This page collects notes about cloud usage, Docker/Kubernetes, and how they fit with self-hosting.

From discussions, the main reasons to use cloud providers instead of only home servers:

• High availability: you need uptime that survives home power or ISP failures.
• Global latency: you have users spread across regions.
• Bandwidth: your home upload bandwidth is too low.

Examples where cloud can make sense:

• Public-facing APIs and websites.
• Latency-sensitive services (e.g. game backends).
• Off-site backups.

For purely personal or family tools (RSS, blog, small git hosting), many participants preferred self-hosting at home for privacy and cost reasons.

Key patterns around Docker:

• Use docker-compose to describe services:
  • One `docker-compose.yml` per application stack.
  • Declare networks, volumes, environment variables explicitly.

• Persist data in named volumes or host directories:
  • For databases: always mount `/var/lib/postgresql` (or equivalent) somewhere persistent.
  • Keep config files versioned in Git.

• Common services containerized:
  • Reverse proxies (nginx, Traefik).
  • Databases (PostgreSQL, MariaDB).
  • Queue systems (Redis, RabbitMQ).
  • Web apps (Node.js, Python, Go, Rust).

Kubernetes was mentioned a few times, usually with a note of caution:

• Powerful but heavy for small personal setups.
• High cognitive overhead:
  • Pods, deployments, services, ingress, secrets, persistent volumes, etc.

General consensus from the tone of the discussions:

• For small-scale self-hosting, Kubernetes is often overkill.
• Prefer:
  • Simple Docker-Compose stacks.
  • Orchestrators like Nomad if you truly need scheduling.
  • Or even plain systemd services for a few long-lived daemons.

When you do host services in the cloud:

• Infrastructure as Code (IaC):
  • Use Terraform, Ansible, or similar to capture VM and network configuration.
  • Avoid hand-tuning servers with undocumented changes.

• Backups and secrets:
  • Keep encrypted secrets (e.g. with Ansible Vault, sops).
  • Store backups in multiple regions if running in a single cloud vendor.

• Cost awareness:
  • Watch for creeping costs: log storage, managed DBs, data transfer.
  • Use basic monitoring (Grafana/Prometheus, netdata, provider dashboards) to avoid surprises.