Permalink
Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
140 lines (107 sloc) 10.6 KB

Overview

This repository contains examples illustrating how Terraform, Apache Brooklyn, and Cloudsoft AMP work, similarities and differences, so people can evaluate when to use each.

The blueprints deploy a simple Node.js “TODO” application. Items are added and removed to a list via a web interface. Items are persisted to a MySQL backend. In the Brooklyn and AMP versions, the Node.js servers autoscale with demand, and traffic is automatically load-balanced between instances.

Please refer to the documentation within each implementation for deployment directions:

Deployment

For clarity, all implementations use shell-based provisioning, and for brevity minimal capabilities for each component are configured. For example, the MySQL database is not general-purpose or highly reusable; it is built for the purposes of this demonstration. (Better MySQL blueprints are available in both the Brooklyn/AMP and the Terraform communities).

It is straightforward in either Brooklyn/AMP or Terraform to write fuller blueprints which are generalized and encapsulated for reuse as building blocks.

Terraform

For Terraform we deploy 3 components:

  • A single node MySQL/MariaDB database, configured with a database/table appropriate for the “TODO” application
  • A cluster of Node.JS servers in an AWS-specific Auto-Scaling Group
  • An AWS Elastic Load Balancer: we chose to use ELB as it makes it simple for us to leverage an AWS autoscaling group (this is the standard pattern when using AWS and TF)

Each component is placed on the same subnet. The database and Node instances use a security group that allows ingress from all addresses on the subnet. The ELB component uses a security group that allows ingress on port 80 from 0.0.0.0/0.

Apache Brooklyn

The Brooklyn blueprint defines the entities or components that are used in the Brooklyn and AMP deployments.

  • A single node MySQL/MariaDB database, configured with a database/table appropriate for the “TODO” application
  • A cluster of Node.JS servers that provide the application logic, persisting to the database. Connectivity to the database is determined dynamically
  • An Nginx-based load balancer. AMP can also configure platform and cloud-specific load balancers, and F5 boxes (as done for the Weather Company project); however Nginx is a quick and easy way to achieve immediate portability

The nodejs.bom and mariadb.bom are generic blueprints for those two components, with no special application logic for this use case and no special assumptions for AWS or a specific cloud. The brooklyn-app.bom blueprint passes config to those elements to set up the “TODO” application.

Cloudsoft AMP

The blueprint deployed and managed by Cloudsoft AMP reuses the entities defined in the Apache Brooklyn blueprint.

The AMP blueprint deploys in the same way as the Apache Brooklyn Blueprint, with the addition that the blueprint also configures resources for “Application Network Security”: Network segregation and accessibility is determined through assigning named network entities within the blueprint, and the security controls are applied using the mechanisms available on the target cloud or platform. Application Network Security is currently supported on AWS and OpenStack.

In this case, only the load balancer will be exposed on the public internet, and other resources will be isolated.

In-life management

Terraform

For Terraform, an autoscaling group was configured to scale Node.JS based on CPU utilization. Terraform does not offer facilities that automate failure detection and node/service replacement.

Apache Brooklyn

In Apache Brooklyn, in-life management is accomplished primarily by applying “policies” to entities within the blueprint. Policies observe certain conditions of the application, the infrastructure on which it runs, or external dependencies, make decisions about actions to take based on those observations, and then take the appropriate action.

For this application, we’ve placed four standard policies on the Node.JS entity or cluster of entities.

  1. There is a service restarter on the Node.JS instance. Should the Node.JS process fail, it is automatically restarted
  2. There is a failure detector on the JodeJS instance. This emits an “Entity Failed” event whenever a failure is detected, and similarly an “Entity Recovered” event when recovered.
  3. On the Node.JS cluster there is a service replacer, which will replace a failed node with a newly provisioned one.
  4. Finally, also on the Node.JS cluster there is an auto-scaler policy. Auto-scaling, and all policies, in AMP are extremely flexible, and may be triggered on any observable condition, individually or taken in aggregate. In this case we are scaling the Node.JS cluster on a metric provided by the Node.JS application itself [metric, and link to line]. In other cases we may choose to scale on CPU usage, requests per second, etc.

Cloudsoft AMP

In-life management in the AMP deployment of this application is identical to that mentioned above for Apache Brooklyn.

Conclusion

This repository shows a simple Brooklyn/AMP-managed application side-by-side with the analagous Terraform-deployed application. The following table briefly summarizes other Brooklyn/AMP capabilities as they relate to Terraform.

Capability Apache Brooklyn/Cloudsoft AMP Terraform
Provision Secure 3-tier Autoscaling App in AWS YES YES
Deploy elswhere: Softlayer, OpenStack, Azure, Kubernetes YES - The same blueprint works in all these clouds PARTLY - New blueprints need to be written for each environment. Many items (ELB, auto-scaling) are specific to particular clouds.
Live view of deployed resoures YES PARTLY - Infrastructure Only. TF will show resources created by its providers but does not model processes or systems and it does not integrate with monitoring tools to show status
Custom Policies YES - This is the main objective of autonomic management NO. Not in Scope
Support for Cloud Service (DNS, S3, RedShift, etc) PARTLY - Blueprints can be written for any cloud service, with some available (e.g. AWS S3, R53, ELB) but the main focus is on portability, not on every native offering. PARTLY, provision only - Excellent coverage for provisioning native services, but does not interact subsequently; relies on cloud-provider supplied integrations, e.g. AWS ELB + auto-scaling group
Integrations (Salt, Chef, Bash, Docker Images, REST calls, logstash, monitoring, ServiceNOW) YES - Extensive integrations with authoring, monitoring, and management tools — you can even use Terraform PARTLY - Custom provisioners can be written but there are not very many.
Community Size Small (enterprise-focused) Medium
Foundation-run Community Yes - Apache Sofware Foundation No - HashiCorp
Deploys Upgrades Partly - improvements in progress Yes