Mini Cloud Architecture

High Availability TODO application with Vagrant, HAProxy, Node.js, and MariaDB Galera Cluster.

Architecture

graph TB
    subgraph "Host Machine"
        Browser["Browser<br/>:8080"]
        Stats["HAProxy Stats<br/>:8404"]
    end
    
    subgraph "Frontend Network - 192.168.50.0/24"
        LB["Load Balancer<br/>192.168.50.10<br/>HAProxy"]
        APP1["App Server 1<br/>192.168.50.11<br/>Node.js + garbd"]
        APP2["App Server 2<br/>192.168.50.12<br/>Node.js + garbd"]
    end
    
    subgraph "Backend Network - 192.168.60.0/24"
        VIP["Virtual IP<br/>192.168.60.30<br/>Keepalived"]
        DB1["Database 1<br/>192.168.60.21<br/>MariaDB Galera"]
        DB2["Database 2<br/>192.168.60.22<br/>MariaDB Galera"]
    end
    
    Browser --> LB
    Stats -.-> LB
    LB -->|Round Robin| APP1
    LB -->|Round Robin| APP2
    APP1 <-->|rsync<br/>File Sync| APP2
    APP1 --> VIP
    APP2 --> VIP
    VIP --> DB1
    VIP --> DB2
    DB1 <-->|Multi-Master<br/>Replication| DB2

Loading

Key Features:

• Load Balancing: HAProxy distributes traffic across 2 app servers
• Database HA: Galera multi-master cluster with VIP failover (Keepalived)
• File Sync: Real-time bidirectional rsync + inotify between app servers
• Arbitrators: 2 garbd instances on app servers provide quorum (4-node voting)
• Networks: Separated frontend (50.x), backend (60.x), and management (70.x)

VMs

VM	Role	RAM	Frontend IP	Backend IP
lb	HAProxy + UFW	512 MB	192.168.50.10	-
app1	Node.js + garbd + inotify-tools + UFW	512 MB	192.168.50.11	192.168.60.11
app2	Node.js + garbd + inotify-tools + UFW	512 MB	192.168.50.12	192.168.60.12
db1	MariaDB Galera + Keepalived + rsync + UFW	1024 MB	-	192.168.60.21
db2	MariaDB Galera + Keepalived + rsync + UFW	1024 MB	-	192.168.60.22

Quick Start

Prerequisites: VirtualBox 6.1+, Vagrant 2.2+, 4GB RAM, 10GB disk

Recommended Quick Start (recommended order to ensure Galera quorum)

# Start DB VMs
vagrant up db1 db2 --provision

# Start app servers
vagrant up app1 app2 --provision

# Start load balancer
vagrant up lb --provision

# NOTE: --provision is needed the first time.

Check status:

vagrant status

Access services from the host:

• TODO App: http://localhost:8080
• HAProxy Stats: http://localhost:8404/haproxy_stats (admin/pass)

Common tasks:

# SSH
vagrant ssh app1

# Stop all VMs
vagrant halt

# Destroy all VMs
vagrant destroy -f

Common Commands

Monitoring

# Service status
vagrant ssh app1 -c "sudo systemctl status todo-app"
vagrant ssh db1 -c "sudo systemctl status mariadb"

# Application logs
vagrant ssh app1 -c "tail -f /home/vagrant/app.log"

# Cluster size
vagrant ssh db1 -c "mysql -u root -e 'SHOW STATUS LIKE \"wsrep_cluster_size\";'"

# VIP location
vagrant ssh db1 -c "ip addr show eth1 | grep 192.168.60.30"

# File sync status
vagrant ssh app1 -c "sudo systemctl status upload-sync"
vagrant ssh app1 -c "sudo tail -f /var/log/upload-sync.log"

Database

# Access MySQL
vagrant ssh db1 -c "mysql -u root"

# Query todos
vagrant ssh db1 -c "mysql -u root -e 'SELECT * FROM todo_app.todos;'"

Application Updates

The ./app folder syncs automatically. After editing server.js:

vagrant ssh app1 -c "sudo systemctl restart todo-app"
vagrant ssh app2 -c "sudo systemctl restart todo-app"

High Availability Testing

App Failover:

vagrant ssh app1 -c "sudo systemctl stop todo-app"
curl http://localhost:8080  # Still works via app2
vagrant ssh app1 -c "sudo systemctl start todo-app"

Database Failover:

vagrant ssh db1 -c "sudo systemctl stop mariadb"
curl http://localhost:8080  # Still works via db2 (VIP fails over)
vagrant ssh db1 -c "sudo systemctl start mariadb"

File Sync:

# Upload image via web UI, then verify sync:
vagrant ssh app1 -c "ls /home/vagrant/app/public/uploads/"
vagrant ssh app2 -c "ls /home/vagrant/app/public/uploads/"