Files
clan-core/clanServices/sshd/README.md

91 lines
3.2 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Clan service: sshd
What it does
- Generates and persists SSH host keys via `vars`.
- Optionally issues CAsigned host certificates for servers.
- Installs the `server` CA public key into `clients` `known_hosts` for TOFUless verification.
When to use it
- ZeroTOFU SSH for dynamic fleets: admins/CI can connect to frequently rebuilt hosts (e.g., server-1.example.com) without prompts or perhost `known_hosts` churn.
Roles
- Server: runs sshd, presents a CAsigned host certificate for `<machine>.<domain>`.
- Client: trusts the CA for the given domains to verify servers certificates.
Tip: assign both roles to a machine if it should both present a cert and verify others.
Quick start (with host certificates)
Useful if you never want to get a prompt about trusting the ssh fingerprint.
```nix
{
inventory.instances = {
sshd-with-certs = {
module = { name = "sshd"; input = "clan-core"; };
# Servers present certificates for <machine>.example.com
roles.server.tags.all = { };
roles.server.settings = {
certificate.searchDomains = [ "example.com" ];
# Optional: also add RSA host keys
# hostKeys.rsa.enable = true;
};
# Clients trust the CA for *.example.com
roles.client.tags.all = { };
roles.client.settings = {
certificate.searchDomains = [ "example.com" ];
};
};
};
}
```
Basic: only add persistent host keys (ed25519), no certificates
Useful if you want to get an ssh "trust this server" prompt once and then never again.
```nix
{
inventory.instances = {
sshd-basic = {
module = {
name = "sshd";
input = "clan-core";
};
roles.server.tags.all = { };
};
};
}
```
Example: selective trust per environment
Admins should trust only production; CI should trust prod and staging. Servers are reachable under both domains.
```nix
{
inventory.instances = {
sshd-env-scoped = {
module = { name = "sshd"; input = "clan-core"; };
# Servers present certs for both prod and staging FQDNs
roles.server.tags.all = { };
roles.server.settings = {
certificate.searchDomains = [ "prod.example.com" "staging.example.com" ];
};
# Admin laptop: trust prod only
roles.client.machines."admin-laptop".settings = {
certificate.searchDomains = [ "prod.example.com" ];
};
# CI runner: trust prod and staging
roles.client.machines."ci-runner-1".settings = {
certificate.searchDomains = [ "prod.example.com" "staging.example.com" ];
};
};
};
}
```
- Admin -> server1.prod.example.com: zeroTOFU (verified via cert).
- Admin -> server1.staging.example.com: falls back to TOFU (or is blocked by policy).
- CI -> either prod or staging: zeroTOFU for both.
Note: server and client searchDomains dont have to be identical; they only need to overlap for the hostnames you actually use.
Notes
- Connect using a name that matches a cert principal (e.g., `server1.example.com`); wildcards are not allowed inside the certificate.
- CA private key stays in `vars` (not deployed); only the CA public key is distributed.
- Logins still require your user SSH keys on the server (passwords are disabled).