Eatzy — Two Data Centers, Zero Downtime
We migrated UK food ordering platform Eatzy, serving 1,000+ restaurants, from a single cPanel server to a dual-DC active/active cluster spanning two data centers in Amsterdam.
From cPanel to Dual-DC Active/Active Cluster
As Eatzy grew, a single server could no longer keep up. VeriTeknik transformed the platform into an infrastructure spanning two data centers with fully automatic failover and zero-downtime target.
| Before | After | |
|---|---|---|
| Hosting | Single cPanel server | 2 DC, 12+ servers, Docker Swarm |
| Database | Single MySQL | MariaDB 11.4 Galera (3-node, synchronous) |
| Failover | None | PowerDNS GSLB, 15s TTL, automatic |
| SSL | Manual | Let's Encrypt auto-SSL (Traefik + DNS challenge) |
| Deploy | FTP upload | Ansible AWX + Docker Registry |
| Monitoring | None | Prometheus + Grafana + ELK Stack |
| RPO | Unknown | < 1 second (Galera sync replication) |
| RTO | Hours+ (manual) | < 30 seconds (GSLB automatic) |
Technical Details
Core components and architectural decisions behind Eatzy's infrastructure.
Database: MariaDB Galera
3-node synchronous replication with near-zero data loss.
- RPO < 1 second (sync replication)
- Quorum voting for split-brain protection
- IST (Incremental State Transfer) for fast recovery
- At least 1 node per DC
Orchestration: Docker Swarm
Scalable architecture with per-restaurant isolated containers.
- 4 managers, 6 workers, 2 edge nodes
- Per-restaurant container isolation
- Overlay network for cross-DC communication
- Rolling updates for zero-downtime deploys
DNS/GSLB: PowerDNS + Lua
Galera-backed DNS with automatic failover.
- Lua health check scripts
- 15-second TTL for fast failover
- DNS record replication via Galera
- Per-restaurant subdomain management
Edge Routing: Traefik v3
Automatic SSL and dynamic service discovery.
- Let's Encrypt DNS-01 challenge auto-SSL
- Per-restaurant host-based routing
- Docker Swarm provider for auto-discovery
- Active edge node in each DC
Monitoring: Prometheus + ELK
End-to-end observability and alerting.
- Node, cAdvisor, MySQL, PHP-FPM exporters
- Grafana dashboards
- AlertManager → Slack notifications
- ELK Stack for centralized log management
EatzyControlDash
A custom management panel built by VeriTeknik specifically for Eatzy. Enables managing all infrastructure operations from a single interface. It is concrete proof of VeriTeknik's "Build & Operate" approach.
EatzyControlDash
Centralized Infrastructure Management Panel
Disaster Recovery
Automatic recovery mechanisms designed and tested for every scenario.
| Scenario | Impact | Recovery |
|---|---|---|
| 1 Galera node fails | None (2/3 quorum) | Automatic IST |
| 1 Swarm worker fails | Rescheduled to other workers | Automatic (< 2 min) |
| 1 Traefik fails | GSLB routes to other DC | Automatic (15s TTL) |
| All of MainCubes fails | Equinix takes over | Automatic GSLB |
| All of Equinix fails | MainCubes takes over | Automatic GSLB |