AWS Terraform ECS deployment
- blake.wilkinson
Introduction
Panintelligence is the easy-to-use and quick-to-deploy solution to unlocking powerful data insights. We're the only software solution on the market that brings together business intelligence and machine learning to a new self-service level, empowering you to gain full control over your data.
Access unprecedented styling options to make our software look just like your own, and hook up cloud data warehousing and ETL (Extract, Transform and Load) tools, as well as your core product, to create a truly seamless analytics experience.
Simply connect your databases to the Panintelligence dashboard. The Panintelligence dashboard only has read access so it won’t modify any of your data.
The Terraform scripts provide a template from which you can derrive your own bespoke deployment based on your individual needs.
Panintelligence does not collect your data or move it.
This project supports multi-availability zone deployment. For more details please take a look at System architecture and Multi-Availability zone deployment. In addition, the monitoring and logging will also explain in more detail.
Why this approach?
At Panintelligence, we believe in scripting as much as we possibility can and lean heavily on the community to supply us with open source tooling to facilitate this goal. Hashicorp has long been one of the biggest players in the devops scene and they Infrastructure as code tool “Terraform” offers an idempotent and platform agnostic method to deliver change across an estate. The software is light and portable. Since we already had a great many of the scripts written in terraform to deploy our software in an idempotent manner, it seemed appropriate to share these scripts with partners.
The architecture is driven chiefly by ECS within the deployment in AWS. ECS represents a free way to orchestrate and manage fleets of containers. Backed by fargate, it offers a low complexity method to create a compute layer. There are other container management options available such as kubernetes and docker swarm where we can assist with the installation of our software onto these platforms, but for AWS, nothing pleases a Yorkshireman more than the prospect of a freebie!
We encourage our partners to fork our code and change it for their own purposes, but we make one request. if you stumble across something that’s not quite right, please do raise a merge request with a fix for our base code as you would be helping out a great number of people, and we treasure our community.
Installation of Key Tools
Panintelligence
We host the container images for panintelligence at github container registry (GHCR). In order to access our images, please create a support request with our team at support@panintelligence.com providing your github username or email. The team will respond with a list of images to which you have been granted access.
It’s highly recommended and endorsed that you pull our images and push to your own private image repository so you can maintain service continuity without relying on a third party.
Terraform
Please review the documentation provided by Hashicorp for the most comprehensive and up to date documentation pertaining to installation of Terraform on your chosen platform https://developer.hashicorp.com/terraform/tutorials/aws-get-started/install-cli
AWS
You will need a target AWS account with the scope to provision resources defined in the service quotas explained later in this document.
System Architecture
Diagram
Resources
Resource | Description | How is it used? |
|---|---|---|
Route 53 | Amazon Route 53 is a highly available and scalable cloud Domain Name System (DNS) web service. For more information, please see Setting up Amazon Route 53 documentation | You can attach your domain name to the AWS Application Load Balancer to point to the Panintelligence dashboardAWS ACM |
AWS ACM | AWS Certificate Manager is a service that lets you easily provision, manage, and deploy public and private Secure Sockets Layer/Transport Layer Security (SSL/TLS) certificates for use with AWS services and your internal connected resources. For more information, please see Setting up - AWS Certificate Manager documentation | In order to use port 443/HTTPS in the AWS Application Load Balancer, you will need an SSL certificate. |
AWS S3 Bucket | Amazon Simple Storage Service (Amazon S3) is an object storage service that offers industry-leading scalability, data availability, security, and performance. For more information, please see AWS S3 Bucket documentation | The architecture requires the user to upload a lambda zip provided in the Git repository and another s3 bucket is created to store images, themes and excel-data. |
AWS Internet gateway | An internet gateway is a horizontally scaled, redundant, and highly available VPC component that allows communication between your VPC and the internet. For more information, please see Internet Gateways documentation | The Panintelligence dashboard requires web browser access. |
AWS IAM | AWS Identity and Access Management (IAM) enables you to manage access to AWS services and resources securely. Using IAM, you can create and manage AWS users and groups, and use permissions to allow and deny their access to AWS resources. For more information, please see AWS IAM documentation | IAM permissions allows you to have fine grain control on who and what has access to resources. |
AWS Security groups | A security group acts as a virtual firewall for your instance to control inbound and outbound traffic. For more information, please AWS Security Groups documentation | Increase protection to your infrastructure. |
AWS Application Load Balancer | Elastic Load Balancing automatically distributes your incoming traffic across multiple targets, such as EC2 instances, containers, and IP addresses, in one or more Availability Zones. It monitors the health of its registered targets, and routes traffic only to the healthy targets. For more information, please see AWS Application Load Balancer documentation | The ALB directs traffic to the healthy EC2 targets. |
AWS VPC | Amazon Virtual Private Cloud (Amazon VPC) enables you to launch AWS resources into a virtual network that you've defined. This virtual network closely resembles a traditional network that you'd operate in your own data centre, with the benefits of using the scalable infrastructure of AWS. For more information, please see AWS VPC documentation | We use the AWS VPC to launch resources in the virtual network. |
Subnets | You need to specify a logical address to specific resources. For more information, please see Subnets documentation | Configure resources to specific subnet cidr blocks. |
NACL | A network access control list (ACL) is an optional layer of security for your VPC that acts as a firewall for controlling traffic in and out of one or more subnets. For more information, please see NACL documentation | Configure additional security. |
AWS Lambda | AWS Lambda is a serverless compute service that lets you run code without provisioning or managing servers, creating workload-aware cluster scaling logic, maintaining event integrations, or managing runtimes. For more information, please see AWS Lambda documentation | The infrastructure uses AWS Lambda to side load S3 objects to AWS EFS |
AWS RDS MariaDB | Amazon Relational Database Service (Amazon RDS) is a web service that makes it easier to set up, operate, and scale a relational database in the AWS Cloud. For more information, please see AWS RDS MariaDB documentation | The Panintelligence dashboard uses AWS RDS MariaDB as an external DB. |
AWS EFS | Amazon Elastic File System (Amazon EFS) provides a simple, serverless, set-and-forget elastic file system for use with AWS Cloud services and on-premises resources. For more information, please see AWS EFS documentation | AWS EFS is used to keep persistent data for themes, images, SVG and custom jdbc |
AWS Auto scaling | AWS Auto Scaling monitors your applications and automatically adjusts capacity to maintain steady, predictable performance at the lowest possible cost. For more information, please see AWS Auto scaling documentation | Auto scaling is used to increase or decrease the EC2 instances depending on traffic. |
AWS NAT gateway | A NAT gateway is a Network Address Translation (NAT) service. You can use a NAT gateway so that instances in a private subnet can connect to services outside your VPC but external services cannot initiate a connection with those instances. Please see AWS NAT gateway documentation | Allows you to use Panintelligence Automated Lincence Manager |
AWS ECS | Amazon Elastic Container Service (ECS) is a fully managed container orchestration service that simplifies running, scaling, and securing Docker container applications on AWS. It integrates seamlessly with other AWS services. | manages the containers in the microservice modelA |
AWS Cloudwatch | Amazon CloudWatch is a monitoring and observability service designed for DevOps engineers, developers, and IT managers. It provides real-time monitoring, logging, and alarms for AWS resources, applications, and services, ensuring operational health and performance optimization. | stores logs and container metrics and insights for monitoring, logging and scaling purposes. |
Service Quotas
AWS Resource | Service quotas that the infrastructure uses | Notes |
|---|---|---|
AWS ECS (cluster) | 1 |
|
AWS ECS (task) | 5 - 8 | Scaling policies will dictate how many tasks are in operation |
AWS ECS (Service) | 5 |
|
Target groups per Auto Scaling group | 2 |
|
AWS Auto scaling per region | 2 |
|
AWS VPC | 1 | VPC configuration is conducted outside the main terraform script |
AWS Internet gateway | 1 |
|
General Purpose SSD (gp2) volume storage | 200 |
|
AWS Application load balancer | 1 |
|
AWS Lambda Function time out | 15 |
|
AWS Lambda temporary storage |
| AWS Default quota value is 512MB, A Panintelligence theme or image would not hit the limit. |
EFS per VPC | 1 |
|
EFS Mount targets | 2 |
|
EFS attached security group | 1 |
|
Interface VPC endpoints per VPC | 8 |
|
Route tables per VPC | n/a |
|
Routes per route table | 10 |
|
Subnets per VPC | 6 |
|
Security groups | 10 |
|
Network ACLs per VPC | 10 |
|
Secrets per account | 1 |
|
DB instance | 1 |
|
DB Parameter group | 1 |
|
DB subnet group | 1 |
|
DB security group | 1 |
|
AWS NAT Gateway | 1 |
|
Technical Data sheet requirements
The technical datasheet offers some guidance on how much resources you would require depending on your infrastructure and users.
Please take a look at this link.
Sizing and recommendations
Service | CPU | Memory | Storage | Notes |
|---|---|---|---|---|
Dashboard | 1 | 2Gb | 100Gb |
|
renderer | 2 | 4gb |
| The more resource you give, the faster resources render. This is only for ad-hoc request or reports. |
pirana | 0.5 | 512mb |
|
|
scheduler | 0.5 | 512mb |
|
|
|
|
|
|
|
Storage requirements
Database storage
EFS storage
it’s recommended to leave around 1Gb storage for images, themes, “SVG” files, custom JDBC and locale information. This will flex based on your own requirements and is dependent on the size of the images you use. Since Pi is a web-based system, it’s recommended that you compress images to the smallest storage requirements while maintaining image quality to facilitate transfer speeds.
Security
Panintelligence
Public Access
The dashboard container is accessible through port 443 via the load balancer.
Private Access
There is no direct access expressly added to the repository database. access to the database can be conducted via adminer. If
AWS RDS MariaDb
AWS EFS
EFS is encrypted at rest.
AWS IAM
The terraform scripts follow least possible permissions where possible.
AWS Lambda
There is one lambda function that configures the EFS. This lambda function only has access to the EFS volume and MUST only be executed once. repeated execution of this lamdba function will overwrite resources in the EFS, eliminating all changes.
Multi AZ deployment
Auto-scaling deployment
The dashboard autoscaling is set initially for a min and max capacity of 1. This should be set higher should you wish to enable autoscaling. the policy is set to scale based on a CPU exhaustion.
because the dashboard container runs using tomcat, you should not set the autoscaling rules to use memory as this would cause the dashboard to prematurely scale out, and never back in.
Persistent Storage on AWS EFS
Database Respository (metadata) on AWS RDS
Model | Core Count | vCPU* | CPU Credits/hour | Memory (GiB) | Network Performance (Gbps) |
|---|---|---|---|---|---|
db.t3.micro | 1 | 2 | 12 | 1 | Up to 5 |
db.t3.small | 1 | 2 | 24 | 2 | Up to 5 |
Panintelligence
Obtaining a licence key
To obtain your licence, please contact your CSM at panintelligence. If you’re unaware who your CSM is, please contact the support desk support@panintelligence.com and your request will be appropriately routed.
Automated Licence Manager
those looking to authenticate their product using the ALM must place the customer key against the PI_LICENCE environment variable on the dashboard service. This service must have the capability to talk out to the internet. when the dashboard first starts, the process will reach out to our system and obtain the licence pertaining to your customer ID. In the event of an unsuccessful initial callback, you will receive a invalid licence notification.
Offline Licencing
Customers who are unable to provision a system that has egress access to the internet from the application layer may be delivered an offline licence. This licence contains everything required to fully licence panintelligence. This licence is non transferrable. A new licence for each deployment should be sought from your CSM. The licence should be stripped of all carriage returns and line spacing and placed inline in the PI_LICENCE environment variable. Please note that these licences are time limited and thus represents and operational overhead to ensure the licences are kept up to date. Also these licences are of fixed scope. If you require more users or additional modules, you will contact your CSM who can create you a new licence.
Key Skills
Terraform
AWS VPC (security groups, networking, subnets, nacls, route tables)
AWS Lambda
AWS ECS
Containerisation
MySQL
AWS EFS
AWS Cloudwatch
AWS Load Balancing (Application Load Balancing)
AWS IAM
AWS Route53
AWS Secrets Manager
SSL/TLS Certificates
Horizontal Scaling
Linux
Key Run Information
Terraform
Although the terraform scripts are self documenting, it helps to become familiar with the HCL notation that hashicorp uses when provisioning systems with terraform. You can find excellent documentation for your first steps into terraform by following the link: https://developer.hashicorp.com/terraform/intro
Networking and security groups
The panintelligence deployment detailed in the terraform scripts makes use of VPC, subnets, security groups and other AWS assets in order to build out a scalable deployment of panintelligence. The code is reuseable and parameterisable. You’re encouraged to fork from our project. We also welcome contributions from our community for improvements to our scripts.
Backup and Restore
Database Disaster Recovery
by default, the database retention period is 3 days. Should you need to restore from a backup, you should follow the AWS backup restore process https://docs.aws.amazon.com/aws-backup/latest/devguide/restoring-rds.html
Please note that this will restore to an entire new database. you will need to connect to this database using an administration tool, migrate your backed up database, then issue a restoration.
one approach to this is to launch an interactive container or ec2 instance with mutual access to both databases. install a mariadb client, then follow the following steps to extract and import your database
backup extract (source database)
mysqldump \
--add-drop-table \
--add-drop-database \
--databases \
-u<admin_user> \
-p<your password> \
-h<database source host endpoint> \
-P3306 \
dashboard \
--ignore-table=dashboard.mis_user_cat_access_view_pi \
--ignore-table=dashboard.test_user_access \
--ignore-table=dashboard.mis_user_cat_access_view \
> sqldump_dashboard<todays_date>.sqlbackup restore (target database)
mysql \
-u<db_user> \
-p<your password> \
-h<database target host endpoint> \
-P3306 < sqldump_dashboard<todays_date>.sqlAWS EFS Disaster Recovery
EFS is backed up using EFS backup vault. please follow the AWS instructions to effect a restore
https://docs.aws.amazon.com/aws-backup/latest/devguide/restoring-efs.html
Time To Deploy / Restore
The script takes approximately 20 minutes to run.
Restoring a database using the delta backup routine from aws takes approximately 20 minutes with additional time required to move relevant extracts around. If you have access set up, the whole process should take no longer than 30 minutes.
Healthchecks, Logging, Troubleshooting
How to check if your Panintelligence dashboard is healthy? How can you troubleshoot if there are issues?
Metrics to look out for:
You want to see if there is any “unhealthy hosts”, if there is any, that’s a sign there is a problem. Look into that particular EC2 instance to find out.
Target groups:
Log into AWS Management console → Go into EC2 console → On the left side, click into the ‘Target Groups’ tab
Click into the Panintelligence Target:
The image below shows where to check for any unhealthy targets
If you do see any unhealthy targets, you can check the ‘Monitoring’ tab to look at the metrics for more details at a specific time
AWS Cloudwatch
Accessing your instance for administration
Because of the nature of containerisation, changes to the container should not be performed interactively or directly. You can set most settings using environment variables for which there is a comprehensive list available here.
You are welcome to make changes to the base container by means of a Dockerfile if you wish to add additional software or packages. Please note that changes made to the base container will fall outside of support. having said that, we’re only too happy to help if you have found an issue. You can obtain assistance by reaching out to our support team at support@panintelligence.com
AWS RDS MariaDB to check if the Panintelligence Repo is healthy:
AWS RDS MariaDB: To view the logs, please follow these instructions linked here:
Log into AWS Management console → Go into AWS RDS console → Click into ‘Databases’ → Click into ‘dashboard’:
You can check the ‘Monitoring’ tab. This is to see if you are hitting max connections, the CPU Utilisation and memory space :
You can even go further and check the RDS MariaDB Logs, select the logs near the bottom and select ‘View’ as highlighted below:
Cost Estimates
All costs are in USD unless otherwise stated.
All Services are deployed to the default region of EU-WEST-1 unless otherwise stated.
Description | Service | Monthly Cost (USD) | Configuration Summary |
|---|---|---|---|
Logging / Monitoring | Amazon CloudWatch | 3.4472 | GetMetricData: Number of metrics requested (20), Standard Logs: Data Ingested (3 GB), Logs Delivered to CloudWatch Logs: Data Ingested (3 GB), Logs Delivered to S3: Data Ingested (0 GB) |
adminer | AWS Fargate | 18.03 | Operating system (Linux), CPU Architecture (x86), Average duration (1 days), Number of tasks or pods (1 per day), Amount of ephemeral storage allocated for Amazon ECS (20 GB), Amount of memory allocated (1 GB) |
renderer | AWS Fargate | 36.04 | Operating system (Linux), CPU Architecture (x86), Average duration (1 days), Number of tasks or pods (1 per day), Amount of ephemeral storage allocated for Amazon ECS (20 GB), Amount of memory allocated (2 GB) |
dashboard | AWS Fargate | 36.04 | Operating system (Linux), CPU Architecture (x86), Average duration (1 days), Number of tasks or pods (1 per day), Amount of ephemeral storage allocated for Amazon ECS (20 GB), Amount of memory allocated (2 GB) |
scheduler | AWS Fargate | 36.04 | Operating system (Linux), CPU Architecture (x86), Average duration (1 days), Number of tasks or pods (1 per day), Amount of ephemeral storage allocated for Amazon ECS (20 GB), Amount of memory allocated (2 GB) |
pipredict | AWS Fargate | 36.04 | Operating system (Linux), CPU Architecture (x86), Average duration (1 days), Number of tasks or pods (1 per day), Amount of ephemeral storage allocated for Amazon ECS (20 GB), Amount of memory allocated (2 GB) |
persistent storage | Amazon Elastic File System (EFS) | 0.17 | Desired Storage Capacity (0.5 GB per month), Infrequent Access requests (0 GB per month) |
Ingress | Application Load Balancer | 18.52 | Number of Application Load Balancers (1) |
Egress | Network Address Translation (NAT) Gateway | 35.09 | Number of NAT Gateways (1) |
| Public IPv4 Address | 18.25 | Number of In-use public IPv4 addresses (5), Number of Idle public IPv4 addresses (0) |
Repo Database | Amazon RDS for MariaDB | 112.08 | Storage amount (20 GB), Nodes (1), Instance type (db.t3.medium), Utilization (On-Demand only) (100 %Utilized/Month), Deployment selection (Multi-AZ), Pricing strategy (OnDemand), Storage volume (General Purpose SSD (gp2)), Additional backup storage (20 GB) |
docker credentials secret | AWS Secrets Manager | 0.4 | Number of secrets (1), Average duration of each secret (30 days), Number of API calls (1 per month) |
Cost Summary
Estimated Montly hosting cost (without licence) $350.15
Cost over 12 Months $4201.80
Please be informed that certain services (e.g., adminer, a database administration tool) and modules that are not utilized in accordance with the terms of your Panintelligence license do not require deployment. This situation presents an opportunity for cost savings.
How to Obtain Support assistance
Our team are on hand during UK office hours and can be raised by emailing support@panintelligence.com
If you require assistance with Panintelligence to help with visualisations, it may be advantageous for you to book time with one of our consultants. Please reach out to your CSM who will schedule this with you. They can be raised at csm@panintelligence.com
CUSTOMER NEWS - Our August 24 Release Is Now Available - Download It Now!