Hey did someone like it. Unfortunately, there's no yelp for software, you know we don't know or we're not.

A

People are instagramming our uh our endpoints or apis right, unfortunately, and then roll back, you know which could be totally totally subjective, totally manual and when building out a continuous delivery pipeline, you know this is exactly what we're trying to solve for uh and it's actually, ironically, the inverse of building a donut. You know the build. You know from the ci perspective we have down, but, as we start getting into the testing deployment, especially with the more I'd say when we're in production, how do we validate something if something went wrong?

A

How do we roll back and roll forward? Those things can take a lot of toil, and this is where we have to build or start building our pipelines right. So what is a continuous delivery pipeline? Well, basically, we want to automate all of that. So, for example, if one of the hard things to get right with this cd pipeline, I'm sure some of the other speakers in different talks are touching on. This is that you always want to have two radio dials.

A

You want to make sure that you're maintaining innovation on one radio dial, but also you want to make sure that you're maintaining business controls and so, as you progress stage stage, one of the things to make sure is having a particular pipeline that is robust enough to have different sets of rule sets right so in the first stage kind of leaving our dev machines.

A

uh Really it shouldn't be any sort of surprise that the rule sets for under, for example, third-party open source analysis unit tests, some sort of integration test that we're having even our code coverage with the static code analysis kind of going on, I like to call it tightening the screw or tightening the rope type of model. Is that now we're taking a look at the environment as a whole?

A

We're just start doing smoke tests, we start doing perf test uh if there's any sort of, let's say an environmental security test that we need to see for compensating controls, um you're kind of running that in a different stage of our pipeline and then finally, we get to production.

A

Just you know, instead of taking minutes like or for example, we want to take minutes like the donuts, not hours like it usually takes that. Are we doing some sort of canary? If it's a you know currently in place deployment? Are we continuously running verification? Are we consistently running, say perf test uh with actual production uh features and production horsepower? You know soap, tests and whatnot.

A

This is where we get to production all this to make sure our customers are happy and the rationale behind all of this is continuous delivery right and the rationale behind this again is making sure that you're able to deploy as much as you wanted to play or make as many donuts as you want to make donuts right so funny, here's a presentation in a presentation I actually gave a webinar, uh maybe a few months back and uh with the actually with the cd foundation here, and one of the things I used to talk about is a fear of deployment.

A

You know, like fear of missing out, if you, google, you know how many times a day should you deploy or amazon deploy like oh, they deploy less than every 12 seconds. If your organization is not there or your donut shop or software organization's, not there, you might feel left out. Don't worry you know, don't let the lean enterprise book make you feel bad. Everybody's organization is different. Everybody has a different style. Everybody has different tastes every you know your organization the best, uh and with that you know, I will leave you with hey.

A

Don't be afraid to innovate, don't be afraid to make that rose or whiskey, flavored doughnut, making a treat is similar to making software, and I'm sure your software will delight your your organization, your clients, internal externally. Thank you. Robbie.