Hello everyone, I would like you to welcome to this talk. Today we will be discussing about feature flags in Terraform. Before we begin, I would like to introduce myself. I'm Pravesh Sapper and I've been working in the field of SRV and DevOps for quite some time. The key topic we'll be discussing today are what are feature flags and why should you use them? How to implement feature toggles in terraform and how we can pass feature flag from configuration file and we'll do a quick demo. So what is terraform? Terraform is an API driven infrastructure as code tool created by Hashicorp. It uses declarative language called etc. As we all know that features have been implemented in software engineering for quite some time. Let's talk a little bit about them. So what are feature flags? It is a software engineering technique which allow you to enable or disable feature without additional code. They are also known as toggles or switches. There are various benefits of using feature flags like running rollout or rollback operation or allowing developer to test code in control environment. Feature flag are passed from either configuration file environment variables or dedicated feature flag management platforms such as launch darkly. Next we'll discuss about why we should use it in our infrastructure as code. The first reason is it increases the flexibility. Feature flag allows granular control over rollout of changes and can be used to turn specific feature or functionality on and off for different environment and different user group. This allows for more efficient and flexible development, testing and deployment of specific feature, increasing the flexibility. You can selectively provision resource based on their environment using feature flags. That is, you can have a smaller instance on dev but larger more capable instance on prod. Using feature flag in infrastructure as code it helps to improve scalability has we know that feature flags can be used to enable or disable sorting feature or functionality based on current load on your infrastructure. Now what it does is it allows us to scale infrastructure up or down as necessary without need to make changes to the code, hence improving the scalability. In general, using feature flag we get better control of over our deployment using techniques such as blue, green or canary. We can test and validate changes in a controlled environment and quickly roll back changes if necessary. It also helps to improve safety by reducing the risk of introducing breaking changes. This helps to ensure that your infrastructure is always in a stable and working state. Using feature flags as part of your infrastructure as code configuration, we are able to perform effective monitoring and debugging. Let me tell you why, because if we use feature flag, we are aware which changes were introduced incrementally and knowledge of this allows us to isolate specific issue of feature or functionality, hence improving the debugging. Let's look into how we can achieve this. The key player is count meta argument. It allows us to create a specific resource, a specific number of identical resources. Here in the example I have a resource block to create AWS instance. You can see in the count argument that we are asking terraform to provision four identical AWS instances. One key thing to remember here is we can pass zero to count as well, which means if we set count to zero, then the resource will not be created at all. Another key player in the story is conditional expression. Conditional expression in terraform uses value of a boolean expression to select one of two values. So using count and conditional expression, we can implement feature toggles in terraform. Let's look into some implementations. First we will look into creating environment based feature toggles. So basically environment based feature toggles are those kind of feature flags which helps you to toggle between feature based on the environment you are deploying to. To do that, we will first create a variable named n and then we will use that variable along with count and conditional expression in our resource block. Here I have two resource block one to create a small or puny Vm and another to create a beefy vm based on the environment. If the value of n is dev, then the value of count in puny Vm is set to one and hence smaller Vm is created. And if the value of n is prod, then the value of count in beefy Vm is set to three and hence three instance of BFBM is created. We'll look into this in the demo as well. Now let's look into how we can create feature toggles based on specific value to toggle specific resources. What we can do is similar to the thing that we discussed before. We will create a local variable to enable that flag and then pass it to the resource just like we did. For environment based toggle. The working mechanism is same, but instead of relying on environment here, we rely on a specific feature flag to enable or disable creation of that resource. Another way you can implement feature flag are at module level. We create a flag for that module and pass it directly to the module instead of passing it at resource level. This has been possible after terraform 00:13 which was a good leap forward. The usage is similar to implementing feature flag in resource. We created a dedicated flag, but instead of passing to a resource level, we pass the flag directly to the module and if the flag is set, it will provision the resource. There are other ways you can use or leverage feature flag like you could use it for blue green deployments using floating ip in this lotion. Or if you are using AWS, you can perform canary release using AWS load balancer target group. I mean, possibilities are endless. Let's talk about how to organize these flags. There are various ways to do this, but first one is defining the feature flag in local variable. We saw how to do this in our previous slides as well. We define a flag as part of our local variable and then use them as needed in other resources or module. Another way we can organize feature flag is using configuration file. This is my personal favorite as we can get a complete information about the infrastructure without having to dig down deep into variables or terraform configuration file. Here we define variable in configuration file and then load this variable into terraform and use it as needed. It's kind of neat, right? Anyways, let's move on to the demo. Okay, in this demo we'll be looking into the things that we discussed in the slides. So I have environment based toggle created which is a variable named n and I'm passing that n variable to our droplet. Let's look into the droplet. So I pass the droplet into the count with conditional expression saying if the value of environment is dev then it will provision a smaller vm. And if the value of environment is broad then it will provision of beefy vm. Let's run terraform plan and see what happens. So as you can see that the value of inn was defaulted to dev and based on that it provisioned puny vm which can be verified from the output. But let's see what happens if we override the value of n to prod. So as you can see, it detected the environment is prod and provisioned beefy instance. The another thing that I want to show you today is the feature flag using configuration file. So I have two feature flags created here which provisions the load balancer and provisions the database. Let's see here in the module. Basically what I'm doing is to provision a load balancer. I'm passing count to module itself. So if the value of provision lb or value of provision load balancer flag is set to true, the count will be one and it will provision the load balancer. So let's set the value of provision load balancer flags to true and run the code again. Now we can see it provisioned the load balancer for us. Let's set the value of database to true as well and run the code. If I run terraform plan again you can see it added four resource to change and if you scroll up and see it has created the database as well. So that's all for the demo. Before we end the talk, here are the key takeaway I want to repeat again which are feature flag is a very powerful technique which has been implemented in software engineering for a long time. I think it's time we start using it as part of our infrastructure as code as well as it helps in adding flexibility, scalability and maintainability to our infrastructure as code. Here's the link to the source code shown in the demo if you are keen a look and here are the references I used to prepare the talk. Thank you for listening. I hope this talk was useful. I would like to connect with you all and talk more about terraform feature flags and DevOps in general. Here are the links to my socials. Thanks again and enjoy the rest of the conference.