Terraform is our main infrastructure as code at the bank. Our entire deployment site and the AWS solution are based on Terraform. We use EBridge and the Dochub solution on the Bitbucket pipeline. Terraform runs on a container in the Bitbucket pipeline, but to create my entire cloud solution, I use EBridge.

Terraform centralizes all our applications and it’s the best way to do it. When you work with the container and the EC2 clusters, you need to create each service manually. Terraform allows you to create a lot of them at once and check if they are being created correctly. It is a huge timesaver.

I find almost all the features valuable. It is the customization of Terraform's modules that I find most valuable.

Terraform could create more examples in the documentation. It is an enterprise/free solution, and you have to do a lot to customize the tools. A huge example I faced before that drove me nuts was when I created an entire data lake using Terraform. A DMS solution using Oracle didn't read some options in AWS on the Terraform module. I opened a ticket to support, asking, "Could you improve this module, only adding these features as variables?" After four months, my ticket was closed by a bot because support was not looking for it. I don't know if there were many issues or tickets, but support should listen to the Terraform community better and make adjustments to their tools.

I've been using HashiCorp Terraform since 2017.

Terraform is stable. I don't use Terraform Cloud, though they offered it to me previously. Using other clouds, we don't face many issues or instability in our ecosystem. The only issue or instability I have faced was because the AWS cloud was unstable, so Terraform got timed out and could not finish the task, trying again and again and being stalled for two hours. That is the only issue I remember because we use the EBridge environment. We try to compensate for Terraform's instability and make it viable using other platforms. Using Terraform's cloud solution brings more stability, but for what we are using right now, Terraform only breaks if we make a mistake in the code.

Terraform is hugely scalable. If I have everything in place, you can create an entire environment, like a data lake, in one simple deployment. But it is complex to scale. You will face some limitations if you try to work with multiple servers or accounts. Terraform works similarly to a FIFO context, but you must wait a long time to create the entire deployment before going to the next account. If deployment on one account fails, you cannot proceed to the next one because Terraform's main focus is on the security of the infrastructure. My entire squad of four uses Terraform because we centralized the DevOps for Terraform. We are planning to expand this tool to our entire development team. We will have at least 15 to 30 developers using the tool in the future.

Their customer support is the worst. I opened a ticket, and I never got an answer, and the community does not listen to the most common issues. But I understand why I was left out because I asked a hugely specific question about a little bug in the code. My experience was not the best, but I no longer need to ask them because I make my own workarounds. Besides, we don't face issues that require us to talk to support.

The customer should receive an answer regardless of the question. Even answers like, "This isn't a feature," "We can't do this right now," or "This isn't the roadmap."

Neutral

I compare Terraform with Ansible. I work with multiple servers in Ansible in a FIFO method. You have a list of servers you apply one after the other. But you can work with those servers in parallel, even if one server fails to deploy. In Terraform, you must finish deployment on each account before going to the next one.

When I joined my current company, they were already using Terraform, and we had to create the environment on AWS manually. We are not using any IaC tools.

The initial setup was complex. You need to understand what you want to do exactly. You are a child in a playground with many toys to play with, and you can create anything you want. But if you pick up the wrong "toy," you can hurt yourself and the environment. Using the wrong module can bring a huge cost to your infrastructure, and it won't perform as well as you initially expected. Initial setup is more difficult to start if you have some infrastructure completed.

If you are setting up from the start with baby steps, it is a little hard to understand the documentation, though it makes the setup look much simpler. You need some knowledge to understand how Terraform can apply to your environment. After that, it's easier. And just like riding a bike, you never forget.

The time taken to deploy the solution depends on what you are creating. For example, if I create a DMS instance for my data lake, it takes 15 minutes to deploy. If I'm creating a different data lake on the Athena database, it takes exactly 23 seconds to create. A DMS takes much more time, but not more than half an hour.

While deploying Terraform, we checked the Terraform statement on AWS. Each account has its own statement. We checked the difference between statements, made a Terraform image, made a plan, and checked the plan to see if it was correct and applied.

I did the deployment myself, looking at the documentation.

We use the free platform using Dochub. Terraform is a little expensive. In a basic scenario, the price is too high to use such a complex tool in the long term. As a small company starting to use Terraform, the price is too high, so I recommend the free solution. If you are a huge company with a huge environment and infrastructure, the price will seem cheaper because you can control your entire infrastructure using Terraform. The solution then pays for itself. Choosing the paid version depends on the size of the company and your team's expertise.

Terraform licenses are per account.

We currently use a statement for replication, trying to embed Terraform into our applications, but that's a new feature. We are applying some other features, for example, to check for vulnerabilities on your first code or to scan if you have hard-coded passwords, but we have not found anyone to help with these use cases. One of my co-workers is certified on Terraform, and they bring similar solutions also being used on Terraform for scanning. For myself, I want them to make that code work on as many accounts as possible. The same code must work on our deployment, homologation/staging, and production accounts. These are the three environments the code needs to work in. We tried making this a few months ago.

When deploying the solution, I did it by myself, looking at their documentation. Their documentation is good and bad. Their documentation is good because it gives a huge amount of information for several different possibilities, but it's bad because the documentation does not have a lot of real-life examples. Terraform prefers documentation as much as hard-coded information. For example, you might be using a certain string to do something, but in reality, you can do that in many other ways, and the documentation won't show all of those. You need to do trial and error or take a course with the Hashicorp organization to understand the different ways you can do something or use the platform.

I rate Terraform a seven out of ten because of its complexity.