Lifecycle, persistence and pricing

What survives a restart, what does not, the v1 limits, and what each state costs

What survives a restart

Stop is a clean shutdown. Start is a fresh boot of the same machine. Across that cycle:

Survives: everything on disk. Installed packages, cloned repositories, virtualenvs, configuration, your data, and the machine's ssh host identity. The Computer also keeps the same IP address for its whole life.

Does not survive: running processes. A fresh boot starts with a clean process table, exactly like rebooting any Linux machine.

If something should come back after a restart, give it a systemd unit or an @reboot cron entry inside the Computer:

# /etc/systemd/system/myapp.service
[Unit]
Description=My app
After=network.target
 
[Service]
ExecStart=/root/venv/bin/python /root/app.py
Restart=on-failure
 
[Install]
WantedBy=multi-user.target

systemctl enable --now myapp

v1 limits, stated plainly

• Host-bound. A Computer's disk lives on one physical host. Loss of that host means loss of the Computer. Keep anything irreplaceable backed up externally.
• No checkpoint or restore. Stop and start is a shutdown and a fresh boot. There is no snapshot of running state.
• No migration. A Computer does not move between hosts.
• Disk is fixed at create. Size it for the workload up front.

Pricing

A Computer bills for what it holds. Two states, one rule: a running Computer bills runtime and disk, a stopped Computer bills parked disk and nothing else.

Worked example: a 2 vCPU, 2 GB, 20 GB Computer running around the clock costs about $153 per month. The same machine used two hours a day costs about $15 per month. Parked, it costs $2.70 per month.

That last number is the design: park a Computer when you are not using it and it becomes a $2.70-per-month reservation that wakes in seconds with everything intact.