1. Reproducible server - kill it all and regenerate. Only things lost are user accounts and user-specific settings.
2. Like Terraform: manages VPC, subnets, elasticIP (unused for now), route tables, security groups, etc.
3. Updates are transactional and allows for rollback.
4. Fail2ban to quiet the noise of the internet
5. sslh to allow 443 be used for SSH, HTTP, SSL, TINC, etc.
6. Tinc VPN is a mesh-capable VPN.
7. Mosh for fast ssh
8. Can support larger networks of machines.

1. Using a non-released and bleeding-edge version of nixops. This can be relaxed soon and we can clean up the boilerplate.
2. Collaborating in a team involves sharing the statefile. This can be done with GPG sharing, but is clunky. S3 support soon.
3. Supports Virtual Box, NixOS, AWS, GCE, Azure, Hetzner, Digital Ocean, libvirtd

These commands require three things:

1. nix or run make nix in the top of this repo
2. access to /server/<NAME>.nixops: this is a statefile which manages deployments
3. access to /server/secrets directory (recommend to use git-crypt and the git-crypt.attributes file)

Both statefile and secrets are encrypted. Decrypt by having an admin add you to the git-crypt repo. You must have a GPG key. This is a rough overview of how to do this:

nix-env -iA nixpkgs.gnupg
gpg2 --full-generate-key
gpg2 -k  # this will show you your keys, find the id of your pub key
gpg2 --send-key <YOUR-PUB-KEY-ID>
# Contact an Admin who will add your key to the git repo, then in the repo:
git-crypt unlock

For a better GPG walkthrough, see this page or GitHub's tutorial.

Ensure you have your authtoken in ./secrets/DO_token.txt or wherever configuration.nix points to. Pick a NAME such that the following will create NAME.nixops file in the current directory to track the state of this deployment:

./manage NAME create '<configuration.nix>'

or

./manage NAME create ./configuration.nix

Note: the former is harder to understand, but will be filesystem-location-independent. (You can move this folder around and everything will still work, portable). Then:

./manage NAME deploy

This will create the server, infect it with NixOS, deploy the configuration, activate services, etc.

At the moment, this will create a single machine, but can be easily extend such that this deployment manages multiple machines to whatever infrastructure desired. The following will SSH into your newly deployed box.

./manage NAME ssh machine

./manage NAME deploy

./manage NAME destroy
./manage NAME delete

nix-env -iA nixpkgs.tinc_pre

SSH into the server, adjust the /var/run/tinc.NAME*/ file permissions and create an invitation. As root on the server:

./manage live ssh server
chown tinc.name /var/run/tinc.name*
tinc -n name invite <YOUR-COMPUTERS-NAME>
systemctl restart tinc.name

The output of the last command is used to join the network. You must also establish an IP. Avahi is the best. On your own computer:

tinc -n name join <output-of-invite-command>
echo "avahi-autoipd name -D" >> /etc/tinc/name/tinc-up
tinc -n name start

Then a cheap SOCKSv5 proxy for browsing can be made using the Avahi local link hostname:

ssh -D 8080 username@your-server -N

git-crypt add-gpg-user USERID