DEV Community: Mohamed Wasim

Deploy Fider as a Private App on AWS with CloudFront VPC Origin

Mohamed Wasim — Wed, 26 Mar 2025 04:49:33 +0000

When architecting solutions on AWS, minimizing the attack surface and ensuring secure deployments is essential. CloudFront VPC Origins provides a way to deliver content from private VPC subnet, reducing direct exposure to the internet. This makes it an ideal choice for those looking to host a private app while securely exposing it to the public internet, offering a more secure alternative to exposing applications from a public subnet.

In this guide, I’ll explain:

What is CloudFront VPC Origin
Why you should deploy Fider as a private application
Create an Internal Application Load Balancer
Create CloudFront VPC origin
Create a CloudFront Distribution
Create Secrets manager secret and Lambda function
Create a Custom Header rule in AWS WAF
Spin up ECS and RDS
Monitor CloudFront and WAF logs
Do You Really Need Zero Trust for Every App
Is CloudFront Truly Zero Trust Compliant
Deciding on Enterprise Grade Deployment for Fider
Caveats to consider
Final thoughts

1. What is CloudFront VPC Origin

Traditionally, when using CloudFront with an origin resource (such as an ALB or EC2 instance), the origin needs to be in a public subnet with a public IP address, implementing Access Control Lists (ACLs) and other controls to restrict access effectively. Users needed to invest ongoing effort to implement and maintain these solutions, resulting in undifferentiated heavy lifting.

However, With CloudFront VPC Origins, users can host their applications in a private VPC, without requiring any direct route to the internet and make sure CloudFront is the only entry point to their applications. When CloudFront VPC Origin is set up as an origin for the CloudFront distribution, traffic stays on the high-throughput AWS Backbone network all the way to your AWS origin, making sure of optimized performance and low latency.

This is designed to prevent end users from discovering or bypassing CloudFront to access web applications directly. By implementing network-level isolation, the origin servers remain hidden on the internet (Obfuscating AWS resources), significantly reducing the attack surface (attack surface reduction) and enhancing the overall security posture. At the same time, users continue to benefit from the CloudFront global scale and high-performance capabilities.

2. Why you should deploy Fider as a private application

Alright, here’s the deal: With this setup, even your ALB can be internal, and it'll route traffic straight to your ECS instances in those private subnets. You heard that right, no public access. This keeps your backend services completely hidden from the internet. User requests go from CloudFront to the VPC origins over a private, secure connection, providing additional security for your applications.

Now, with CloudFront VPC origin, you can stick that ALB in a private subnet where nobody can get to it directly. CloudFront is the only way in, so you can rest easy knowing the attack surface is way smaller. Plus, that ALB? The DNS name only resolves to private IPs, so no internet users can mess with it. By doing so, you significantly reduce the exposure of your internal ALB to DDoS attacks.

Even if someone were to somehow discover your ALB’s ARN (which is very unlikely), they still couldn’t send traffic to it using their own CloudFront distribution. That’s because internal ALBs are only accessible from within the same AWS account and VPC.

And if you’re really serious about security, you can slap on Origin Custom Headers to make sure only your CloudFront distribution is allowed to talk to the ALB. This header verification ensures the traffic legitimacy. But wait, it gets better, those custom headers can be dynamically rotated using AWS Secrets Manager and a Lambda function automation, keeping everything secure without manual intervention.

3. Create an Internal Application Load Balancer

Click on create load balancer, and choose Application Load Balancer

Under Basic Configuration choose Internal as scheme,

Under Network Mapping choose the VPC you created or refer this guide to create a VPC, creating a single public subnet should suffice this time (for placing NAT gateway).

Choose to launch the Application Load Balancer in at least 2 Azs, in private subnets.

Choose the Security group for your internal ALB and create an HTTPS listener,

Navigate to the VPC console and click on the Managed prefix lists, choose pl-b6a144df and map this to your ALB's inbound SG rules(HTTPS), by doing so you are only letting the CloudFront's IP ranges to communicate with your ALB. This is the recommended approach according to AWS Best Practices for DDoS Resiliency.

Under Secure listener settings choose the certificate you created from ACM, if not you can refer this guide and forward to Step 3.

Once you are done with requesting the public certificate in ACM, your certificate will appear in the drop-down select it, scroll down, and click on create load balancer.

The internal ALB is now created with the traffic originating from the CloudFront IP ranges.

4. Create CloudFront VPC origin

Now that you are done with the internal ALB creation, Navigate to the CloudFront service console and choose VPC origins from the navigation pane.

Click on VPC origins and create a VPC origin,

Provide a Name and choose internal ALB's ARN from the Origin ARN drop-down,

Choose the protocol to HTTPS only and click on create VPC origin.

The VPC origin creation is now complete.

5. Create a CloudFront Distribution

Navigate to the CloudFront service console and click on Create a CloudFront distribution,

Under the Origin section, select the VPC origin you created. Enable Origin Shield for an additional layer of caching, and don't forget to add the custom header, name it x-origin-verify.

Under Default cache behaviour choose the following settings, under Allowed HTTP methods choose the third option, if you want to know why then check out this guide for reason.

Under Cache key and origin requests choose Cache policy and origin request policy (recommended). Choose the default cache policy and origin request policy. Under Response headers policy choose CORS-with-preflight-and-SecurityHeadersPolicy, it is an AWS-managed response header policy. Use this policy to specify security-related HTTP response headers that CloudFront adds to HTTP responses that it sends to viewers.

The AWS-managed policy provides the following security headers, it ultimately narrows down to your use case, you can also create a Custom policy and attach it under Response headers policy.

Enable Web Application Firewall for your CloudFront distribution, toggle on use monitor mode you can later disable it once you get know how much of your requests are being blocked by this WAF configuration(False positives and red herrings).

Check SQL protections and Rate limiting(DDoS protection) and take a note of price estimation.

Add the CNAME through which you would like to access your application publicly and choose the custom SSL certificate for your CNAME.

Configure standard logging for your CloudFront distribution to monitor the performance and usage metrics.

The CloudFront distribution creation is now complete.

Navigate to the WAF & Shield console, click on Web ACLs and choose Global(CloudFront) in the region drop-down. Click on the Web ACL and make a note of the rules and WCUs.

6. Create Secrets manager secret and Lambda function

Choose secret type:

Navigate to the Secrets manager service console, and click on Store a new secret.

Under Secret type choose Other type of secret, and provide HEADERVALUE as a Key and paste the same header value you used in the CloudFront distribution. Click on next.

Configure secret:

Choose the secret name and click on next, take a note of Resource permissions step, we will come back to this step once we are done creating the lambda function.

Review:

Skip to the review section leaving behind the configure rotation step, as we will configure that once we create and configure the lambda function.

The secret creation is now complete.

Create a rotator lambda function:

Navigate to the lambda console and click on Create function,

Choose Author from scratch and choose Python 3.9 as runtime with architecture as x86_64. Under Execution role choose create a new role, this creates a new service role for the function.

Under Additional Configurations enable tags and VPC, i am placing lambda function inside VPC to enable VPC flow logs to monitor network traffic, which is critical for auditing and incident response. Before moving onto choosing the security group create a new one in the EC2 console and then revert back here.

Since this rotator Lambda function will be triggered by Secrets Manager, it won't be needing any inbound rules, as it's event-driven. It still needs one outbound rule, to reach out to the Secrets manager endpoint via HTTPS protocol, as the traffic (for HTTPS) will be routed through public internet via a NAT-gateway.

Now, reference the created SG under the lambda's configuration parameters drop-down and click on Create function.

The rotator-lambda function is now created.

Configuration parameters - rotator lambda function:

Use the python code https://gist.github.com/WasimTTY/ff092402d09c56f6496b359036e42aa0 from this gist to deploy the code onto the lambda function.

This code was modified to work with the CloudFront WAF configuration.

Once pasted, click on deploy to deploy the code to the lambda function. Additionally, this code does come with an external dependency, so create a lambda layer and attach it to the rotator-lambda function.

You can find the artifact in this repository: https://github.com/aws-samples/amazon-cloudfront-waf-secretsmanager/tree/master/artifacts

Now, under the Configuration tab choose the Timeout for the rotator-lambda function, i have set it to 5 mins 3 sec with the memory and the Ephemeral storage left to defaults.

Take a note of the Existing role attached to the lambda function. we will be needing it attach the policies to let the lambda function interact with the CloudFront distribution and WAF for rotating the secret-custom-header.

Navigate to the IAM console and under Roles tab search for the service role created by the rotator-lambda function.

Create an inline policy for accessing the CloudFront distribution. I’ve provided only the permissions necessary for the service to function. This policy grants the rotator Lambda function the minimal privileges required to modify the Secret-Custom-Header on the CloudFront distribution.

  {
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "VisualEditor0",
            "Effect": "Allow",
            "Action": [
                "cloudfront:GetDistribution",
                "cloudfront:UpdateDistribution",
                "cloudfront:GetDistributionConfig"
            ],
            "Resource": "arn:aws:cloudfront::<AWS ACCOUNT ID>:distribution/E3R3ZKP77W4214",
            "Condition": {
                "StringEquals": {
                    "aws:PrincipalArn": "arn:aws:iam::<AWS ACCOUNT ID>:role/service-role/header-rotate-function-role-tinh6qnu"
                }
            }
        }
    ]
}

I’ve applied the same approach for the Secrets Manager, CloudFront WAF, and Lambda VPC policies as well, granting only the minimal permissions required for each service.

  {
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "VisualEditor0",
            "Effect": "Allow",
            "Action": [
                "secretsmanager:GetSecretValue",
                "secretsmanager:DescribeSecret",
                "secretsmanager:PutSecretValue",
                "secretsmanager:UpdateSecretVersionStage"
            ],
            "Resource": "arn:aws:secretsmanager:us-east-2:<AWS ACCOUNT ID>:secret:Custom-Header-lQWRrQ",
            "Condition": {
                "StringEquals": {
                    "aws:PrincipalArn": "arn:aws:iam::<AWS ACCOUNT ID>:role/service-role/header-rotate-function-role-tinh6qnu"
                }
            }
        },
        {
            "Sid": "VisualEditor1",
            "Effect": "Allow",
            "Action": "secretsmanager:GetRandomPassword",
            "Resource": "*",
            "Condition": {
                "StringEquals": {
                    "aws:PrincipalArn": "arn:aws:iam::<AWS ACCOUNT ID>:role/service-role/header-rotate-function-role-tinh6qnu"
                }
            }
        }
    ]
}

  {
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "VisualEditor0",
            "Effect": "Allow",
            "Action": [
                "ec2:CreateNetworkInterface",
                "ec2:DescribeNetworkInterfaces",
                "ec2:DeleteNetworkInterface",
                "ec2:UnassignPrivateIpAddresses",
                "ec2:DescribeSubnets",
                "ec2:AssignPrivateIpAddresses"
            ],
            "Resource": "*",
            "Condition": {
                "StringEquals": {
                    "aws:PrincipalArn": "arn:aws:iam::<AWS ACCOUNT ID>:role/service-role/header-rotate-function-role-tinh6qnu"
                }
            }
        }
    ]
}

  {
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "VisualEditor0",
            "Effect": "Allow",
            "Action": [
                "wafv2:UpdateWebACL",
                "wafv2:GetWebACL"
            ],
            "Resource": [
                "arn:aws:wafv2:*:<AWS ACCOUNT ID>:*/regexpatternset/*/*",
                "arn:aws:wafv2:*:<AWS ACCOUNT ID>:*/managedruleset/*/*",
                "arn:aws:wafv2:*:<AWS ACCOUNT ID>:*/rulegroup/*/*",
                "arn:aws:wafv2:us-east-1:<AWS ACCOUNT ID>:global/webacl/CreatedByCloudFront-4ffb490d-8e92-419b-a38b-017361fd80aa/a2512452-7f08-4591-bf1a-fec1923ac75a",
                "arn:aws:wafv2:*:<AWS ACCOUNT ID>:*/ipset/*/*"
            ],
            "Condition": {
                "StringEquals": {
                    "aws:PrincipalArn": "arn:aws:iam::<AWS ACCOUNT ID>:role/service-role/header-rotate-function-role-tinh6qnu"
                }
            }
        }
    ]
}

The IAM policies crafted above are identity-based policies scoped down to a specific IAM role using a PrincipalArn condition, enforcing least privilege.

The rotator-lambda function also needs a resource-based policy in order for the Secrets manager to invoke the function. I have used AWS CLI to add this policy to the lambda function.

  aws lambda add-permission \
  --function-name secret-header-rotator-function \
  --statement-id AllowSecretsManagerInvoke \
  --action lambda:InvokeFunction \
  --principal secretsmanager.amazonaws.com \
  --source-account <AWS ACCOUNT ID> \
  --source-arn arn:aws:secretsmanager:us-east-2:<AWS ACCOUNT ID>:secret:Custom-Header-lQWRrQ

The resultant policy will look like this,

  {
  "Version": "2012-10-17",
  "Id": "default",
  "Statement": [
    {
      "Sid": "AllowSecretsManagerInvoke",
      "Effect": "Allow",
      "Principal": {
        "Service": "secretsmanager.amazonaws.com"
      },
      "Action": "lambda:InvokeFunction",
      "Resource": "arn:aws:lambda:us-east-2:<AWS ACCOUNT ID>:function:secret-header-rotator-function",
      "Condition": {
        "StringEquals": {
          "AWS:SourceAccount": "<AWS ACCOUNT ID>"
         },
        "ArnLike": {
          "AWS:SourceArn": "arn:aws:secretsmanager:us-east-2:<AWS ACCOUNT ID>:secret:Custom-Header-lQWRrQ"
        }
      }
    }
  ]
}

The resource-based policy above follows the Zero Trust principle by minimizing trust, ensuring access is explicitly verified through conditions like Source Account and Source ARN before granting permissions.

Configure secrets manager - resource policy:

Now, navigate to the Secrets manager service console and under the Resource permissions paste the resource-based policy for invoking the rotator-lambda function.

  {
  "Version" : "2012-10-17",
  "Statement" : [ {
    "Sid" : "AllowLambdaAccess",
    "Effect" : "Allow",
    "Principal" : {
      "AWS" : "arn:aws:iam::<AWS ACCOUNT ID>:role/service-role/header-rotate-function-role-tinh6qnu"
    },
    "Action" : [ "secretsmanager:GetSecretValue", "secretsmanager:PutSecretValue", "secretsmanager:UpdateSecretVersionStage", "secretsmanager:DescribeSecret" ],
    "Resource" : "arn:aws:secretsmanager:us-east-2:<AWS ACCOUNT ID>:secret:Custom-Header-lQWRrQ",
    "Condition" : {
      "StringEquals" : {
        "aws:ResourceTag/Environment" : "production",
        "aws:SourceAccount" : "<AWS ACCOUNT ID>",
        "aws:SourceArn" : "arn:aws:lambda:us-east-2:<AWS ACCOUNT ID>:function:secret-header-rotator-function"
      }
    }
  } ]
}

This policy also aligns with the Zero Trust principle, Even though the request is coming from the rotator-lambda function (which could be considered a trusted service), the policy does not allow access by default. It enforces strict conditions including ABAC (Attribute Based Access Control) using resource tags to validate the request’s origin and ensure the caller is explicitly authorized.

Taking this approach tightens security by enforcing stringent, context-aware access controls.

7. Create a Custom Header rule in AWS WAF

Navigate to the WAF & Shield service console and look-up for Web ACLs created under the Global(CloudFront).

Create a new rule named XOriginVerify, and configure the same secret-custom-header value that was set in the CloudFront distribution and the Secrets manager.

The rotator-lambda function looks Explicitly for this rule name as it is hard-coded in the python code. You can modify it according to your use case.

Once the rotator-lambda function updates the secret-custom-header in both the CloudFront distribution and in WAF, it tests the secret by pinging your domain-name, in this case the request has to propagate through the CloudFront, and WAF will let the request through only if that header matches, it should result in 200 OK. You have to check CloudWatch logs for this.

8. Spin up ECS and RDS

You can refer to this guide for spinning up the ECS and RDS infrastructure.

9. Monitor CloudFront and WAF logs

Monitoring CloudFront and AWS WAF logs is a crucial part of incident response in any secure AWS setup, especially when you're running a public-facing app.

You need visibility into what’s happening at your edge and app layers. That’s where CloudFront and AWS WAF logs come in.

In the AWS management console under the region drop-down choose us-east-1, this is where CloudFront delivers logs to in the CloudWatch log-group.

Each log contains information such as the time the request was received, the processing time, request paths,client-IP and server responses. You can use these access logs to analyze response times and to troubleshoot issues.

When you're absolutely done observing the false positives for WAF rules, navigate to the WAF console and adjust the actions for the rules according to your use case. You can choose whether to set actions such as Allow, Block, Captcha, or Count based on your specific needs. It ultimately narrows down to your requirements and the level of protection you wish to enforce.

It’s important to note that when you set a rule group action to 'Count,' it will override any Allow, Block, or Captcha actions that were originally specified for the rules within that group. As a result, all rules in the group will only count matching requests, regardless of their original action.

By default, the action is set to 'Count' during testing and monitoring of the rule group’s behavior, which allows you to observe how the rules perform before deploying them to production.

AWS WAF now supports sending logs to CloudWatch logs, Hover over to the Logging and metrics tab and click on Enable drop-down and choose Logging destination.

Choose the CloudWatch logs log group as destination and click on save,

Logging is now enabled for AWS WAF.

Hover over to the Sampled requests tab to help you analyze how your AWS WAF rules are evaluating traffic. Each sampled request shows details like the IP address, headers, URI, country, and rule evaluation outcome. Sampled requests are free.

In addition to this, you can also use the default queries provided by WAF to query the number of requests that come from an IP address. Under the Cloudwatch logs insight tab in the WAF console.

10. Do You Really Need Zero Trust for Every App

Before slapping on Zero Trust policies like they’re the new fad, take a step back and do the threat modelling first.

That’s the only way to know whether Zero Trust is even the right fit for your app.

Threat modelling is like a security audit before you go shopping for locks and alarms. It forces you to ask:

What exactly are you protecting ?
Who would want to break in ?
How might they try to do it ?

That last part? That’s basically recon and attackers do it all the time. They scan your public IPs, poke at your APIs, look for leaky S3 buckets or overly permissive roles. So when you threat model, you're doing a defensive version of recon: identifying what an attacker might see and how they'd try to exploit it.

Threat modeling helps you spot risks like this, Once you know that, you can set up your Zero Trust stuff, like least privilege access, network segmentation, and strong authentication all tailored to the actual threats, not just random paranoia.

Even something as focused as the rotator Lambda function, which runs every 4 hours to manage the secret header, carries real risk. Its footprint may be small, but its role is critical. In cloud environments, even well-scoped serverless functions can become attack vectors if misconfigured or exposed.

This strict posture isn't just theoretical. Tools like AWSRedirector demonstrates how serverless components can be exploited in real-world attacks . That's why we lock down even the smallest moving parts.

Without threat modeling, you’d be like, ‘Oh, I’ll just block everything!’ but that’s like building a fortress with no doors. You gotta know who’s allowed in, and under what conditions. That’s how Zero Trust works it’s not about ‘trust no one’ for the sake of it, it’s about verifying every request based on real risks. You should understand the balance between security and usability.

If you’re curious about how AWS approaches Zero Trust in detail, you can check out their Zero Trust Security Model and AWS Zero Trust Architecture principles.

11. Is CloudFront Truly Zero Trust Compliant

Zero Trust architectures are identity-centric, requiring per-request authentication and authorization before traffic reaches the application. Using CloudFront means your application is publicly addressable on the internet, even if access is restricted with WAF, signed URLs, or custom headers. This introduces a public entry point, contrary to Zero Trust principles of default deny and private access.

CloudFront does not support integration with IAM Identity Center, Azure AD, Okta, or other identity providers (IdPs) to enforce access control. While it offers IP-level and path-level access logs, it lacks identity-based audit trails, making per-user auditing and continuous trust evaluation impossible. Additionally, it does not inspect or enforce device posture or context, meaning it cannot verify whether requests originate from secure, approved environments.

If you're looking to implement true Zero Trust at the edge, Cloudflare and Akamai (EAA) offer the most complete solutions today. Others like Fastly are catching up, but platforms like AWS CloudFront, Azure Front Door, and Google Cloud CDN still fall short of Zero Trust standards.

12. Deciding on Enterprise Grade Deployment for Fider

Fider is just a feedback platform, and most of what it contains are suggestions and feedback. We don’t store any sensitive Personally Identifiable Information (PII) like Social Security Numbers or financial data. Even in the event of a compromise, the risk is minimal, typically limited to names and email addresses. However, if you're using Fider internally within your organization, the risk could be higher, as internal feedback may include more sensitive or proprietary information

Given Fider’s risk profile, a Lambda-based secret header rotator backed by Secrets Manager offers a strong security baseline. This level of protection is more than sufficient if you’re using Fider to collect public feedback and suggestions without overengineering.

AWS Verified Access is indeed the right choice if you’re planning to host everything within AWS and require enterprise-level security controls. But, it’s generally overkill unless you have tighter access control policies in your enterprise setup, and costs more. You are also charged per GB for data processed by Verified Access. You can calculate the estimated cost here AWS Verified Access Pricing. It functions more as an access control solution than as traditional VPN tunnels. It only works with AWS hosted systems.

If you are curious about this setup then you can check this out https://pages.nist.gov/zero-trust-architecture/VolumeC/HowTo-E4B5.html

Cloudflare Zero Trust is the sweet spot and comes out as the better choice when considering both cost and performance. It supports edge services, DNS, and mTLS out of the box, providing enhanced security, performance, and seamless encryption with Cloudflare’s global network.You can check out its pricing here.

13. Caveats to consider

VPC Endpoints:

You can still choose to go with VPC interface endpoints (Powered by AWS PrivateLink) if you haven't integrated external OAuth providers (like Google, GitHub, etc.) because these services are not available via AWS PrivateLink.

When you configure OAuth providers with Fider, the authentication requests go to the OAuth provider's endpoints over the internet. This involves external APIs that cannot be routed through VPC Interface Endpoints directly, since these OAuth services are external to your AWS infrastructure and not hosted within your VPC or private AWS environment.

It will also cost you more than the NAT gateway, you can checkout this post from the official AWS forum for even more clarification. Only choose this method if you consider security as paramount.

⚠️ Network Isolation ≠ Identity Validation:

Using a VPC origin does make the app private, but it does not guarantee identity-based access control. CloudFront doesn't authenticate requesters the way solutions like Cloudflare Zero Trust do, leaving the application vulnerable without proper identity verification.

14. Final thoughts

Let’s be real no system is made of adamantium. Given enough time, resources, or misconfigurations, even the most locked-down architecture can be poked at. But this setup doesn’t aim for invincibility, it aims to minimize exposure, limit blast radius, and make exploitation significantly harder and riskier.

This architecture trades convenience for control, reducing your attack surface, slowing down adversaries, and giving you time and telemetry to respond.

Deploying Fider on AWS ECS: A Step-by-Step Guide to Deploy a Feedback Platform

Mohamed Wasim — Tue, 05 Nov 2024 12:12:17 +0000

This guide provides detailed instructions for deploying Fider on AWS ECS, with tasks running inside a private subnet and an internet-facing Application Load Balancer (ALB) set up as a custom origin for CloudFront(Cloudfront + WAF).

This is a detailed article, so buckle up! Since this setup closely mimics a production deployment and let’s be real, deploying Fider on AWS isn’t an easy feat, you’ll want to follow along carefully.

AWS services used

Elastic container registry(ECR)
Elastic container service(ECS Fargate launch type)
Application Load Balancer(ALB)
Relational Database Service(RDS)
Systems Manager Parameter Store
Simple Storage Service(S3)
Simple Email Service(SES)
Cloudfront(CDN)
Web Application Firewall(WAF)
Key Management Service(KMS)
Amazon Certificate Manager (ACM)

What is Fider ?
Fider's Tech Stack
Pre-flight Checks
Install AWS CLI
Push the Docker Images to ECR
Create a VPC
Create Subnet Group for RDS
Create an RDS PostgreSQL DB
Create Parameters in Systems Manager Parameter Store
Create an IAM User
Create an S3 Bucket
Create an Identity in SES
Why Use ECS Instead of EC2
Create an ECS Cluster
Create Task Definition
Create an ECS Service
Create CloudFront Distribution with WAF

What is Fider ?

Okay, okay, so, listen up! Fider is like this thing where people can go and, like, share their opinions on stuff, right? Imagine you’ve got a big ol' bag of potato chips don’t ask me why I’m talking about chips, it just seems right and you wanna know what everyone thinks about the flavor. Fider’s the place where people can give feedback on stuff like that, but, you know, for real stuff like websites and apps.

So, you go there, type up your thoughts, and boom other people can vote on whether they agree or disagree, like they're choosing between watching Family Guy or some other dumb show. But here’s the kicker, Fider lets people actually vote and prioritize the feedback they care about!

So, yeah, that’s Fider. Kinda like a big opinion poll but with a lot more techy stuff behind it. And you can totally use it for, you know, not chips...

You can checkout the Fider's official page for more details: https://fider.io/

This guide focuses on deploying Fider on AWS, but you can also self-host it on other platforms like Heroku, Azure, or your own servers. For more details, check out the official Fider self-hosting documentation.

Fider's Tech Stack

Backend: Go (Golang), with custom HTTP handling.
Frontend: React, using SCSS(Sassy CSS) for styling.
Database: PostgreSQL.
Authentication: OAuth 2.0.
Background Jobs: Go routines.
Cloud Storage and Email Integration: S3 for blob storage and SES for email notifications.

Pre-flight checks

Before starting, ensure that you have an AWS account with access to the free tier or a valid payment method on file.

Install AWS CLI

Step 1:

You need the awscli package installed for pushing the images built locally to Elastic Container Registry(ECR).

Step 2:

If you use ubuntu distro, Check the snap version by using the following command.

  snap version

Run the following snap install command for the AWS CLI.

  sudo snap install aws-cli --classic

After the installation is complete, verify the installation by using the command,

  aws --version

With Snap package you always get the latest version of AWS CLI as snap packages automatically refresh.

For command line installer check out the documentation: https://docs.aws.amazon.com/cli/latest/userguide/getting-started-install.html

Step 4: Configure AWS CLI

Use this command to configure the AWS CLI.

  aws configure

Once configured, verify your AWS CLI credentials using the following command.

  aws configure list

Step 5: Pull official docker images from DockerHub

Pull the official docker image for Fider from the DockerHub registry at: https://hub.docker.com/r/getfider/fider/tags

Use the docker pull command to pull the image from the DockerHub.

  docker pull getfider/fider:main

Check if the docker image is available in your instance.

  docker images

Push the Docker images to ECR

Step 1: Visit the ECR console

Create a public or private repository based on your use case by clicking on the create repository prompt in the top right corner.

Name the repo getfider/fider

Check if the repository is created.

Step 2: View the push commands

View the push commands by clicking on the respective repository.

Use the AWS CLI to retrieve an authentication token and authenticate your Docker client to your registry.

tag your pulled image so you can push the image to this repository.

Run the following command to push this image to your newly created AWS repository.

The docker images should now be available in the repository you created.

Create a VPC

Step 1: Create a standalone VPC for deploying Fider application

Create VPC only for now, you can create public and private subnets manually and can attach these subnets to particular route tables.

A route table will also get created with route to local traffic.

Step 2: Create an Internet gateway

Create an IGW and attach it to the VPC you created.

The IGW is now attached to the VPC.

Associate the IGW with the route table by clicking on edit routes.

By clicking on Add route attach the the IGW to the route table.

The route table now has a route to the IGW, which means the subnets associated with this route table can have Internet access.

Step 3: Create Public Subnets

Create at least two public subnets, so that the Application Load Balancer can reside in it.

Navigate to the VPC Console, under the subnets section create two public subnets.

Decide the subnet CIDR range and provision it accordingly.

Associate the public subnets to the public route table where the IGW is also attached.

Step 4: Create Private Subnets

Create at least two private subnets, so that your ECS tasks and RDS can reside in it.

Navigate to the VPC Console, under the subnets section create at least two private subnets.

Under the actions dropdown choose Edit subnet settings and uncheck the Enable auto-assign public IPv4 address.

Step 5: Create a Private route table

Create a private route table in the VPC console under the route tables section to route traffic locally.

Associate the private subnets to the private route table, like how you did with the public route table, and follow the same steps.

Step 6: Create an Elastic-ip

Allocate an elastic-ip address and attach it with the NAT gateway.

Click on Allocate Elastic IP address

From public ipv4 address pool, choose Amazon's pool of ipv4 address and click Allocate

You should now be able to check the static Elastic IP allocated to you. It can now be attached to the NAT gateway.

Step 7: Create NAT gateway

Create a NAT gateway to allow tasks in the private subnet to connect to the internet.

Navigate to NAT gateway section under the VPC console and create a NAT gateway.

Associate it with a public subnet and also associate an Elastic IP address to it.

Set the connectivity type to public.

Attach the NAT gateway to the private route table that has three private subnets associated with it.

The private route table now has a route to the NAT gateway.

Your VPC lineage should now look like this.

Create Subnet group for RDS

Navigate to the RDS sevice console and click on subnet groups,

Click on create DB subnet group

Fill out the subnet group details and add subnets and choose the availability zones(azs) accordingly(private subnets).

If you choose to launch multi-AZ DB clusters, you must select 3 subnets in three different availability zones.

Click on create,your subnet group will now be available in the subnet groups dashboard.

Create an RDS PostreSQL DB

Navigate to the RDS service console from the management console and click on DB instance, to create an RDS DB instance.

Click on create database

Choose standard create and choose PostgreSQL as engine type.

Choose the latest PostgreSQL engine version and for this guide i choose to go with single DB instance under the availability and durability section.

Under credential settings provide postgres as master user name and choose self managed to use your custom password for authenticating with RDS.

For this guide, I chose to go with Burstable classes, depending on your use case you can fine tune the DB instance class. Storage will be left as default, unless you choose to customise it.

Under connectivity section, choose the VPC that you created(fider-vpc) and the subnet group you created previously would appear under the DB subnet group section.

Choose NO under public access section.

Now, before proceeding to VPC security group section, navigate to VPC console and create a security group for both RDS and ECS.

The above sg rules are for ECS service.

The above sg rule is for RDS and it has no outbound rules set.

Now back to the RDS console where we left,

Choose the security group you created and choose the availability zone where you would want to host the DB instance.

Set database authentication as password based, I have disabled monitoring you can enable it if you want.

Create the initial database as fider in the DB instance.

I chose to go with the above settings, you can enable it if you want.

Click on create database to provision an RDS PostgrSQL DB.

The RDS PostgreSQL DB is now available!

Create Parameters in Systems manager Parameter Store

Since Fider currently does not appear to have built-in logic specifically to fetch credentials or secrets from AWS Secrets Manager, I have used parameter store to store Fider's environment variables so the tasks in ECS can fetch these credentials when they are being fired up.

Fider supports the above environment variables. I have used the standard tier and secureString type to encrypt sensitive data using KMS keys.

Create an IAM user

Fider supports the use of Amazon S3 for blob storage and Amazon SES for sending emails. However, simply passing the access keys through environment variables will not automatically make the images appear in the S3 bucket, nor will it enable SES for email sending.

The application will assume the role assigned to the user, which will then allow it to perform actions on the user's behalf.

Navigate to the IAM console and create a user and assign the roles necessary for accessing the S3 bucket and SES service,

I've granted full access to S3 and SES for this guide, but this isn't the recommended approach. Instead, you should create fine-grained access policies and assign them to the role.

Create an S3 bucket

Navigate to the S3 bucket console and create a general purpose S3 bucket for blob storage.

Once you've created the bucket, navigate to the Permissions tab and check the Bucket Policy section. It’s likely empty, right? Well, it shouldn’t be, if left empty the S3 bucket will remain empty forever.

  {
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Principal": {
                "AWS": "arn:aws:iam::<ACCOUNT_ID>:user/<USER_NAME>"
            },
            "Action": "s3:*",
            "Resource": "arn:aws:s3:::<BUCKET_NAME>"
        }
    ]
}

This policy is too permissive; you can create more fine-grained policies to restrict access.

Create an identity in SES

The application requires email verification as part of the signup process to prevent spam and ensure only valid users access the platform. I chose the SES API route instead of SMTP.

Navigate to the SES service console and start creating an identity.

I chose to go with domain verification. Once you've created the identity via the domain, make sure to validate the records with your DNS provider.

You can send a test mail from the verified identities to the check the workflow.

If you plan to run this setup in production, you can request production access from Amazon support. Otherwise, you can remain in the sandbox environment.

Why use ECS instead of EC2

There is one major reason to use ECS instead of EC2,

1. Reduced Operational Overhead:

EC2 is cheaper on paper, but that requires careful capacity management and container resource planning plus reserved instance commitments or spot fleet configurations. In the end it feels like 99% of the time, the admin overhead doesn't outweigh the savings.

Depends on how much control you want. If you want less control, go with Fargate. And yes, if you go with standard ECS, the EC2 instances are run and managed by you.

Create an ECS cluster

Step 1:

Navigate to the ECS console and create an ECS cluster with Fargate launch type.

Click on Create cluster.

I have turned on container insights for this guide, to see aggregated metrics at cluster and service level. In this way you can run the deep dive analysis.

This will create an ECS cluster via Cloudformation in the backend.

Create Task Definition

Step 1:

Navigate to the task definition section under the ECS console and create a task definition named fider-app-task-definition.

The task definition can be created via both the console and JSON.

Refer to the following JSON task definition for creating the task definition via JSON.

  {
    "taskDefinitionArn": "arn:aws:ecs:us-east-2:624184658995:task-definition/fider-app-task-definition:7",
    "containerDefinitions": [
        {
            "name": "fider-app",
            "image": "624184658995.dkr.ecr.us-east-2.amazonaws.com/getfider/fider:main",
            "cpu": 2048,
            "memory": 8192,
            "memoryReservation": 4096,
            "portMappings": [
                {
                    "name": "fider-app-3000-tcp",
                    "containerPort": 3000,
                    "hostPort": 3000,
                    "protocol": "tcp",
                    "appProtocol": "http"
                }
            ],
            "essential": true,
            "environment": [],
            "mountPoints": [],
            "volumesFrom": [],
            "secrets": [
                {
                    "name": "BASE_URL",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/BASE_URL"
                },
                {
                    "name": "DATABASE_URL",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/DATABASE_URL"
                },
                {
                    "name": "JWT_SECRET",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/JWT_SECRET"
                },
                {
                    "name": "GO_ENV",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/GO_ENV"
                },
                {
                    "name": "EMAIL_NOREPLY",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/EMAIL_NOREPLY"
                },
                {
                    "name": "EMAIL_AWSSES_ACCESS_KEY_ID",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/EMAIL_AWSSES_ACCESS_KEY_ID"
                },
                {
                    "name": "EMAIL_AWSSES_SECRET_ACCESS_KEY",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/EMAIL_AWSSES_SECRET_ACCESS_KEY"
                },
                {
                    "name": "EMAIL_AWSSES_REGION",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/EMAIL_AWSSES_REGION"
                },
                {
                    "name": "BLOB_STORAGE",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/BLOB_STORAGE"
                },
                {
                    "name": "BLOB_STORAGE_S3_ACCESS_KEY_ID",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/BLOB_STORAGE_S3_ACCESS_KEY_ID"
                },
                {
                    "name": "BLOB_STORAGE_S3_BUCKET",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/BLOB_STORAGE_S3_BUCKET"
                },
                {
                    "name": "BLOB_STORAGE_S3_ENDPOINT_URL",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/BLOB_STORAGE_S3_ENDPOINT_URL"
                },
                {
                    "name": "BLOB_STORAGE_S3_REGION",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/BLOB_STORAGE_S3_REGION"
                },
                {
                    "name": "BLOB_STORAGE_S3_SECRET_ACCESS_KEY",
                    "valueFrom": "arn:aws:ssm:us-east-2:624184658995:parameter/BLOB_STORAGE_S3_SECRET_ACCESS_KEY"
                }
            ],
            "logConfiguration": {
                "logDriver": "awslogs",
                "options": {
                    "awslogs-group": "/ecs/fider-app-task",
                    "mode": "non-blocking",
                    "awslogs-create-group": "true",
                    "max-buffer-size": "25m",
                    "awslogs-region": "us-east-2",
                    "awslogs-stream-prefix": "ecs"
                },
                "secretOptions": []
            },
            "systemControls": []
        }
    ],
    "family": "fider-app-task-definition",
    "taskRoleArn": "arn:aws:iam::624184658995:role/ecsTaskExecutionRole",
    "executionRoleArn": "arn:aws:iam::624184658995:role/ecsTaskExecutionRole",
    "networkMode": "awsvpc",
    "revision": 7,
    "volumes": [],
    "status": "ACTIVE",
    "requiresAttributes": [
        {
            "name": "ecs.capability.execution-role-awslogs"
        },
        {
            "name": "com.amazonaws.ecs.capability.ecr-auth"
        },
        {
            "name": "com.amazonaws.ecs.capability.docker-remote-api.1.28"
        },
        {
            "name": "com.amazonaws.ecs.capability.docker-remote-api.1.21"
        },
        {
            "name": "com.amazonaws.ecs.capability.task-iam-role"
        },
        {
            "name": "ecs.capability.execution-role-ecr-pull"
        },
        {
            "name": "ecs.capability.secrets.ssm.environment-variables"
        },
        {
            "name": "com.amazonaws.ecs.capability.docker-remote-api.1.18"
        },
        {
            "name": "ecs.capability.task-eni"
        },
        {
            "name": "com.amazonaws.ecs.capability.docker-remote-api.1.29"
        },
        {
            "name": "com.amazonaws.ecs.capability.logging-driver.awslogs"
        },
        {
            "name": "com.amazonaws.ecs.capability.docker-remote-api.1.19"
        }
    ],
    "placementConstraints": [],
    "compatibilities": [
        "EC2",
        "FARGATE"
    ],
    "requiresCompatibilities": [
        "FARGATE"
    ],
    "cpu": "2048",
    "memory": "8192",
    "runtimePlatform": {
        "cpuArchitecture": "X86_64",
        "operatingSystemFamily": "LINUX"
    },
    "registeredAt": "2025-02-18T06:06:55.123Z",
    "registeredBy": "arn:aws:sts::624184658995:assumed-role/AWSReservedSSO_AdministratorAccess_263ab8d7ae88c1c4/wasim",
    "tags": []
}

I have referenced the environment variables created in systems manager parameter store by their resource arn.

You also need the following policies to be attached to ecsTaskExecutionRole along with AmazonECSTaskExecutionRolePolicy,

AmazonSSMFullAccess
ROSAKMSProviderPolicy

The policies attached provides the ECS agent the required permissions to fetch the KMS encrypted environment variables from the systems manager parameter store, decrypt it and provide it to the application hosted inside the ECS tasks.

I have provided full access permission policies, but it's usually not the right way to do it. You can craft the fine grained access policies and attach the same to the ecsTaskExecutionRole.

Create an ECS service

Step 1:

Navigate to the ECS cluster you created and click on it, once inside the cluster click on service.

Step 2: Configuration parameters

In the compute configuration choose the capacity provider strategy and choose FARGATE as capacity provider with base 1 and weight as 100

Choose service as the application type in the deployment configuration and Replica as the service type with Desired tasks set to 1. You can also set it 2, I have used one for this guide.

In the networking configuration choose the VPC you created.

Turn off the public IP address as the tasks will use NAT gateway.

Configure the Application load balancer and place it inside the public subnet as front-facing for your application.

Create a target group on the fly in the ECS console itself and set up an HTTPS listener for your ALB. For this deployment, I’ve chosen to handle the HTTP to HTTPS redirect at the CDN level, so you don’t need an HTTP listener for your ALB, as redirects won’t happen at that level.

For the SSL certificate to appear in the dropdown, skip to step 3 from here and request a public certificate from the ACM console.

The health check path for Fider is _/health

I have not enabled service auto scaling for this guide, you can enable it if you want.

Click on create in the end and navigate to Cloudformation to view the status.

The password authentication with RDS PostgreSQL is successful and the records got migrated!

Since we have also enabled container insights, we can check the performance metrics of clusters, services and tasks.

Note: If you are not able to run the tasks or if you ran into errors like this,

It's because you placed the task in the public subnet with no public IP assigned to it, when you created the application load balancer in the ECS service console itself. Even i ran into this issue and then i had to troubleshoot the failed tasks.

I used the Systems Manager's AWSSupport-TroubleshootECSTaskFailedToStart automation document. It diagnosed the issue for me.

Note: Create the application load balancer and target group separately, note the availability zones(AZs) you assinged for ALB, and when you spin up the ECS tasks via service make sure you provide the private subnets from the same availability zones(AZs).

Feel free to check this documentation when you run into issues!

Step 3: Use ACM for SSL certificates

The service is up and running now.

Navigate to the ACM console and click on request a certificate.

Request a public certificate and click on Next.

Provide your domain name and choose the DNS validation method.

Click on request and ACM provides you with the DNS records, verify the records by mapping them to your DNS registrar settings.

Your SSL certificate is now ready to be used with the Application Load Balancer. You can map it under the Listener configuration for the ALB in the certificate section of the ECS service console.

Since I’m setting up CloudFront, I’ve also chosen to enable WAF at the CDN level for this deployment, as CloudFront acts as the first point of contact and can filter traffic before it reaches the ALB.

Since WAF is enabled at the CloudFront level and all traffic originates from there, you don’t need to enable WAF at the ALB level. Instead, I’ve added an X-Custom-Header to restrict access to the Application Load Balancer (ALB), ensuring that only requests from CloudFront are allowed.

Create Cloudfront distribution with WAF

Navigate to the integrations tab under the application load balancer

Click on manage cloudfront + WAF integration.

click on the check box to enable the cloudfront IPs as inbound to your application load balancer.

create an SSL/TLS certificate for your domain, from which you would like to host Fider on. Cloudfront only accepts SSL/TLS certificates from us-east-1(N.virginia) region.

Once you click create, WAF rules and Cloudfront distribution will get enabled for your ALB.

Now,navigate to the Cloudfront distribution and click on the behaviours tab to edit the settings of default cache behaviour,

I chose to handle the HTTP to HTTPS redirect at this level.

Since Fider is a dynamic application, Features like submitting feedback, voting, and commenting are processed in real-time, and the application relies on a database to store and manage user data dynamically. So set the Allowed HTTP methods to GET,HEAD,OPTIONS,PUT,POST,PATCH,DELETE . If not you won't be able to perform PUT operation if you choose from first two.

Save it and now hover over to the origins tab and choose to edit the ALB origin settings.

Set the protocol to HTTPS only, I choose to let Cloudfront talk to my ALB over HTTPS and scroll further down,

Add the Custom header which you added in the previous step with ALB, Cloudfront passes this header in every request it makes to the ALB. Hitting the ALB DNS without this header would result in 403 forbidden error.

This is how your ALB resource map should look. It will only forward traffic to the target group if the string you configured in CloudFront is present in the request header.

Next, map the distribution domain to the domain where you want to host Fider in your DNS records(CNAME record). Don't hit the distribution domain name it will only result in 502 error.

Hit the domain name you configured in the DNS entry, You should be able to see the landing page.

The images you upload to the webiste will get stored as an object in the S3 bucket, and when you access it from the browser via UI. you should be able to see this in the network tab under response headers (X-cache: Hit from Cloudfront)

You can navigate to the AWS WAF in the console and look for the request metrics.

You can also configure the managed rules in the WAF console under the Rules tab.

Viola! There you have it. Fider's up and running with the tasks securely inside their own private subnet, and the ALB routing requests to them. Additionally, with CloudFront and AWS WAF configured, traffic is efficiently delivered via the CDN while being protected at the edge, ensuring secure and optimized access.

DEV Community: Mohamed Wasim

Deploy Fider as a Private App on AWS with CloudFront VPC Origin

1. What is CloudFront VPC Origin

2. Why you should deploy Fider as a private application

3. Create an Internal Application Load Balancer

4. Create CloudFront VPC origin

5. Create a CloudFront Distribution

6. Create Secrets manager secret and Lambda function

7. Create a Custom Header rule in AWS WAF

8. Spin up ECS and RDS

9. Monitor CloudFront and WAF logs

10. Do You Really Need Zero Trust for Every App

11. Is CloudFront Truly Zero Trust Compliant

12. Deciding on Enterprise Grade Deployment for Fider

13. Caveats to consider

14. Final thoughts

Deploying Fider on AWS ECS: A Step-by-Step Guide to Deploy a Feedback Platform

AWS services used

Table of Contents

What is Fider ?

Fider's Tech Stack

Pre-flight checks

Install AWS CLI

Step 1:

Step 2:

Step 4: Configure AWS CLI

Step 5: Pull official docker images from DockerHub

Push the Docker images to ECR

Step 1: Visit the ECR console

Step 2: View the push commands

Create a VPC

Step 1: Create a standalone VPC for deploying Fider application

Step 2: Create an Internet gateway

Step 3: Create Public Subnets

Step 4: Create Private Subnets

Step 5: Create a Private route table

Step 6: Create an Elastic-ip

Step 7: Create NAT gateway

Create Subnet group for RDS

Create an RDS PostreSQL DB

Create Parameters in Systems manager Parameter Store

Create an IAM user

Create an S3 bucket

Create an identity in SES

Why use ECS instead of EC2

1. Reduced Operational Overhead:

Create an ECS cluster

Step 1:

Create Task Definition

Step 1:

Create an ECS service

Step 1:

Step 2: Configuration parameters

Step 3: Use ACM for SSL certificates

Create Cloudfront distribution with WAF