enroll/README.md

Enroll

enroll inspects a Linux machine (currently Debian-only) and generates Ansible roles for things it finds running on the machine.

It aims to be optimistic and noninteractive:

• Detects packages that have been installed
• Detects Debian package ownership of /etc files using dpkg's local database.
• Captures config that has changed from packaged defaults (dpkg conffile hashes + package md5sums when available).
• Also captures service-relevant custom/unowned files under /etc/<service>/... (e.g. drop-in config includes).
• Defensively excludes likely secrets (path denylist + content sniff + size caps).
• Captures non-system users that exist on the system, and their SSH public keys
• Captures miscellaneous /etc files that it can't attribute to a package, and installs it in an etc_custom role
• Avoids trying to start systemd services that were detected as being Inactive during harvest

---

Two modes: single-site vs multi-site (--fqdn)

enroll has two distinct ways to generate Ansible:

1) Single-site mode (default: no --fqdn)

Use this when you're enrolling one server (or you're generating a "golden" role set you intend to reuse).

What you get

• Config, templates, and defaults are primarily contained inside each role.
• Raw config files (when not templated) live in the role's files/.
• Template variables (when templated) live in the role's defaults/main.yml.

Pros

• Roles are more self-contained and easier to understand.
• Better starting point for provisioning new servers, because the role contains most of what it needs.
• Less inventory abstraction/duplication.

Cons

• Less convenient for quickly enrolling multiple hosts with divergent configs (you'll do more manual work to make roles flexible across hosts).

2) Multi-site mode (--fqdn)

Use this when you want to enroll several existing servers quickly, especially if they differ.

What you get

• Roles are shared across hosts, but host-specific data lives in inventory.
• Host inventory drives what's managed:
  • which files to deploy for that host
  • which packages are relevant for that host
  • which services should be enabled/started for that host
• For non-templated config, raw files live in host-specific inventory under .files/ (per role).

Pros

• Fastest way to retrofit multiple servers into config management.
• Avoids shared-role "host A breaks host B" problems by keeping host-specific state in inventory.
• Better fit when you already have a fleet and want to capture/reflect reality first.

Cons

• More abstraction: roles become more "data-driven".
• Potential duplication: raw files may exist per-host in inventory (even if identical).
• Harder to use the roles to provision a brand-new server without also building an inventory for that new host, because multi-site output assumes the server already exists and is being retrofitted.

Rule of thumb

• If your goal is "make this one server reproducible / provisionable" → start with single-site.
• If your goal is "get several already-running servers under management quickly" → use multi-site.

---

Key concepts

Harvest

enroll begins by 'harvesting' known state about your host. This includes detecting what running services exist, what packages have been installed 'manually' (that is, stuff that doesn't come out of the box with the OS), and anything 'custom' in /etc that it can't attribute to a specific package.

It also detects if any config files have been changed from their packaged defaults. If they have, it will attempt to 'harvest' them. If the config file is identical to how it comes with the package, then it doesn't bother harvesting it, because there's little value in config-managing it if it's identical to what you get by simply installing the package!

The harvest writes a state.json file explaining all the data it harvested and, if it chose not to harvest something, explanations as to why that is the case (see below: sensitive data).

Remote harvesting (workstation → remote)

If you'd prefer not to install enroll on the target host, you can run the harvest over SSH from your workstation and pull the harvest bundle back locally:

enroll harvest --remote-host myhost.example.com --remote-user myuser --out /tmp/enroll-harvest

• --remote-port defaults to 22
• --remote-user defaults to your local $USER

This uploads a self-contained enroll zipapp to a temporary directory on the remote host, runs harvest there, then downloads the resulting harvest bundle to the --out directory on your workstation.

Privilege note: A "full" harvest typically needs root access. Remote harvesting assumes the remote user can run sudo without a password prompt (NOPASSWD) so the harvest can run non-interactively. If you don't want this, pass --no-sudo as well.

JinjaTurtle note: If you want to take advantage of JinjaTurtle to turn configs into templates (see below note on JinjaTurtle integration), you'll still need to install JinjaTurtle on the remote host first.

Sensitive data

enroll doesn't make any assumptions about how you might handle sensitive data from your config files, in Ansible. Some people might use SOPS, others might use Vault, others might do something else entirely.

For this reason, enroll will attempt to read config files, and if it detects data that looks like a sensitive SSH/SSL private key, or password, or API key, etc, then it won't harvest it for config management.

This inevitably means that it will deliberately miss some important config files that you probably want to manage in Ansible.

Nonetheless, in the Harvest 'state' file, there should be an explanation of 'excluded files'. You can parse or inspect this file to find what it chose to ignore, and then you know what you might want to augment the results with later, once you 'manifest' the harvest into Ansible configuration.

Nonetheless, in some cases it may be appropriate to truly grab as much as you can, including secrets. For that, read on for the --dangerous flag.

Opting in to fetching sensitive data: --dangerous

WARNING: --dangerous disables enroll's "likely a secret" safety checks. This can cause private keys, TLS key material, API tokens, database passwords, and other credentials to be copied into your harvest output in plaintext.

Only use --dangerous if you explicitly want to scoop up sensitive files and you understand where the harvest output is stored, who can read it, and how it will be handled (backups, git commits, etc, as well as risk of using --out with a shared /tmp location where other users could see the data). We offer no liability if your sensitive data is compromised through the use of this tool!

Strong recommendation: If you plan to keep harvested files long-term (especially in git), encrypt secrets at rest. A common approach is to use SOPS and then use the community.sops Ansible collection to load/decrypt encrypted content during deploy.

Install the collection:

ansible-galaxy collection install community.sops

Then you can use the collection's lookup/vars plugins or modules to decrypt or load SOPS-encrypted vars at runtime.

Manifest

The 'manifest' subcommand expects to be given a path to the 'harvest' obtained in the first step. It will then attempt to generate Ansible roles and playbooks (and potentially 'inventory') from that harvest.

Manifesting is the most complex step because a lot of people will have opinions on how Ansible roles and inventory should work. No solution is perfect for everyone. However, enroll tries to strike a reasonable balance.

Remember, the purpose of this tool is to save time getting your systems into a decently-managed state. It's still up to you to wrangle it into a form that works for you on an ongoing basis.

---

Single-shot mode for the impatient sysadmin

enroll has a 'single-shot' subcommand which combines the two other phases (harvest and manifest) into one. Use it to generate both the harvest and then manifest ansible from that harvest all in one go. Perfect if you're in a hurry!

---

JinjaTurtle integration (both modes)

If you also have my other tool JinjaTurtle installed, enroll will attempt to create Jinja2 templates for any ini/json/xml/toml style configuration that it finds.

• Templates live in the role (roles/<role>/templates/...)
• Variables live in:
  • single-site: roles/<role>/defaults/main.yml
  • multi-site (--fqdn): inventory/host_vars/<fqdn>/<role>.yml

JinjaTurtle will be used automatically if it is detected on the $PATH. You can also be explicit and pass --jinjaturtle, but this will throw an error if JinjaTurtle is not on the $PATH.

If you do have JinjaTurtle installed, but don't wish to make use of it, you can use --no-jinjaturtle, in which case all config files will be kept as 'raw' files.

Remote mode: if you are using the --remote-xxx flags for manifest or single-shot subcommands, and want to take advantage of the JinjaTurtle integration, you'll still need to install JinjaTurtle on the remote host in advance.

---

How multi-site avoids "shared role breaks a host"

In multi-site mode, roles are data-driven. The role contains generic tasks like:

• "deploy all files listed for this host"
• "install packages listed for this host"
• "apply systemd enable/start state listed for this host"

The host inventory is what decides which files/packages/services apply to that host. This prevents the classic failure mode where host2 adds a config file to a shared role and host1 then fails trying to deploy a file it never had.

Raw non-templated files are stored under:

• inventory/host_vars/<fqdn>/<role>/.files/...

…and the host's role variables describe which of those files should be deployed.

---

Install

Ubuntu/Debian apt repository

sudo mkdir -p /usr/share/keyrings
curl -fsSL https://mig5.net/static/mig5.asc | sudo gpg --dearmor -o /usr/share/keyrings/mig5.gpg
echo "deb [arch=amd64 signed-by=/usr/share/keyrings/mig5.gpg] https://apt.mig5.net $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/mig5.list
sudo apt update
sudo apt install enroll

AppImage

Download the AppImage file from the Releases page (verify with GPG if you wish, my fingerprint is here), then make it executable and run it:

chmod +x Enroll.AppImage
./Enroll.AppImage

Pip/PipX

pip install enroll

Poetry

Clone this repository with git, then:

poetry install
poetry run enroll --help

---

Usage

1. Harvest state/information about the host

On the host (root recommended to harvest as much data as possible):

enroll harvest --out /tmp/enroll-harvest

Remote harvest over SSH (no enroll install required on the remote host)

enroll harvest --remote-host myhost.example.com --remote-user myuser --out /tmp/enroll-harvest

--dangerous (captures potentially sensitive files — read the warning above)

enroll harvest --out /tmp/enroll-harvest --dangerous

Remote + dangerous:

enroll harvest --remote-host myhost.example.com --remote-user myuser --out /tmp/enroll-harvest --dangerous

2. Generate Ansible manifests (roles/playbook) from that harvest

Single-site (default: no --fqdn)

Good for one server, or for producing roles you want to reuse to provision new machines:

enroll manifest --harvest /tmp/enroll-harvest --out /tmp/enroll-ansible

Multi-site (--fqdn)

Best when enrolling multiple already-running servers into one repo:

enroll manifest --harvest /tmp/enroll-harvest --out /tmp/enroll-ansible --fqdn "$(hostname -f)"

Single-shot

Alternatively, do both steps in one shot:

enroll single-shot --harvest /tmp/enroll-harvest --out /tmp/enroll-ansible --fqdn "$(hostname -f)"

Remote single-shot (run harvest over SSH, then manifest locally):

enroll single-shot --remote-host myhost.example.com --remote-user myuser --harvest /tmp/enroll-harvest --out /tmp/enroll-ansible --fqdn "myhost.example.com"

In multi-site mode (--fqdn), you can run single-shot repeatedly against multiple hosts while reusing the same --out directory so each host merges into the existing Ansible repo.

3. Run Ansible

Single-site

You can run it however you prefer (local connection or your own inventory). Example:

ansible-playbook -i "localhost," -c local /tmp/enroll-ansible/playbook.yml

Multi-site (--fqdn)

In multi-site mode, enroll generates an ansible.cfg, host_vars inventory, and a host-specific playbook:

ansible-playbook /tmp/enroll-ansible/playbooks/"$(hostname -f)".yml

---

Found a bug, have a suggestion?

My Forgejo doesn't yet support proper federation, and for that reason I've not opened up registration/login to use the issue queue.

Instead, you can e-mail me (see the pyproject.toml for details) or contact me on the Fediverse:

https://goto.mig5.net/@mig5