Skip to content
Work
InfrastructurePublic

Infrastructure and mail systems

Infrastructure & DevOpsCybersecurityCloud Providers

Hosting, containers, mail, isolation, backup, monitoring

Infrastructure and reliability engineering

Context

Mail and sites were accumulated hosts. Failure was discovered by a user. Backup was a file on the same disk. Migration meant hope.

System boundary

In scope

  • web hosting
  • mail (Mailcow)
  • Docker deployments
  • backup
  • restore tests
  • monitoring
  • migration

Out of scope

building the hosted applications themselves

Architecture

Edge (Cloudflare/Nginx) in front of application containers and mail. Backups leave the machine. Monitoring is a separate channel from the thing it watches.

Production infrastructure for hosting and email: containers, dedicated servers, isolation between tenants, backup, migration, and monitoring.
  1. Cloudflare
  2. Nginx
  3. App containers
  4. Mailcow
  5. Storage
  6. Off-site backup
  7. Monitoring
  • Edge → compute → store
  • Monitoring is off the host
Published sequence
  1. Cloudflare
  2. Nginx
  3. App containers
  4. Storage
  5. Off-site backup
  6. Mailcow
  7. Monitoring
Roles in the diagram
  • external1
  • service2
  • store2
  • interface2

Counts from the published record. They describe structure, not a measured result.

What is watched away from the host?

edge

  • externalCloudflare
  • interfaceNginx

runtime

  • serviceApp containers
  • serviceMailcow

persistence

  • storeStorage

resilience

  • storeOff-site backup

observation

  • interfaceMonitoring
  1. CloudflareNginx
  2. NginxApp containers
  3. NginxMailcow
  4. App containersStorage
  5. MailcowStorage
  6. StorageOff-site backup
  7. NginxMonitoring
  8. MailcowMonitoring

Critical flows

  1. DNS/edge

  2. reverse proxy

  3. app or mail container

  4. storage. Backup jobs copy off-box. Alerts fire from health checks, not from the app’s own uptime page alone.

Stages

  • Tenant isolation
  • Mail
  • Containers
  • Off-site backup
  • Monitoring

Decisions

01

Mail is a system

It gets isolation, backup, and monitoring like anything else that can take a company down.

02

Backups that have never been restored are not backups

Restore tests are scheduled work, not folklore.

03

Monitoring off the host

A box cannot be the only witness to its own death.

When it fails

ConditionDetectionResponse
Disk fillFilesystem alertsShed logs; block non-essential writes
Mail queue explosionQueue depthRate-limit outbound; inspect relay
Failed restore testScheduled restore jobPage; treat backup as broken
Certificate or DNS lapseExpiry monitorsRenew path; status page

Infrastructure

  • Dedicated servers
  • Docker
  • Nginx
  • Mailcow
  • off-site backup
  • Cloudflare
  • uptime and log monitoring

Outcome

Hosting and mail run as operated systems: isolated, backed up, watched, and movable.

Current state

In production. Migrations and capacity work follow the same topology.

  • Docker
  • Nginx
  • Mailcow
  • Cloudflare
  • Linux
  • Backup
  • Monitoring
Contact