Continuous Delivery Foundation cdCon 2020, 20 Oct 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Managed Releases with Spinnaker - Nirmalya Sen , OpsMx

Description

Managed Releases with Spinnaker - Nirmalya Sen , OpsMx

Spinnaker is a continuous delivery platform for releasing software changes with high velocity and confidence. But what if you are not doing continuous delivery? What if you are doing managed releases where the release manager based on lot of organizational and business criteria decides when to release, what to release. Can they use Spinnaker? In this talk, Nirmalya will show case how you can adopt Spinnaker's continuous delivery features to be able to make managed releases.

A

Hello and welcome to this session on managed releases with spinnaker.

A

Before I get started. Let me tell you about the company that I represent. We are op semex, a leading provider of continuous delivery solutions.

A

We have. We are an experienced team with people who have worked previously in companies like netflix, cisco, oracle aws. We have offices all over the u.s and india. We have 80 plus employees, and we are very active in the open source community working with spinnaker kubernetes jenkins, and we have you know significant contributions that we have made to each of these projects as service, offering we provide 24x7 support, manage spinnaker and as well as professional services, in helping you onboard and automate your cd pipelines around spinnaker.

A

With that, let's get into and look at what is uh spinnaker. Spinnaker is a continuous delivery platform. It originated out of netflix uh where they have been doing continuous delivery at the rate of almost 8 000 a day, and this number is probably from a year ago, so that number may have even gone up so where, where does a typical enterprise set? Is they're typically able to make releases? You know earlier was seven eight months is, is really a good release cycle, but maybe they're even longer.

A

Now people want to bring that down to a few months, to maybe few weeks to days to a maturity level of netflix, where you are able to make multiple releases a day and spinnaker is a tool that allows you to achieve that maturity level.

A

uh So there are quite a few continuous delivery tools out there. So what's the benefit or advantage of a spinnaker, it includes support of deployment to multiple cloud environments, be it kubernetes or flavors of kubernetes like openshift or aws azure, and and that's just to name a few of the major ones. uh So it abstracts out the cloud and allows you to deal at an abstraction level uh with each of these clouds.

A

I'll take a few minutes right after this to show you spinnaker, so that you know you get a hang of it before we get into how we do manage releases uh like I said it allows you to do automated releases safe deployments. Like you know, if you are doing that many releases a day, how do you know that a release went out, that's bad, and then how do you recover from it?

A

So there are built-in rollback mechanisms so that you can test, and you know if you find out that your new release is not as good. You can easily and quickly roll back automatically, and the other nice thing about this is that it provides you, visibility and control across your applications uh across the different cloud providers. You don't need to log into different uis consoles for different vendors and providers, and and have visibility and control over what's going on and especially in context to the application.

A

So it's a foundational platform for continuous delivery that can be quite easily extended to match your uh requirements.

A

Maybe at this time we'll I'll spend a few minutes to really show you what spinnaker is uh some of you may be aware of it uh if you're not aware of it, it's really good to those of you, um probably I'll. Take four or five minutes to spend uh to to show you. Spinnaker uh spinnaker actually has applications right. Applications are uh things that your departments, your organizations, deploy your sales applications, your marketing applications.

A

So if you go into any applications, you can see what are the infrastructure behind it right. You have your clusters, which are your server groups that serving that applications you can have load balancers behind them. You can have firewalls even right now, spinnaker stop started supporting serverless functions uh and you can manage them right from here, so you can think about an application being represented here as the inventory of all the resources that are part of your application. Right, uh like you can see here, you can have different uh providers here.

A

You can manage them from one, so you can have the same application being deployed to different uh environments, some to aws some to google cloud, some to uh some kubernetes environment. So only you can do all that in the context of the same application and you can have also different accounts in there. The main thing is especially when you're doing continuous delivery.

A

You don't want a developer to push code straight to production right. You want dates in between. uh So that's where accounts come into play and you have access control uh to control who has access to what accounts.

A

So what you can control is that users who have the permissions can deploy to dev or stage environments, but only the ops folks can promote to uh production.

A

Just to quickly show you, uh you know um the ability to view and control it. uh So if I go look at this say there is this application uh right over here. uh Let me just find one that I, that seems to be more current.

A

Let's say this application. That's over there. I can see all aspects of that application. I can control this application right from here, including scaling it out uh if some- and it will also show me all the different like instances running in that application, this being a kubernetes application. These are the parts that are running and I can see even the output of those fonts right right from here. So I don't need to give out cube cutter files and all that for people to access.

A

Now, how do I get these uh uh infrastructure components uh up and running it's by virtue of these uh pipelines? That's really the power of spinnaker, where, as you make releases, you can create pipelines.

A

So what you can do is you know you can define your workflow in a ui way, but spinnaker is basically what it you can call as uh pipeliners code. uh It enables github type operations, so you can have all of these as a code.

A

Everything is represented as a json file, so this whole pipeline, the visual representation you are seeing is available as a json file as part of the pipeline, you can configure and have automated triggers triggers can come in anywhere from a git check-in or to a new image being published to a docker repository.

A

uh You could even have them scheduled to run at frequent intervals right. You can pass in parameters to your pipeline so that you know you can use the same pipeline to be applied to stage one first and stage two or even stage and production to just uh vary the parameters and uh deploy them, and then you can have also notifications being sent out at something goes wrong. It didn't it integrates with slack email.

A

So once you have configured the initial setup, you can add stages that do different. Things like this is a stage that calls a jenkins job.

A

This is a stage that actually deploys to uh kubernetes, so there are tons of stages built in first class objects or actions that are built in that allows you to build a pretty complex pipeline and also a spinnaker allows you to extend it very easily. In fact, you know, part of my demo later on would include extending spinnaker to enable you to do a release management.

A

So that is about building a pipeline where you can build a complex pipelines and run them and once they are run, you can actually monitor what's going on over there.

A

You know this is all good, so not much to look at, but if something went wrong, ceo jenkins will fail. You have a link straight from here to jenkins. You can go and look at the jenkins pipeline uh to see what went wrong in log file also uh as part of the jenkins bill.

A

It did a bill and it actually passes back the build number which is used as a tag in the image, and this is used in subsequent stages to deploy that image right to make sure that you are deploying the image that was built.

A

Similarly, you know you can see the results of uh deployments like this was a a deployment that happened. You can go. Look at the actual yaml got deployed. You will notice. Spinnaker adds a lot of extra annotations so that it can manage these resources successfully and you can even go look at the the resource, so it will show you that is a resource right over here. uh Everything about that resource.

A

So that's, in short, like you know what spinnaker is about it, basically it it's like an uh inventory of your application. uh It shows you all the infrastructure behind that application and then it has pipelines for you to uh deploy that application in a continuous fashion right.

A

So that's a good tool, uh spinnaker right, it does continuous delivery. Now, like I showed in the previous slide, each company organization is at a different maturity level.

A

So what if I am not doing continuous delivery, I do not want uh code moving in from checking by a developer to production in a matter of minutes, if not in matter of hours, if not minutes.

A

What? If I want to choose what I want to promote to the next stage? What if I want to choose the features that want to go on to the next stage? What if I want to plan and schedule a release where you work with the stakeholders, product managers qa leads to define, uh what's going out and then also have these different stakeholders sign off. What needs to be promoted, sort of like an approval process is spinnaker not for me, then.

A

Yes, spinnaker, you can use spinnaker and not without, like you know, writing a lot of custom scripts and all that you can use spinnaker to do exactly for what what you want it to do.

A

uh The question that comes up hey, there's a lot of tools out there that orchestrate this process. I just talked about all these- manage the releases. What goes into a release but talking to a lot of companies unless you have built a custom solution.

A

Yes, those processes are there, but they automate the business aspect of it, getting approvals from different process providing inputs as to what features are in there. But that tool doesn't really integrate very well with the actual deployment process, at least a few customers.

A

Where I interacted with uh you know they have the process and then, when it's ready to be done, they create a jira ticket assigned to someone. Then that person gets the jira ticket run some scripts. Once it's successful, they assign the jira ticket back to someone to do the next part of the delivery process.

A

So that whole process becomes quite long and in in the process you know sometimes a step gets, missed or or the ball gets dropped and you have a broken release, uh so the cost of delivering a release that way is pretty high right.

A

It would be nice if you could actually integrate all that business process associated with software delivery along with actually delivering the software right, and that's what you know. I will look at how you can do that using spec.

A

uh So an approach to do that using spinnaker is you know you use a bill of materials to define what gets released based on everything uh you know your say: release manager decides that you know of this 10 releases, I'm going to release version a of the software uh version, bro version a of a component of a software version b of a component of a software, and then you integrate all the approval process in it.

A

Like you know, building it then deploying it to some staging environment run tests with it, have people validated uh and then get approvals to move it on to the next stage right. All that can be achieved using a spinnaker pipeline. What does it look like? I showed you a pipeline, so example of such a spinnaker pipeline could be. Is that I manage my this bill of materials and I'll come to an example of a bill of materials, so we used a few customers.

A

Is I basically manage the bill of materials in some source control right once I make a change to that and commit it. It's propagated.

A

It triggers this pipeline, which actually would read that file and inject it into the pipeline, which would go ahead and deploy the different services, the version of the services.

A

Then you do a validate of deployment and, if you're ready, you will promote it to the next stage right.

A

Let's look at what an example of a bomb can be like it could have all the list of services that's being installed.

A

You may not install all service in the release right or or promote all the services, so you just determine if the service is being released or not, and what's the version of the release or the image that's going to be published here would be doing it with a kubernetes application, and you can also include other parameters that may be part of your release.

A

Things like hey. I want to make this release to only say uh u.s region, or I want to release this only to uh india region right.

A

So all those also you can include as part of this bomb file, and I have picked up your pipeline pick it up and deploy uh having this pipeline being source controlled, it's versioned, so you will know what got released when uh you can have an approval process right if this is managed and get you edit this, and this becomes your source of truth of what's being released and approved, and once it's approved and merged it can trigger of your uh deployment and being part of gate or any other source control.

A

You would have an audit history of you know who made change what changes were made. So it really gives you a very good idea about what's being released. Okay,.

A

Let's look at a demo of that, uh but before we go there uh you know you could include even approvals at that process. Right approvals could include, you know, uh integrations with jira or service now or policy checks with confirm uh for uh uh conformance.

A

Like you know, before you release anything, uh you need to have sonar, cube coverage of, say, 80 or better, or you want to have your image scanned with no critical or high vulnerabilities detected.

A

So you can have all those checks built in not rely on someone to run the check and give you a thumbs up automatically included and then finally, especially when you're promoting to production, uh if you are not ready for continuous delivery, you want some manual checks in there before you promote and that's what you can actually do so to look at a demo um here. I have a pipeline just like we talked about.

A

Let's look at this one here.

A

You know this is exactly the pipeline. I talked about it's uh triggered by a bomb file change which I will show it to you and make a change and trigger it. It takes in a parameter of a namespace so that you know to show that I can control deployment to different uh spaces.

A

uh Then, moving on to the next stage, this is where the sort of the secret sauces, where we leverage spinnaker to actually read the file and uh and inject the contents of that file into a spinnaker pipeline. You know here is a little script to show that, but typically what you would do is you would encapsulate this in a custom stage and make it available in spinnaker.

A

I have it here so that you know I can use it to show people what exactly is going on then, based on that, you know you deploy it to your different services. You know I'm taking a very simple case of deploying an nginx and postgres, and you know here you will see that the image is actually I am making it a parameter, and this parameter was injected into the pipeline. In this stage,.

A

Once I deploy the services I'll do a validate if everyone's satisfied with the validation, then I'll, promote it uh to production and then in deploy to production, it picks up another pipeline. Does that does exactly the same? So in this, while I'm deploying to production, I am passing in the same bom file, uh the reference to the bom file and the version to the deploy to production so that it picks up the exact same bom file and pretty much goes through the same process injected the contents into the pipeline and deploy the services right.

A

Let's go look at the bom file, which I showed you pretty much like this right. You have the different services, the images and you know. Let's say uh I want to decide that. Okay, I'm ready we'll be upgrading everything to the next version. So I go to that nginx, I'm moving to the next version. Postgres are moving to the next version. Everybody looks at it, comments, everything and then okay. I approve the change it's getting committed.

A

When I go back and look at my pipeline, I see that it has started running it triggered this pipeline and it is running, and so what this is doing right now is going and fetching the details. So, while it's running, let me show you some of these from an existing pipeline that run already.

A

So, if you look at this pipeline, this is the stage that run and when I look at the output of the stage, you will see that it actually fetched that exact file content right over here right and then these are made available into a spherical pipeline so that anyone can ask or use these as variables.

A

uh Next, uh actually, you know that it did it. I have this evaluate variable stage to show that, uh like you know, if I look at what ngnx is, it tells me just this image version install true namespace. It shows me all those values in there.

A

It also shows me the trigger that got passed in it tells me this is the reference to the kit. This is the version that came in, so I can use that in in subsequent stages to fetch it.

A

So here we see that it all ran over here right and it's waiting for me to say. Should I promote it to the next stage or not? Let me go ahead and do that. I say: okay, everything looks great, let's promote it and I say continue.

A

What we'll see is that I have the second pipeline right over here that will get triggered and if I look at the parameters, it gets parameters past the bomb file and the version that came in from the previous stage, and so it will pick these up and start deploying.

A

uh Let's look at the infrastructure if you remember, we had earlier versions and the new version should have got deployed.

A

Let's wait for this page to refresh uh because we asked it to deploy the next versions of it. ah Let me refresh this piece: spinnaker does take a few minutes to refresh and then sync this uh information. So here you can see that spinnaker identified that the next version. Actually it's getting bigger to target. Let's look at target.

A

uh You will see that nginx 117 is already deployed over here and the old version has been started right.

A

So what you see here is how you can use spinnaker to control what versions of all the services get deployed, not necessarily have continuous delivery right. You plan releases, determine what releases would get promoted and promote them through the whole process in an automated fashion.

A

That's all I wanted to demo for now hope you found it useful, and you know if there are any questions, uh I'd be happy to uh answer them right now, hello. um Hopefully uh you found the session informative. um So uh if you have any questions uh on the session, you can type it in the chat and I'd be happy to answer them yeah for everyone.

A

If you have any questions, um you know you can use the chat uh to ask it and I'd be happy to uh answer uh any questions you have, or you know, if you have any other questions later on, feel free to reach out to me. um I think it may have been in the schedule, but it's really my first name at opsamextonio as well. You can reach out to me uh there's this question uh about. Does spinnaker work alongside jenkins? Yes, absolutely spinnaker integrates with a lot of uh existing tools.

A

You don't need to really uh throw away and rewrite everything you have. You can use your existing infrastructure, including jenkins.

A

uh So uh if you have uh jenkins, there are stages available that you can use to invoke uh the jenkins to, like you know, do the build jobs and all, and even the it was the other way around is that you can trigger a spinnaker pipeline to do a deployment once uh jenkins job successfully completes.

A

So there is that both bi-directional integration available uh with spinnaker so hope uh that answered the question for you josh.

A

um There is another question: how to start with hello worldish app with spinnaker um installing spinnaker is very easy. um You know um there is a website, or there is a blog that I have written and published.

A

um If you go and search on the opsamex blog blog.opsomix.com about installing spinnaker with helm, uh you will find a blog that you can use to install spinnaker in a kubernetes cluster in less than 10 minutes and create a hello world uh type application. uh It actually does walk you through deploying nginx. You know, step by step. Really people have followed it and then been able to get up and running with spinnaker, so go to blog.opsimx.com and search for installing spinnaker with helm or spinnaker helm, and you will find the blog.

A

There is another question: what are your thoughts on using and or a miniker? Is it good tool to deploy or just stick with spinnaker?

A

uh I think uh you can stick with spinnaker, so what menika tries to do is just you know, provide a developer version of it, uh bring it up, but you know if you already have a kubernetes cluster available or even have mini, cube running on your machine, uh it's really about running and helm chart and that will get you up and running with spinnaker.

A

uh So what you should do is you know the blog that I just talked about? You can go there and just uh run the blog in less than 10 minutes. You would have a spinnaker up and running so stick with spinnaker uh how's the install process for spinnaker uh similar question so which is good, I think, um just to play around with spinnaker. It's simple easy. um You know really, if you have a kubernetes cluster available less than 10 minutes really, in my view, um using the helm chart that's available so uh now.

A

Obviously, if you want to go to production, um have thousands of pipelines running? uh You need to do some planning around that? uh How you scale up the components how you set up for aha and all that and that's where I think we come in and help you can help you guys, but you know just for playing around and using it it's it's really uh simple.

A

There's one more question: does app summits provide service to help with proof of concepts? Absolutely you can, and you know any tool like this is simple as well as complex. uh You know it. It takes a while for you to understand the tool uh how to use it. So that's where definitely we can come in and help you, you know, understand your use case and how you can uh best leverage the power and features of spinnaker to your advantage.

A

So you know, if you think uh you need help, definitely reach out to us uh at ops. Okay, we do have a booth uh in our uh in in the booth area, so definitely drop by or reach out to us at opsonics.com.

A

Okay, I think I'll take one last question, um because the next session starts in five minutes and this one is spinnaker was built as a cloud agnostic, how much how it is much flexible with non-cloud environments.

A

Yes, spinnaker has been built for cloud environments, but spinnaker legacy is very extensible right. uh So if you have apis available to interact with your non-cloud environments, you could definitely um use uh spinnaker to interact with and deploy to non-cloud environments too. It's a very extensible tool that you can integrate and extend to do other things.

A

So with that, I I think we're out of time. uh Thank you, everyone for attending thanks for all those great questions. I hope you found this uh informative and if you have further questions, you can either reach out to me directly at nermalia, opsamex, dot, io or visit our booth.

A

Thank you and enjoy the rest of the virtual summit.