mirror of
https://github.com/enpaul/vault2vault.git
synced 2025-12-17 17:00:38 +00:00
Compare commits
12 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
| 226c717684 | |||
| b55af77051 | |||
| 4550a73404 | |||
| bdb62993a2 | |||
| 3f6f5cf7e0 | |||
| 2f75180623 | |||
| c729414b03 | |||
| ba6b71687e | |||
| d61d2cb1a1 | |||
| c7c2a87ebb | |||
| 8e9df58f43 | |||
| 9943dd112c |
20
CHANGELOG.md
20
CHANGELOG.md
@@ -2,6 +2,26 @@
|
||||
|
||||
See also: [Github Release Page](https://github.com/enpaul/vault2vault/releases).
|
||||
|
||||
## Version 0.1.2
|
||||
|
||||
View this release on: [Github](https://github.com/enpaul/vault2vault/releases/tag/0.1.2),
|
||||
[PyPI](https://pypi.org/project/vault2vault/0.1.2/)
|
||||
|
||||
- Add user documentation
|
||||
- Add project road map
|
||||
- Fix incorrect and missing docstrings for internal functions
|
||||
|
||||
## Version 0.1.1
|
||||
|
||||
View this release on: [Github](https://github.com/enpaul/vault2vault/releases/tag/0.1.1),
|
||||
[PyPI](https://pypi.org/project/vault2vault/0.1.1/)
|
||||
|
||||
- Fix bug causing stack trace when the same vaulted block appears in a YAML file more than
|
||||
once
|
||||
- Fix bug where the `--ignore-undecryptable` option was not respected for vaulted variables
|
||||
in YAML files
|
||||
- Update logging messages and levels to improve verbose output
|
||||
|
||||
## Version 0.1.0
|
||||
|
||||
View this release on: [Github](https://github.com/enpaul/vault2vault/releases/tag/0.1.0),
|
||||
|
||||
141
README.md
141
README.md
@@ -10,27 +10,30 @@ but works recursively on encrypted files and in-line variables
|
||||
[](https://www.python.org)
|
||||
[](https://github.com/psf/black)
|
||||
|
||||
⚠️ **This project is alpha software and is under active development** ⚠️
|
||||
⚠️ **This project is beta software and is under active development** ⚠️
|
||||
|
||||
- [What is this?](#what-is-this)
|
||||
- [Installing](#installing)
|
||||
- [Using](#using)
|
||||
- [Usage](#usage)
|
||||
- [Recovering from a failed migration](#recovering-from-a-failed-migration)
|
||||
- [Roadmap](#roadmap)
|
||||
- [Developing](#developer-documentation)
|
||||
|
||||
## What is this?
|
||||
|
||||
If you use [Ansible Vault](https://docs.ansible.com/ansible/latest/user_guide/vault.html)
|
||||
then you may have encountered the problem of needing to role your vault password. Maybe
|
||||
then you may have encountered the problem of needing to roll your vault password. Maybe
|
||||
you found it written down on a sticky note, maybe a coworker who knows it left the
|
||||
company, maybe you accidentally typed it into Slack when you thought the focus was on your
|
||||
terminal. Whatever, these things happen.
|
||||
|
||||
The builtin tool Ansible provides,
|
||||
The built-in tool Ansible provides,
|
||||
[`ansible-vault rekey`](https://docs.ansible.com/ansible/latest/cli/ansible-vault.html#rekey),
|
||||
works suffers from two main drawbacks: first, it only works on vault encrypted files and
|
||||
not on vault encrypted YAML data; and second, it only works on a single vault encrypted
|
||||
file at a time. To rekey everything in a large project you'd need to write a script that
|
||||
goes through every file and rekeys everything in every format it can find.
|
||||
suffers from two main drawbacks: first, it only works on vault encrypted files and not on
|
||||
vault encrypted YAML data; and second, it only works on a single vault encrypted file at
|
||||
a time. To rekey everything in a large project you'd need to write a script that
|
||||
recursively goes through every file and rekeys every encrypted file and YAML variable all
|
||||
at once.
|
||||
|
||||
This is that script.
|
||||
|
||||
@@ -55,42 +58,104 @@ install `vault2vault` using [PipX](https://pypa.github.io/pipx/) and the `ansibl
|
||||
pipx install vault2vault[ansible]
|
||||
```
|
||||
|
||||
**Note: vault2vault requires an Ansible installation to function. If you are installing to a standalone virtual environment (like with PipX) then you must install it with the `ansible` extra to ensure a version of Ansible is available to the application.**
|
||||
> Note: vault2vault requires an Ansible installation to function. If you are installing to a
|
||||
> standalone virtual environment (like with PipX) then you must install it with the
|
||||
> `ansible` extra to ensure a version of Ansible is available to the application.\*\*
|
||||
|
||||
## Using
|
||||
## Usage
|
||||
|
||||
These docs are pretty sparse, largely because this project is still under active design
|
||||
and redevelopment. Here are the command line options:
|
||||
> Note: the full command reference is available by running `vault2vault --help`
|
||||
|
||||
```
|
||||
> vault2vault --help
|
||||
usage: vault2vault [-h] [--version] [--interactive] [-v] [-b] [-i VAULT_ID] [--ignore-undecryptable] [--old-pass-file OLD_PASS_FILE]
|
||||
[--new-pass-file NEW_PASS_FILE]
|
||||
[paths ...]
|
||||
Vault2Vault works with files in any arbitrary directory structures, so there is no need to
|
||||
have your Ansible project(s) structured in a specific way for the tool to work. The
|
||||
simplest usage of Vault2Vault is by passing the path to your Ansible project directory to
|
||||
the command:
|
||||
|
||||
Recursively rekey ansible-vault encrypted files and in-line variables
|
||||
|
||||
positional arguments:
|
||||
paths Paths to search for Ansible Vault encrypted content
|
||||
|
||||
options:
|
||||
-h, --help show this help message and exit
|
||||
--version Show program version and exit
|
||||
--interactive Step through files and variables interactively, prompting for confirmation before making each change
|
||||
-v, --verbose Increase verbosity; can be repeated
|
||||
-b, --backup Write a backup of every file to be modified, suffixed with '.bak'
|
||||
-i VAULT_ID, --vault-id VAULT_ID
|
||||
Limit rekeying to encrypted secrets with the specified Vault ID
|
||||
--ignore-undecryptable
|
||||
Ignore any file or variable that is not decryptable with the provided vault secret instead of raising an error
|
||||
--old-pass-file OLD_PASS_FILE
|
||||
Path to a file with the old vault password to decrypt secrets with
|
||||
--new-pass-file NEW_PASS_FILE
|
||||
Path to a file with the new vault password to rekey secrets with
|
||||
```bash
|
||||
vault2vault ./my-ansible-project/
|
||||
```
|
||||
|
||||
Please report any bugs or issues you encounter on
|
||||
[Github](https://github.com/enpaul/vault2vault/issues).
|
||||
The tool will prompt for the current vault password and the new vault password and then
|
||||
process every file under the provided path. You can also specify multiple paths and
|
||||
they'll all be processed together:
|
||||
|
||||
```bash
|
||||
vault2vault \
|
||||
./my-ansible-project/playbooks/ \
|
||||
./my-ansible-project/host_vars/ \
|
||||
./my-ansible-project/group_vars/
|
||||
```
|
||||
|
||||
To skip the interactive password prompts you can put the password in a file and have the
|
||||
tool read it in at runtime. The `--old-pass-file` and `--new-pass-file` parameters work
|
||||
the same way as the `--vault-password-file` option from the `ansible` command:
|
||||
|
||||
```bash
|
||||
vault2vault ./my-ansible-project/ \
|
||||
--old-pass-file=./oldpass.txt \
|
||||
--new-pass-file=./newpass.txt
|
||||
```
|
||||
|
||||
If you use multiple vault passwords in your project and want to roll them you'll need to
|
||||
run `vault2vault` once for each password you want to change. By default, `vault2vault`
|
||||
will fail with an error if it encounters vaulted data that it cannot decrypt with the
|
||||
provided current vault password. To change this behavior and instead just ignore any
|
||||
vaulted data that can't be decrypted (like, for example, if you have data encrypted with
|
||||
multiple vault passwords) you can pass the `--ignore-undecryptable` flag to turn the
|
||||
errors into warnings.
|
||||
|
||||
> Please report any bugs or issues you encounter on
|
||||
> [Github](https://github.com/enpaul/vault2vault/issues).
|
||||
|
||||
### Recovering from a failed migration
|
||||
|
||||
This tool is still pretty early in it's development, and to be honest it hooks into
|
||||
Ansible's functionality in some fragile ways. I've tested as best I can to ensure it
|
||||
covers as many edge cases as possible, but there is still the chance that you might get
|
||||
partway through a password migration and then have the tool fail out, leaving half of your
|
||||
data successfully rekeyed and the other half not.
|
||||
|
||||
In the spirit of the
|
||||
[Unix philosophy](https://hackaday.com/2018/09/10/doing-one-thing-well-the-unix-philosophy/)
|
||||
this tool does not include any built-in way to recover from this state. However, it can
|
||||
be done very effectively using a version control tool.
|
||||
|
||||
If you are using Git to track your project files then you can use the command
|
||||
`git reset --hard` to restore all files to the state of the currently checked out commit.
|
||||
This does have the side effect of erasing any other un-committed work in the repository,
|
||||
so it's recommended to always have a clean working tree when using Vault2Vault.
|
||||
|
||||
If you are not using a version control system to track your project files then you can
|
||||
create a temporary Git repository to use in the event of a migration failure:
|
||||
|
||||
```bash
|
||||
cd my-project/
|
||||
|
||||
# Initialize the new repository
|
||||
git init
|
||||
|
||||
# Add and commit all your existing files to the git tree
|
||||
git add .
|
||||
git commit -m "initial commit"
|
||||
|
||||
# Run vault migrations
|
||||
vault2vault ...
|
||||
|
||||
# If no recovery is necessary, delete the git repository data
|
||||
rm -rf .git
|
||||
```
|
||||
|
||||
## Roadmap
|
||||
|
||||
This project is considered feature complete as of the
|
||||
[0.1.1](https://github.com/enpaul/vault2vault/releases/tag/0.1.1) release. As a result the
|
||||
roadmap focuses on stability and user experience ahead of a 1.0 release.
|
||||
|
||||
- [ ] Reimplement core vaulted data processing function to enable multithreading
|
||||
- [ ] Implement multithreading for performance in large environments
|
||||
- [ ] Add unit tests
|
||||
- [ ] Add integration tests
|
||||
- [ ] Redesign logging messages to improve clarity and consistency
|
||||
|
||||
## Developer Documentation
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
[tool.poetry]
|
||||
name = "vault2vault"
|
||||
version = "0.1.0"
|
||||
version = "0.1.2"
|
||||
license = "MIT"
|
||||
authors = ["Ethan Paul <24588726+enpaul@users.noreply.github.com>"]
|
||||
description = "Recursively rekey ansible-vault encrypted files and in-line variables"
|
||||
@@ -12,7 +12,7 @@ packages = [
|
||||
keywords = ["ansible", "vault", "playbook", "yaml", "password"]
|
||||
readme = "README.md"
|
||||
classifiers = [
|
||||
"Development Status :: 3 - Alpha",
|
||||
"Development Status :: 4 - Beta",
|
||||
"Environment :: Console",
|
||||
"Framework :: Ansible",
|
||||
"Intended Audience :: Developers",
|
||||
|
||||
481
vault2vault.py
481
vault2vault.py
@@ -9,24 +9,26 @@ from pathlib import Path
|
||||
from typing import Any
|
||||
from typing import Iterable
|
||||
from typing import List
|
||||
from typing import Tuple
|
||||
from typing import Union
|
||||
from typing import Optional
|
||||
|
||||
import ruamel.yaml
|
||||
|
||||
try:
|
||||
import ansible.constants
|
||||
import ansible.parsing.vault
|
||||
from ansible.parsing.vault import VaultSecret
|
||||
from ansible.parsing.vault import VaultLib
|
||||
from ansible.parsing.vault import AnsibleVaultError
|
||||
except ImportError:
|
||||
print(
|
||||
"FATAL: No supported version of Ansible could be imported under the current python interpreter"
|
||||
"FATAL: No supported version of Ansible could be imported under the current python interpreter",
|
||||
file=sys.stderr,
|
||||
)
|
||||
sys.exit(1)
|
||||
|
||||
|
||||
__title__ = "vault2vault"
|
||||
__summary__ = "Recursively rekey ansible-vault encrypted files and in-line variables"
|
||||
__version__ = "0.1.0"
|
||||
__version__ = "0.1.2"
|
||||
__url__ = "https://github.com/enpaul/vault2vault/"
|
||||
__license__ = "MIT"
|
||||
__authors__ = ["Ethan Paul <24588726+enpaul@users.noreply.github.com>"]
|
||||
@@ -44,8 +46,8 @@ ruamel.yaml.add_constructor(
|
||||
|
||||
|
||||
def rekey(
|
||||
old: ansible.parsing.vault.VaultLib,
|
||||
new: ansible.parsing.vault.VaultLib,
|
||||
old: VaultLib,
|
||||
new: VaultLib,
|
||||
content: bytes,
|
||||
) -> bytes:
|
||||
"""Rekey vaulted content to use a new vault password
|
||||
@@ -61,6 +63,208 @@ def rekey(
|
||||
return new.encrypt(old.decrypt(content))
|
||||
|
||||
|
||||
# This whole function needs to be rebuilt from the ground up so I don't
|
||||
# feel bad about disabling this warning
|
||||
def _process_file( # pylint: disable=too-many-statements
|
||||
path: Path,
|
||||
old: VaultLib,
|
||||
new: VaultLib,
|
||||
interactive: bool,
|
||||
backup: bool,
|
||||
ignore: bool,
|
||||
) -> None:
|
||||
"""Determine whether a filepath includes vaulted data and if so, rekey it
|
||||
|
||||
:param path: Path to the file to check
|
||||
:param old: VaultLib object with the current (old) vault password encoded in it
|
||||
:param new: VaultLib object with the target (new) vault password encoded in it
|
||||
:param interactive: Whether to prompt interactively for confirmation before each
|
||||
rekey operation
|
||||
:param backup: Whether to copy the original file to a backup before making any
|
||||
in-place changes
|
||||
:param ignore: Whether to ignore any errors that come from failing to decrypt
|
||||
any vaulted data
|
||||
"""
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
logger.debug(f"Processing file {path}")
|
||||
|
||||
def _process_yaml_data( # pylint: disable=too-many-locals
|
||||
content: bytes, data: Any, ignore: bool, name: str = ""
|
||||
):
|
||||
if isinstance(data, dict):
|
||||
for key, value in data.items():
|
||||
content = _process_yaml_data(
|
||||
content, value, ignore, name=f"{name}.{key}"
|
||||
)
|
||||
elif isinstance(data, list):
|
||||
for index, item in enumerate(data):
|
||||
content = _process_yaml_data(
|
||||
content, item, ignore, name=f"{name}.{index}"
|
||||
)
|
||||
elif isinstance(data, ruamel.yaml.comments.TaggedScalar) and old.is_encrypted(
|
||||
data.value
|
||||
):
|
||||
logger.info(f"Identified vaulted content in {path} at {name}")
|
||||
confirm = (
|
||||
_confirm(f"Rekey vault encrypted variable {name} in file {path}?")
|
||||
if interactive
|
||||
else True
|
||||
)
|
||||
|
||||
if not confirm:
|
||||
logger.debug(
|
||||
f"User skipped vault encrypted content in {path} at {name} via interactive mode"
|
||||
)
|
||||
return content
|
||||
|
||||
try:
|
||||
new_data = rekey(old, new, data.value.encode())
|
||||
except AnsibleVaultError as err:
|
||||
msg = f"Failed to decrypt vault encrypted data in {path} at {name} with provided vault secret"
|
||||
if ignore:
|
||||
logger.warning(msg)
|
||||
return content
|
||||
raise RuntimeError(msg) from err
|
||||
content_decoded = content.decode("utf-8")
|
||||
|
||||
# Ok so this next section is probably the worst possible way to do this, but I did
|
||||
# it this way to solve a very specific problem that would absolutely prevent people
|
||||
# from using this tool: round trip YAML format preservation. Namely, that it's impossible.
|
||||
# Ruamel gets the closest to achieving this: it can do round trip format preservation
|
||||
# when the starting state is in _some_ known state (this is better than competitors which
|
||||
# require the starting state to be in a _specific_ known state). But given how many
|
||||
# ways there are to write YAML- and by extension, how many opinions there are on the
|
||||
# "correct" way to write YAML- it is not possible to configure ruamel to account for all of
|
||||
# them, even if everyones YAML style was compatible with ruamel's roundtrip formatting (note:
|
||||
# they aren't). So there's the problem: to be useful, this tool would need to reformat every
|
||||
# YAML file it touched, which means nobody would use it.
|
||||
#
|
||||
# To avoid the YAML formatting problem, we need a way to replace the target content
|
||||
# in the raw text of the file without dumping the parsed YAML. We want to preserve
|
||||
# indendation, remove any extra newlines that would be left over, add any necessary
|
||||
# newlines without clobbering the following lines, and ideally avoid reimplementing
|
||||
# a YAML formatter. The answer to this problem- as the answer to so many stupid problems
|
||||
# seems to be- is a regex. If this is too janky for you (I know it is for me) go support
|
||||
# the estraven project I'm trying to get off the ground: https://github.com/enpaul/estraven
|
||||
#
|
||||
# Ok, thanks for sticking with me as I was poetic about this. The solution below...
|
||||
# is awful, I can admit that. But it does work, so I'll leave it up to
|
||||
# your judgement as to whether it's worthwhile or not. Here's how it works:
|
||||
#
|
||||
# 1. First we take the first line of the original (unmodified) vaulted content. This line
|
||||
# of text has several important qualities: 1) it exists in the raw text of the file, 2)
|
||||
# it is pseudo-guaranteed to be unique, and 3) it is guaranteed to exist (vaulted content
|
||||
# will be at least one line long, but possibly no more)
|
||||
search_data = data.value.split("\n")[1]
|
||||
try:
|
||||
# 2. Next we use a regex to grab the full line of text from the file that includes the above
|
||||
# string. This is important because the full line of text will include the leading
|
||||
# whitespace, which ruamel helpfully strips out from the parsed data.
|
||||
# 3. Next we grab the number of leading spaces on the line using the capture group from the
|
||||
# regex
|
||||
padding = len(
|
||||
re.search(rf"\n(\s*){search_data}\n", content_decoded).groups()[0]
|
||||
)
|
||||
except (TypeError, AttributeError):
|
||||
# This is to handle an edgecase where the vaulted content is actually a yaml anchor. For
|
||||
# example, if a single vaulted secret needs to be stored under multiple variable names.
|
||||
# In that case, the vaulted content iself will only appear once in the file, but the data
|
||||
# parsed by ruamel will include it twice. If we fail to get a match on the first line, then
|
||||
# we check whether the data is a yaml anchor and, if it is, we skip it.
|
||||
if data.anchor.value:
|
||||
logger.debug(
|
||||
f"Content replacement for encrypted content in {path} at {name} was not found, so replacement will be skipped because target is a YAML anchor"
|
||||
)
|
||||
return content
|
||||
raise
|
||||
|
||||
# 4. Now with the leading whitespace padding, we add this same number of spaces to each line
|
||||
# of *both* the old vaulted data and the new vaulted data. It's important to do both because
|
||||
# we'll need to do a replacement in a moment so we need to know both what we're replacing
|
||||
# and what we're replacing it with.
|
||||
padded_old_data = "\n".join(
|
||||
[f"{' ' * padding}{item}" for item in data.value.split("\n") if item]
|
||||
)
|
||||
padded_new_data = "\n".join(
|
||||
[
|
||||
f"{' ' * padding}{item}"
|
||||
for item in new_data.decode("utf-8").split("\n")
|
||||
if item
|
||||
]
|
||||
)
|
||||
|
||||
# 5. Finally, we actually replace the content. This needs to have a count=1 so that if the same
|
||||
# encrypted block appears twice in the same file we only replace the first occurance of it,
|
||||
# otherwise the later replacement attempts will fail. We also need to re-encode it back to
|
||||
# bytes because all file operations with vault are done in bytes mode
|
||||
content = content_decoded.replace(
|
||||
padded_old_data, padded_new_data, 1
|
||||
).encode()
|
||||
return content
|
||||
|
||||
with path.open("rb") as infile:
|
||||
raw = infile.read()
|
||||
|
||||
# The 'is_encrypted' check doesn't rely on the vault secret in the VaultLib matching the
|
||||
# secret the data was encrypted with, it just checks that the data is encrypted with some
|
||||
# vault secret. We could use either `old` or `new` for this check, it doesn't actually matter.
|
||||
if old.is_encrypted(raw):
|
||||
logger.info(f"Identified vault encrypted file: {path}")
|
||||
|
||||
confirm = (
|
||||
_confirm(f"Rekey vault encrypted file {path}?") if interactive else True
|
||||
)
|
||||
|
||||
if not confirm:
|
||||
logger.debug(
|
||||
f"User skipped vault encrypted file {path} via interactive mode"
|
||||
)
|
||||
return
|
||||
|
||||
if backup:
|
||||
path.rename(f"{path}.bak")
|
||||
|
||||
try:
|
||||
updated = rekey(old, new, raw)
|
||||
except AnsibleVaultError:
|
||||
msg = f"Failed to decrypt vault encrypted file {path} with provided vault secret"
|
||||
if ignore:
|
||||
logger.warning(msg)
|
||||
return
|
||||
raise RuntimeError(msg) from None
|
||||
elif path.suffix.lower() in YAML_FILE_EXTENSIONS:
|
||||
logger.debug(f"Identified YAML file: {path}")
|
||||
|
||||
confirm = (
|
||||
_confirm(f"Search YAML file {path} for vault encrypted variables?")
|
||||
if interactive
|
||||
else True
|
||||
)
|
||||
|
||||
data = yaml.load(raw)
|
||||
|
||||
if not confirm:
|
||||
logger.debug(
|
||||
f"User skipped processing YAML file {path} via interactive mode"
|
||||
)
|
||||
return
|
||||
|
||||
if backup:
|
||||
shutil.copy(path, f"{path}.bak")
|
||||
|
||||
updated = _process_yaml_data(raw, data, ignore=ignore)
|
||||
else:
|
||||
logger.debug(f"Skipping non-vault file {path}")
|
||||
return
|
||||
|
||||
logger.debug(f"Writing updated file contents to {path}")
|
||||
|
||||
with path.open("wb") as outfile:
|
||||
outfile.write(updated)
|
||||
|
||||
|
||||
def _get_args() -> argparse.Namespace:
|
||||
parser = argparse.ArgumentParser(
|
||||
prog=__title__,
|
||||
@@ -131,174 +335,6 @@ def _confirm(prompt: str, default: bool = True) -> bool:
|
||||
print("Please input one of the specified options", file=sys.stderr)
|
||||
|
||||
|
||||
# This whole function needs to be rebuilt from the ground up so I don't
|
||||
# feel bad about disabling this warning
|
||||
def _process_file( # pylint: disable=too-many-statements
|
||||
path: Path,
|
||||
old: ansible.parsing.vault.VaultLib,
|
||||
new: ansible.parsing.vault.VaultLib,
|
||||
interactive: bool,
|
||||
backup: bool,
|
||||
ignore: bool,
|
||||
) -> None:
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
logger.debug(f"Processing file {path}")
|
||||
|
||||
def _process_yaml_data(content: bytes, data: Any, name: str = ""):
|
||||
if isinstance(data, dict):
|
||||
for key, value in data.items():
|
||||
content = _process_yaml_data(content, value, f"{name}.{key}")
|
||||
elif isinstance(data, list):
|
||||
for index, item in enumerate(data):
|
||||
content = _process_yaml_data(content, item, f"{name}.{index}")
|
||||
elif isinstance(data, ruamel.yaml.comments.TaggedScalar) and old.is_encrypted(
|
||||
data.value
|
||||
):
|
||||
logger.debug(f"Identified vaulted content in {path} at '{name}'")
|
||||
confirm = (
|
||||
_confirm(f"Rekey vault encrypted variable {name} in file {path}?")
|
||||
if interactive
|
||||
else True
|
||||
)
|
||||
|
||||
if not confirm:
|
||||
logger.debug(
|
||||
f"User skipped vault encrypted content in {path} at '{name}' via interactive mode"
|
||||
)
|
||||
return content
|
||||
|
||||
new_data = rekey(old, new, data.value.encode())
|
||||
content_decoded = content.decode("utf-8")
|
||||
|
||||
# Ok so this next section is probably the worst possible way to do this, but I did
|
||||
# it this way to solve a very specific problem that would absolutely prevent people
|
||||
# from using this tool: round trip YAML format preservation. Namely, that it's impossible.
|
||||
# Ruamel gets the closest to achieving this: it can do round trip format preservation
|
||||
# when the starting state is in _some_ known state (this is better than competitors which
|
||||
# require the starting state to be in a _specific_ known state). But given how many
|
||||
# ways there are to write YAML- and by extension, how many opinions there are on the
|
||||
# "correct" way to write YAML- it is not possible to configure ruamel to account for all of
|
||||
# them, even if everyones YAML style was compatible with ruamel's roundtrip formatting (note:
|
||||
# they aren't). So there's the problem: to be useful, this tool would need to reformat every
|
||||
# YAML file it touched, which means nobody would use it.
|
||||
#
|
||||
# To avoid the YAML formatting problem, we need a way to replace the target content
|
||||
# in the raw text of the file without dumping the parsed YAML. We want to preserve
|
||||
# indendation, remove any extra newlines that would be left over, add any necessary
|
||||
# newlines without clobbering the following lines, and ideally avoid reimplementing
|
||||
# a YAML formatter. The answer to this problem- as the answer to so many stupid problems
|
||||
# seems to be- is a regex. If this is too janky for you (I know it is for me) go support
|
||||
# the estraven project I'm trying to get off the ground: https://github.com/enpaul/estraven
|
||||
#
|
||||
# Ok, thanks for sticking with me as I was poetic about this. The solution below...
|
||||
# is awful, I can admit that. But it does work, so I'll leave it up to
|
||||
# your judgement as to whether it's worthwhile or not. Here's how it works:
|
||||
#
|
||||
# 1. First we take the first line of the original (unmodified) vaulted content. This line
|
||||
# of text has several important qualities: 1) it exists in the raw text of the file, 2)
|
||||
# it is pseudo-guaranteed to be unique, and 3) it is guaranteed to exist (vaulted content
|
||||
# will be at least one line long, but possibly no more)
|
||||
search_data = data.value.split("\n")[1]
|
||||
try:
|
||||
# 2. Next we use a regex to grab the full line of text from the file that includes the above
|
||||
# string. This is important because the full line of text will include the leading
|
||||
# whitespace, which ruamel helpfully strips out from the parsed data.
|
||||
# 3. Next we grab the number of leading spaces on the line using the capture group from the
|
||||
# regex
|
||||
padding = len(
|
||||
re.search(rf"\n(\s*){search_data}\n", content_decoded).groups()[0]
|
||||
)
|
||||
except (TypeError, AttributeError):
|
||||
# This is to handle an edgecase where
|
||||
if data.anchor.value:
|
||||
logger.debug(
|
||||
f"Content replacement for encrypted content in {path} at {name} was not found, so replacement will be skipped because target is a YAML anchor"
|
||||
)
|
||||
return content
|
||||
raise
|
||||
|
||||
# 4. Now with the leading whitespace padding, we add this same number of spaces to each line
|
||||
# of *both* the old vaulted data and the new vaulted data. It's important to do both because
|
||||
# we'll need to do a replacement in a moment so we need to know both what we're replacing
|
||||
# and what we're replacing it with.
|
||||
padded_old_data = "\n".join(
|
||||
[f"{' ' * padding}{item}" for item in data.value.split("\n") if item]
|
||||
)
|
||||
padded_new_data = "\n".join(
|
||||
[
|
||||
f"{' ' * padding}{item}"
|
||||
for item in new_data.decode("utf-8").split("\n")
|
||||
if item
|
||||
]
|
||||
)
|
||||
|
||||
# 5. Finally, we actually replace the content. We also need to re-encode it back to bytes
|
||||
# because all file operations with vault are done in bytes mode
|
||||
content = content_decoded.replace(padded_old_data, padded_new_data).encode()
|
||||
return content
|
||||
|
||||
with path.open("rb") as infile:
|
||||
raw = infile.read()
|
||||
|
||||
# The 'is_encrypted' check doesn't rely on the vault secret in the VaultLib matching the
|
||||
# secret the data was encrypted with, it just checks that the data is encrypted with some
|
||||
# vault secret. We could use either `old` or `new` for this check, it doesn't actually matter.
|
||||
if old.is_encrypted(raw):
|
||||
logger.debug(f"Identified vault encrypted file: {path}")
|
||||
|
||||
confirm = (
|
||||
_confirm(f"Rekey vault encrypted file {path}?") if interactive else True
|
||||
)
|
||||
|
||||
if not confirm:
|
||||
logger.debug(
|
||||
f"User skipped vault encrypted file {path} via interactive mode"
|
||||
)
|
||||
return
|
||||
|
||||
if backup:
|
||||
path.rename(f"{path}.bak")
|
||||
|
||||
try:
|
||||
updated = rekey(old, new, raw)
|
||||
except ansible.parsing.vault.AnsibleVaultError:
|
||||
msg = f"Failed to decrypt vault encrypted file {path} with provided vault secret"
|
||||
if ignore:
|
||||
logger.warning(msg)
|
||||
return
|
||||
raise RuntimeError(msg) from None
|
||||
elif path.suffix.lower() in YAML_FILE_EXTENSIONS:
|
||||
logger.debug(f"Identified YAML file: {path}")
|
||||
|
||||
confirm = (
|
||||
_confirm(f"Search YAML file {path} for vault encrypted variables?")
|
||||
if interactive
|
||||
else True
|
||||
)
|
||||
|
||||
data = yaml.load(raw)
|
||||
|
||||
if not confirm:
|
||||
logger.debug(
|
||||
f"User skipped processing YAML file {path} via interactive mode"
|
||||
)
|
||||
return
|
||||
|
||||
if backup:
|
||||
shutil.copy(path, f"{path}.bak")
|
||||
|
||||
updated = _process_yaml_data(raw, data)
|
||||
else:
|
||||
logger.debug(f"Skipping non-vault file {path}")
|
||||
return
|
||||
|
||||
logger.debug(f"Writing updated file contents to {path}")
|
||||
|
||||
with path.open("wb") as outfile:
|
||||
outfile.write(updated)
|
||||
|
||||
|
||||
def _expand_paths(paths: Iterable[Path]) -> List[Path]:
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
@@ -309,61 +345,53 @@ def _expand_paths(paths: Iterable[Path]) -> List[Path]:
|
||||
logger.debug(f"Including file {path}")
|
||||
results.append(path)
|
||||
elif path.is_dir():
|
||||
logger.debug(f"Descending into subdirectory {path}")
|
||||
logger.debug(f"Identifying files under {path}")
|
||||
results += _expand_paths(path.iterdir())
|
||||
else:
|
||||
logger.debug(f"Discarding path {path}")
|
||||
return results
|
||||
|
||||
|
||||
def _read_vault_pass_file(path: Union[Path, str]) -> str:
|
||||
logger = logging.getLogger(__name__)
|
||||
try:
|
||||
with Path(path).resolve().open(encoding="utf-8") as infile:
|
||||
return infile.read()
|
||||
except (FileNotFoundError, PermissionError):
|
||||
logger.error(
|
||||
f"Specified vault password file '{path}' does not exist or is unreadable"
|
||||
)
|
||||
sys.exit(1)
|
||||
def _load_password(
|
||||
fpath: Optional[str], desc: str = "", confirm: bool = True
|
||||
) -> VaultSecret:
|
||||
"""Load a password from a file or interactively
|
||||
|
||||
:param fpath: Optional path to the file containing the vault password. If not provided then
|
||||
the password will be prompted for interactively.
|
||||
:param desc: Description text to inject into the interactive password prompt. Useful when using
|
||||
this function multiple times to identify different passwords to the user.
|
||||
:param confirm: Whether to prompt twice for the input and check that the two inputs match
|
||||
:returns: Populated vault secret object with the loaded password
|
||||
"""
|
||||
|
||||
def _load_passwords(
|
||||
old_file: str, new_file: str
|
||||
) -> Tuple[ansible.parsing.vault.VaultSecret, ansible.parsing.vault.VaultSecret]:
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
if old_file:
|
||||
old_vault_pass = _read_vault_pass_file(old_file)
|
||||
logger.info(f"Loaded old vault password from {Path(old_file).resolve()}")
|
||||
else:
|
||||
logger.debug(
|
||||
"No old vault password file provided, prompting for old vault password input"
|
||||
)
|
||||
old_vault_pass = getpass.getpass(
|
||||
prompt="Old Ansible Vault password: ", stream=sys.stderr
|
||||
if fpath:
|
||||
try:
|
||||
with Path(fpath).resolve().open("rb", encoding="utf-8") as infile:
|
||||
return VaultSecret(infile.read())
|
||||
except (FileNotFoundError, PermissionError) as err:
|
||||
raise RuntimeError(
|
||||
f"Specified vault password file '{fpath}' does not exist or is unreadable"
|
||||
) from err
|
||||
|
||||
logger.debug("No vault password file provided, prompting for interactive input")
|
||||
|
||||
password_1 = getpass.getpass(
|
||||
prompt=f"Enter {desc} Ansible Vault password: ", stream=sys.stderr
|
||||
)
|
||||
|
||||
if confirm:
|
||||
password_2 = getpass.getpass(
|
||||
prompt=f"Confirm (re-enter) {desc} Ansible Vault password: ",
|
||||
stream=sys.stderr,
|
||||
)
|
||||
|
||||
if new_file:
|
||||
new_vault_pass = _read_vault_pass_file(new_file)
|
||||
logger.info(f"Loaded new vault password from {Path(new_file).resolve()}")
|
||||
else:
|
||||
logger.debug(
|
||||
"No new vault password file provided, prompting for new vault password input"
|
||||
)
|
||||
new_vault_pass = getpass.getpass(
|
||||
prompt="New Ansible Vault password: ", stream=sys.stderr
|
||||
)
|
||||
confirm = getpass.getpass(
|
||||
prompt="Confirm new Ansible Vault password: ", stream=sys.stderr
|
||||
)
|
||||
if new_vault_pass != confirm:
|
||||
logger.error("New vault passwords do not match")
|
||||
sys.exit(1)
|
||||
if password_1 != password_2:
|
||||
raise RuntimeError(f"Provided {desc} passwords do not match")
|
||||
|
||||
return ansible.parsing.vault.VaultSecret(
|
||||
old_vault_pass.encode("utf-8")
|
||||
), ansible.parsing.vault.VaultSecret(new_vault_pass.encode("utf-8"))
|
||||
return VaultSecret(password_1.encode("utf-8"))
|
||||
|
||||
|
||||
def main():
|
||||
@@ -383,17 +411,26 @@ def main():
|
||||
sys.exit(0)
|
||||
|
||||
if not args.paths:
|
||||
logger.warning("No path provided, nothing to do!")
|
||||
logger.warning("No paths provided, nothing to do!")
|
||||
sys.exit(0)
|
||||
|
||||
old_pass, new_pass = _load_passwords(args.old_pass_file, args.new_pass_file)
|
||||
in_vault = ansible.parsing.vault.VaultLib([(args.vault_id, old_pass)])
|
||||
out_vault = ansible.parsing.vault.VaultLib([(args.vault_id, new_pass)])
|
||||
try:
|
||||
old_pass = _load_password(args.old_pass_file, desc="existing", confirm=False)
|
||||
new_pass = _load_password(args.new_pass_file, desc="new", confirm=True)
|
||||
|
||||
logger.debug(
|
||||
in_vault = VaultLib([(args.vault_id, old_pass)])
|
||||
out_vault = VaultLib([(args.vault_id, new_pass)])
|
||||
except RuntimeError as err:
|
||||
logger.error(str(err))
|
||||
sys.exit(1)
|
||||
except KeyboardInterrupt:
|
||||
sys.exit(130)
|
||||
|
||||
logger.info(
|
||||
f"Identifying all files under {len(args.paths)} input paths: {', '.join(args.paths)}"
|
||||
)
|
||||
files = _expand_paths(args.paths)
|
||||
logger.info(f"Identified {len(files)} files for processing")
|
||||
|
||||
for filepath in files:
|
||||
_process_file(
|
||||
|
||||
Reference in New Issue
Block a user