Infrastructure we've
actually shipped.

The Challenge

• CentOS 7 reached end-of-life — no security patches, ticking clock on compliance
• Single MTA node handling all volume — one failure meant total outage
• Ansible roles were brittle, undocumented, and partially broken
• Deliverability reputation at risk due to infrastructure instability
• No monitoring — problems discovered when users complained
• Manual intervention required for most operational tasks

What We Did

• Migrated entire stack from CentOS 7 → Ubuntu 22.04 LTS
• Refactored all Ansible roles — idempotent, documented, tested
• Designed stateless MTA node architecture for horizontal scaling
• Deployed HAProxy for TCP-layer load balancing across MTA nodes
• Integrated Prometheus + Grafana for real-time metrics
• Wired up Graylog for centralized log aggregation and alerting
• Built automated node provisioning — full node in ~2 hours

Architecture

Client Type

• High-volume transactional email sender (anonymized)
• SaaS platform with critical email delivery requirements
• Operated at scale — 50,000+ messages per day sustained
• Global recipient base, deliverability-sensitive