hcp-ansible/docs/architecture_guide.md.md

# Architecture Guide

Hashistack-Ansible allows you to deploy a number of architecture, wether you want to deploy a dev, testing, or production environment. These different architectures are described in this section.

## Dev deployment

If you only want to deploy a test environment, you can simply add a simgle host to each service that you want to deploy.

```ini
[haproxy_servers]

[vault_servers]
test01

[consul_servers]
test01

[nomad_servers]
test01
```

In this example, you will end end with each service running on a single host, with no clustering, and no redundancy. This setup *IS NOT RECOMMENDED** for anything but testing purposes, as it provides zero resiliency, and will break if anything goes down.

For this setup, the only requirement is for the target host to have a network interface that you can ssh into from the deployment host.

The architecture would like something like this:

```mermaid
graph LR;
  client[Client] -->|http| server{
    Vault Server
    Consul Server
    Nomad Server
  };
```

## Testing/Preprod deployment

For testing, of pre-production deployments, running all services on the same nodes might be a good way to cut cost and/or save resources.

## Production deployment

For production use, it is recommended to separate concerns as much as possible. This means that consul, vault and nomad, as well as the haproxy services, should be on different nodes altogether. The **client-facing** and **cluster-facing** interfaces should also be separated.

Ideally, you would need:
  - an odd number (3 to 5) of consul servers
  - an odd number (3 to 5) of vault servers
  - an odd number (3 to 5) of nomad servers
  - multiple (2 to 3) haproxy servers

The **nomad**, **vault** and **consul** servers should have **two network interfaces**, and one of them should be reachable from the haproxy nodes.

The architecture for this infrastructure would look like:

```mermaid
graph TD
  client[Client] -->|https :443| keepalived
  keepalived[VIP] --> haproxy1[HAProxy] & haproxy2[HAProxy]
  subgraph frontends
  direction LR
    haproxy1[HAProxy]
    haproxy2[HAProxy]
  end

  haproxy1[HAProxy] & haproxy2[HAProxy] -->|http :8500| consul

  subgraph consul
  direction LR
    consul1[Consul 01] <--> consul2[Consul 02] & consul3[Consul 03] & consul4[Consul 04] & consul5[Consul 05]
    consul2[Consul 02] <--> consul3[Consul 03] & consul4[Consul 04] & consul5[Consul 05]
    consul3[Consul 03] <--> consul4[Consul 04] & consul5[Consul 05]
    consul4[Consul 04] <--> consul5[Consul 05]

  end

  haproxy1[HAProxy] & haproxy2[HAProxy] -->|http :8200| vault

  subgraph vault
  direction LR
    vault1[Vault 01] <--> vault2[Vault 02]
    vault2[Vault 02] <--> vault3[Vault 03]
    vault3[Vault 03] <--> vault1[Vault 01]
  end

  vault -->|Service registration| consul
  haproxy1[HAProxy] & haproxy2[HAProxy] -->|http :4646| nomad

  subgraph nomad
  direction LR
    nomad1[Nomad 01] <--> nomad2[Nomad 02]
    nomad2[Nomad 02] <--> nomad3[Nomad 03]
    nomad3[Nomad 03] <--> nomad1[Nomad 01]
  end

  nomad -->|Service registration| consul
```
> **Note**: you can substract the haproxy part if using an external load-balancing solution, like AWS ALB,or any other LB technology, for connecting to your platform.