diff --git a/doc/.wordlist.txt b/doc/.wordlist.txt index 8a0ec8a14..1dbfebde8 100644 --- a/doc/.wordlist.txt +++ b/doc/.wordlist.txt @@ -28,6 +28,7 @@ YAML addons balancer boolean +datasource dropdown favicon installable @@ -36,6 +37,8 @@ namespaces observability reST reStructuredText +stdin subdirectories subfolders subtree +validator diff --git a/doc/howto/autoinstall-validation.rst b/doc/howto/autoinstall-validation.rst new file mode 100644 index 000000000..7a2067e43 --- /dev/null +++ b/doc/howto/autoinstall-validation.rst @@ -0,0 +1,264 @@ +.. _autoinstall_validation: + +Autoinstall Validation +===================================== + +The following how-to guide demonstrates how to perform pre-validation of a autoinstall config. + +Autoinstall config is validated against a :doc:`JSON schema <../reference/autoinstall-schema>` during runtime before it is applied. This check ensures existence of required keys and their data types, but does not guarantee total validity of the data provided (see the :ref:`Validator Limitations` section for more details). + +Pre-validating the Autoinstall configuration +-------------------------------------------- + +You can validate autoinstall config prior to install time by using the `validate-autoinstall-user-data script `_ in the Subiquity GitHub repository. + +Getting Started +^^^^^^^^^^^^^^^ + +Running the validation script requires downloading the Subiquity source code and installing the development dependencies. First, clone the Subiquity repository and ``cd`` into the root of the repository: + +.. code:: none + + git clone https://github.com/canonical/subiquity.git && cd subiquity/ + +Then the required dependencies can be installed by running: + +.. code:: none + + make install_deps + + +Now you can invoke the validation script with: + +.. code:: none + + ./scripts/validate-autoinstall-user-data + + +or you can feed the configuration data via stdin: + + +.. code:: none + + # a trivial example + cat | ./scripts/validate-autoinstall-user-data + +.. warning:: + + Never run the validation script as ``sudo``. + +Finally, after running the validation script it will report the result of the validation attempt: + +.. code:: none + + $ ./scripts/validate-autoinstall-user-data.py + Success: The provided autoinstall config validated successfully + +You can also use the exit codes to determine the result: 0 (success) or 1 (failure). + + +Choice of Delivery Method +^^^^^^^^^^^^^^^^^^^^^^^^^ + +By default the validation script will expect your autoinstall configuration to be passed via cloud-config and expects a valid cloud-config file containing an ``autoinstall`` section: + +.. code:: none + + #cloud-config + + # some cloud-init directives + + autoinstall: + # autoinstall directives + +This allows you to use the script directly on your cloud-config data. The validation script will extract the autoinstall configuration from the provided cloud-config data and perform the validation on the extracted autoinstall section directly. + + +If you want to validate autoinstall configurations which will be delivered via the installation media, like the following example: + +.. code:: none + + autoinstall: + # autoinstall directives + +then this can be signalled by passing the ``--no-expect-cloudconfig`` flag. Both formats in this delivery method, with or without a top-level ``autoinstall`` keyword, are supported in this mode. + +.. _validator-limitations: + +Validator Limitations +--------------------- + +The autoinstall validator currently has the following limitations: + +1. The validator makes an assumption about the target installation media that may not necessarily be true about the actual installation media. It assumes that (1) the installation target is ubuntu-server and (2) the only valid install source is :code:`synthesized`. Some cases where this would cause the validator fail otherwise correct autoinstall configurations: + + a. Missing both an :code:`identity` and :code:`user-data` section for a Desktop target, where these sections are fully optional. + b. A :code:`source` section which specifies any :code:`id` other than :code:`synthesized`, where the :code:`id` may really match a valid source on the target ISO. + +2. Validity of the data provided in each section is not guaranteed as some sections cannot be reasonably validated outside of the installation runtime environment (e.g., a bad :ref:`match directive `). + +3. The validator is unable to replicate some of the cloud-config based :ref:`delivery checks `. There are some basic checks performed to catch simple delivery-related errors, which you can read more about in the examples section, but the focus of the validation is on the Autoinstall configuration *after* it has been delivered to the installer. + +.. note:: + See the cloud-init documentation for `how to validate your cloud-config`_. + + +------------ + +Examples +-------- + +Common mistake #1 +^^^^^^^^^^^^^^^^^ + +If a top level ``autoinstall`` keyword is not found in the provided cloud-config during runtime then the installer will miss the autoinstall config and present an interactive session. To prevent occurrences of this issue, the validation script will report a failure if the provided cloud-config does not contain an autoinstall section. *This does not indicate a crash at runtime*, as you can definitely provide cloud-config without autoinstall, but it is a useful result for checking a common formatting mistake. + +.. tabs:: + + .. tab:: Validation output + + + Validating cloud-config which is missing the ``autoinstall`` keyword: + + .. code:: none + + $ ./scripts/validate-autoinstall-user-data.py + AssertionError: Expected data to be wrapped in cloud-config but could not find top level 'autoinstall' key. + Failure: The provided autoinstall config did not validate successfully + + .. tab:: Faulty config + + As an example, the following cloud-config contains an autoinstall section but has misspelled the ``autoinstall`` keyword: + + .. code:: none + + #cloud-config + autoinstll: + # autoinstall directives + + +Common Mistake #2 +^^^^^^^^^^^^^^^^^ + +Another common mistake is to forget the ``#cloud-config`` header in the cloud-config file, which will result in the installer "missing" the autoinstall configuration. + +.. tabs:: + + .. tab:: Validation output + + The validator will fail the provided cloud-config data if it does not contain the right header: + + + .. code:: none + + $ ./scripts/validate-autoinstall-user-data.py + AssertionError: Expected data to be wrapped in cloud-config but first line is not '#cloud-config'. Try passing --no-expect-cloudconfig. + Failure: The provided autoinstall config did not validate successfully + + + .. tab:: Faulty config + + Missing the ``#cloud-config`` header will mean the file is not read by cloud-init: + + .. code:: none + + autoinstall: + # autoinstall directives + + +Again, this is not indicative of a real runtime error that would appear. Instead, this case would result in having the installer presenting a fully interactive install where a partially or fully automated installation was desired instead. + +Common Mistake #3 +^^^^^^^^^^^^^^^^^ + +Another possible mistake is to think that the autoinstall config on the installation media is a cloud-config datasource (it is not): + +.. tabs:: + + .. tab:: Validation output + + When providing the autoinstall configuration using the top-level ``autoinstall`` keyword format, the installer will verify there are no other top-level keys: + + .. code:: none + + $ ./scripts/validate-autoinstall-user-data.py --no-expect-cloudconfig + error: subiquity/load_autoinstall_config/read_config: autoinstall.yaml is not a valid cloud config datasource. + No other keys may be present alongside 'autoinstall' at the top level. + Malformed autoinstall in 'top-level keys' section + Failure: The provided autoinstall config did not validate successfully + + .. tab:: Faulty config + + The following config contains cloud-config directives when it is not expected to contain any: + + .. code:: none + + #cloud-config + + # some cloud-config directives + + autoinstall: + # autoinstall directives + + + +Debugging errors +^^^^^^^^^^^^^^^^ + +By default, the validation script has low verbosity output: + +.. code:: none + + Malformed autoinstall in 'version or interactive-sections' section + Failure: The provided autoinstall config did not validate successfully + +However, you can increase the output level by successively passing the ``-v`` flag. At maximum verbosity, the validation script will report errors the same way they are reported at runtime. This is great for inspecting issues in cases where the short error message isn't yet specific enough to be useful and can be used to inspect specific JSON schema validation errors. + + +.. code:: none + + $ ./scripts/validate-autoinstall-user-data.py autoinstall.yaml -vvv + start: subiquity/load_autoinstall_config: + start: subiquity/load_autoinstall_config/read_config: + finish: subiquity/load_autoinstall_config/read_config: SUCCESS: + start: subiquity/Reporting/load_autoinstall_data: + finish: subiquity/Reporting/load_autoinstall_data: SUCCESS: + start: subiquity/Error/load_autoinstall_data: + finish: subiquity/Error/load_autoinstall_data: SUCCESS: + start: subiquity/core_validation: + finish: subiquity/core_validation: FAIL: Malformed autoinstall in 'version or interactive-sections' section + finish: subiquity/load_autoinstall_config: FAIL: Malformed autoinstall in 'version or interactive-sections' section + Malformed autoinstall in 'version or interactive-sections' section + Traceback (most recent call last): + File ".../subiquity/scripts/../subiquity/server/server.py", line 654, in validate_autoinstall + jsonschema.validate(self.autoinstall_config, self.base_schema) + File "/usr/lib/python3/dist-packages/jsonschema/validators.py", line 1080, in validate + raise error + jsonschema.exceptions.ValidationError: '*' is not of type 'array' + + Failed validating 'type' in schema['properties']['interactive-sections']: + {'items': {'type': 'string'}, 'type': 'array'} + + On instance['interactive-sections']: + '*' + + The above exception was the direct cause of the following exception: + + Traceback (most recent call last): + File ".../subiquity/./scripts/validate-autoinstall-user-data.py", line 186, in verify_autoinstall + app.load_autoinstall_config(only_early=True, context=None) + File ".../subiquity/scripts/../subiquitycore/context.py", line 159, in decorated_sync + return meth(self, **kw) + ^^^^^^^^^^^^^^^^ + File ".../subiquity/scripts/../subiquity/server/server.py", line 734, in load_autoinstall_config + self.validate_autoinstall() + File ".../subiquity/scripts/../subiquity/server/server.py", line 663, in validate_autoinstall + raise new_exception from original_exception + subiquity.server.autoinstall.AutoinstallValidationError: Malformed autoinstall in 'version or interactive-sections' section + Failure: The provided autoinstall config did not validate successfully + +In this case, the above output shows that ``interactive-sections`` section failed to validate against the JSON schema because the type provided was a ``string`` and not an ``array`` of ``string`` s. + +.. LINKS + +.. _how to validate your cloud-config: https://cloudinit.readthedocs.io/en/latest/howto/debug_user_data.html diff --git a/doc/howto/index.rst b/doc/howto/index.rst index 0dc988692..8087d1f54 100644 --- a/doc/howto/index.rst +++ b/doc/howto/index.rst @@ -19,6 +19,7 @@ Getting started with autoinstall autoinstall-quickstart-s390x basic-server-installation configure-storage + autoinstall-validation Found a problem? ---------------- diff --git a/doc/reference/autoinstall-reference.rst b/doc/reference/autoinstall-reference.rst index 8c73ae6c5..cadacef95 100644 --- a/doc/reference/autoinstall-reference.rst +++ b/doc/reference/autoinstall-reference.rst @@ -720,6 +720,8 @@ An example storage section: The extensions to the curtin syntax allow for disk selection and partition or logical-volume sizing. +.. _disk_selection_extensions: + Disk selection extensions ^^^^^^^^^^^^^^^^^^^^^^^^^ diff --git a/doc/reference/autoinstall-schema.rst b/doc/reference/autoinstall-schema.rst index ad4b62741..f99a7479a 100644 --- a/doc/reference/autoinstall-schema.rst +++ b/doc/reference/autoinstall-schema.rst @@ -3,14 +3,30 @@ Autoinstall schema ================== -The server installer validates the provided autoinstall configuration against a :ref:`JSON schema`. +The server installer validates the provided autoinstall configuration against a :ref:`JSON schema`. The end of this reference manual presents the schema as a single document which could be used to manually pre-validate an autoinstall configuration, however the actual runtime validation process is more involved than a simple JSON schema validation. See the provided :doc:`pre-validation script <../howto/autoinstall-validation>` for how to perform autoinstall pre-validation. + +.. _how_the_delivery_is_verified: + +How the delivery is verified +---------------------------- + +To ensure expected runtime behaviour after delivering the autoinstall config, the installer performs some sanity checks to ensure one delivery method is not confused for another. + +cloud-config +^^^^^^^^^^^^ + +When passing autoinstall via cloud-config, the installer will inspect the cloud-config data for any autoinstall-specific keywords outside of the top-level ``autoinstall`` keyword in the config and throw an error if any are encountered. If there are no misplaced keys, the data within the ``autoinstall`` section is passed to the installer. + + +Installation Media +^^^^^^^^^^^^^^^^^^ + +When passing autoinstall via the installation media and using the top-level ``autoinstall`` keyword format, the installer will inspect the passed autoinstall file to guarantee that there are no other top-level keys. This check guarantees that the autoinstall config is not mistaken for a cloud-config datasource. How the configuration is validated ---------------------------------- -This reference manual presents the schema as a single document. Use it pre-validate your configuration. - -At run time, the configuration is not validated against this document. Instead, configuration sections are loaded and validated in this order: +After the configuration has been delivered to the installer successfully, the configuration sections are loaded and validated in this order: 1. The reporting section is loaded, validated and applied. 2. The error commands are loaded and validated. @@ -35,7 +51,7 @@ Regeneration To regenerate the schema, run ``make schema`` in the root directory of the `Subiquity source repository`_. -.. LINKS +.. LINKS .. _JSON schema: https://json-schema.org/ .. _Subiquity source repository: https://github.com/canonical/subiquity diff --git a/doc/tutorial/providing-autoinstall.rst b/doc/tutorial/providing-autoinstall.rst index 4fa7e22c1..ed3a39b49 100644 --- a/doc/tutorial/providing-autoinstall.rst +++ b/doc/tutorial/providing-autoinstall.rst @@ -21,7 +21,7 @@ The suggested way of providing autoinstall configuration to the Ubuntu installer The autoinstall configuration is provided via cloud-init configuration, which is almost endlessly flexible. In most scenarios the easiest way will be to provide user data via the :external+cloud-init:ref:`datasource_nocloud` data source. -When providing autoinstall via cloud-init, the autoinstall configuration is provided as :external+cloud-init:ref:`user_data_formats-cloud_config`. This means it requires a :code:`#cloud-config` header. The autoinstall directives are placed under a top level :code:`autoinstall:` key: +When providing autoinstall via cloud-init, the autoinstall configuration is provided as :external+cloud-init:ref:`user_data_formats-cloud_config`. This means the file requires a :code:`#cloud-config` header and the autoinstall directives are placed under a top level :code:`autoinstall:` key: .. code-block:: yaml @@ -47,7 +47,7 @@ The autoinstall configuration provided in this way is passed to the Ubuntu insta version: 1 .... -Starting in 24.04 (Noble), to be consistent with the cloud-config based format, a top-level :code:`autoinstall:` keyword is allowed: +Starting in 24.04 (Noble), to be consistent with the cloud-config based format, a single top-level :code:`autoinstall:` keyword is allowed: .. code-block:: yaml @@ -69,7 +69,7 @@ Alternatively, you can pass the location of the autoinstall file on the kernel c Order of precedence for autoinstall locations ---------------------------------------------- +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Because there are many ways to specify the autoinstall file, it may happen that multiple locations are specified at the same time. Subiquity searches for the autoinstall file in the following order and uses the first existing one: diff --git a/scripts/runtests.sh b/scripts/runtests.sh index 46dbac2c1..f20a95d5a 100755 --- a/scripts/runtests.sh +++ b/scripts/runtests.sh @@ -49,7 +49,7 @@ validate () { answers-core-desktop|answers-uc24) ;; *) - python3 scripts/validate-autoinstall-user-data.py < $tmpdir/var/log/installer/autoinstall-user-data + python3 scripts/validate-autoinstall-user-data.py --legacy --check-link < $tmpdir/var/log/installer/autoinstall-user-data # After the lunar release and the introduction of mirror testing, it # came to our attention that new Ubuntu installations have the security # repository configured with the primary mirror URL (i.e., diff --git a/scripts/validate-autoinstall-user-data.py b/scripts/validate-autoinstall-user-data.py index 7713b1396..155cb4d06 100755 --- a/scripts/validate-autoinstall-user-data.py +++ b/scripts/validate-autoinstall-user-data.py @@ -14,77 +14,286 @@ # You should have received a copy of the GNU General Public License # along with this program. If not, see . -""" Entry-point to validate autoinstall-user-data against schema. +"""Validate autoinstall-user-data against the autoinstall schema. + By default, we are expecting the autoinstall user-data to be wrapped in a cloud -config format: +config format. Example: -#cloud-config -autoinstall: - + #cloud-config + autoinstall: + To validate the user-data directly, you can pass the --no-expect-cloudconfig switch. """ import argparse +import asyncio import io import json +import sys +import tempfile +import traceback +from argparse import ArgumentParser, Namespace +from pathlib import Path +from typing import Any import jsonschema import yaml +# Python path trickery so we can import subiquity code and still call this +# script without using the makefile. Eventually we should ship this a +# program in the subiquity snap, so users don't even have to checkout the +# source code but that will also require work to make sure Subiquity is +# safe to install on regular systems. +scripts_dir = sys.path[0] +subiquity_root = Path(scripts_dir) / ".." +curtin_root = subiquity_root / "curtin" +probert_root = subiquity_root / "probert" + +sys.path.insert(0, str(subiquity_root)) +sys.path.insert(1, str(curtin_root)) +sys.path.insert(2, str(probert_root)) + +from subiquity.cmd.server import make_server_args_parser # noqa: E402 +from subiquity.server.dryrun import DRConfig # noqa: E402 +from subiquity.server.server import SubiquityServer # noqa: E402 + +DOC_LINK: str = ( + "https://canonical-subiquity.readthedocs-hosted.com/en/latest/reference/autoinstall-reference.html" # noqa: E501 +) + + +SUCCESS_MSG: str = "Success: The provided autoinstall config validated successfully" +FAILURE_MSG: str = "Failure: The provided autoinstall config failed validation" -def main() -> None: - """ Entry point. """ - parser = argparse.ArgumentParser() - parser.add_argument("--json-schema", - help="Path to the JSON schema", - type=argparse.FileType("r"), - default="autoinstall-schema.json") - parser.add_argument("input", nargs="?", - help="Path to the user data instead of stdin", - type=argparse.FileType("r"), - default="-") - parser.add_argument("--no-expect-cloudconfig", dest="expect-cloudconfig", - action="store_false", - help="Assume the data is not wrapped in cloud-config.", - default=True) +def verify_link(data: str) -> bool: + """Verify the autoinstall doc link is in the generated user-data.""" - args = vars(parser.parse_args()) + return DOC_LINK in data - user_data: io.TextIOWrapper = args["input"] - if args["expect-cloudconfig"]: - assert user_data.readline() == "#cloud-config\n" - def get_autoinstall_data(data): return data["autoinstall"] +def parse_cloud_config(data: str) -> dict[str, Any]: + """Parse cloud-config and extra autoinstall data.""" + + # "#cloud-config" header is required for cloud-config data + first_line: str = data.splitlines()[0] + if not first_line == "#cloud-config": + raise AssertionError( + ( + "Expected data to be wrapped in cloud-config " + "but first line is not '#cloud-config'. Try " + "passing --no-expect-cloudconfig." + ) + ) + + cc_data: dict[str, Any] = yaml.safe_load(data) + + # "autoinstall" top-level keyword is required in cloud-config delivery case + if "autoinstall" not in cc_data: + raise AssertionError( + ( + "Expected data to be wrapped in cloud-config " + "but could not find top level 'autoinstall' " + "key." + ) + ) else: - def get_autoinstall_data(data): - try: - cfg = data["autoinstall"] - except KeyError: - cfg = data - return cfg + return cc_data["autoinstall"] - # Verify autoinstall doc link is in the file +def parse_autoinstall(user_data: str, expect_cloudconfig: bool) -> dict[str, Any]: + """Parse stringified user_data and extract autoinstall data.""" + + if expect_cloudconfig: + return parse_cloud_config(user_data) + else: + return yaml.safe_load(user_data) + + +def legacy_verify(ai_data: dict[str, Any], json_schema: io.TextIOWrapper) -> None: + """Legacy verification method for use in CI""" + + # support top-level "autoinstall" in regular autoinstall user data + if "autoinstall" in ai_data: + data: dict[str, Any] = ai_data["autoinstall"] + else: + data: dict[str, Any] = ai_data + + jsonschema.validate(data, json.load(json_schema)) + + +async def make_app() -> SubiquityServer: + parser: ArgumentParser = make_server_args_parser() + opts, unknown = parser.parse_known_args(["--dry-run"]) + app: SubiquityServer = SubiquityServer(opts, "") + # This is needed because the ubuntu-pro server controller accesses dr_cfg + # in the initializer. + app.dr_cfg = DRConfig() + app.base_model = app.make_model() + app.controllers.load_all() + return app + + +async def verify_autoinstall( + app: SubiquityServer, + cfg_path: str, + verbosity: int = 0, +) -> int: + """Verify autoinstall configuration using a SubiquityServer. + + Returns 0 if successfully validated. + Returns 1 if fails to validate. + """ + + # Tell the server where to load the autoinstall + app.autoinstall = cfg_path + + # Suppress start and finish events unless verbosity >=2 + if verbosity < 2: + for el in app.event_listeners: + el.report_start_event = lambda x, y: None + el.report_finish_event = lambda x, y, z: None + # Suppress info events unless verbosity >=1 + if verbosity < 1: + for el in app.event_listeners: + el.report_info_event = lambda x, y: None - stream_pos: int = user_data.tell() + # Make sure all events are printed (we could fail during read, which + # would happen before we setup the reporting controller) + app.controllers.Reporting.config = {"builtin": {"type": "print"}} + app.controllers.Reporting.start() - data: str = user_data.read() + # Validation happens during load phases. Do both phases. + try: + app.load_autoinstall_config(only_early=True, context=None) + app.load_autoinstall_config(only_early=False, context=None) + except Exception as exc: - link: str = "https://canonical-subiquity.readthedocs-hosted.com/en/latest/reference/autoinstall-reference.html" # noqa: E501 + print(exc) # Has the useful error message - assert link in data + # Print the full traceback if verbosity > 2 + if verbosity > 2: + traceback.print_exception(exc) + + print(FAILURE_MSG) + return 1 + + print(SUCCESS_MSG) + return 0 + + +async def _async_main(ai_user_data: dict[str, Any], args: Namespace) -> int: + # Make a dry-run server + app: SubiquityServer = await make_app() + + with tempfile.TemporaryDirectory() as td: + path = Path(td) / "autoinstall.yaml" + yaml_as_text: str = yaml.dump(ai_user_data) + path.write_text(yaml_as_text) + + return await verify_autoinstall( + app=app, + cfg_path=path, + verbosity=args.verbosity, + ) + + +def parse_args() -> Namespace: + """Parse argparse arguments.""" + + parser = argparse.ArgumentParser( + prog="validate-autoinstall-user-data", + description=__doc__, + formatter_class=argparse.RawTextHelpFormatter, + ) + + parser.add_argument( + "input", + nargs="?", + help="Path to the user data instead of stdin", + type=argparse.FileType("r"), + default="-", + ) + parser.add_argument( + "--no-expect-cloudconfig", + dest="expect_cloudconfig", + action="store_false", + help="Assume the data is not wrapped in cloud-config.", + default=True, + ) + + # Hidden validation path we use in CI until the new validation method + # is ready. i.e. continue to validate based on the json schema directly. + parser.add_argument( + "--json-schema", + help=argparse.SUPPRESS, + type=argparse.FileType("r"), + default="autoinstall-schema.json", + ) + + parser.add_argument( + "--legacy", + action="store_true", + help=argparse.SUPPRESS, + default=False, + ) + + # An option we use in CI to make sure Subiquity will insert a link to + # the documentation in the auto-generated autoinstall file post-install. + # There's not need for users to check this. + parser.add_argument( + "--check-link", + dest="check_link", + action="store_true", + help=argparse.SUPPRESS, + default=False, + ) + parser.add_argument( + "-v", + "--verbosity", + action="count", + help=( + "Increase output verbosity. Use -v for more info, -vv for " + "detailed output, and -vvv for fully detailed output." + ), + default=0, + ) + + return parser.parse_args() + + +def main() -> int: + """Entry point.""" + + args: Namespace = parse_args() + + str_user_data: str = args.input.read() + + # Verify autoinstall doc link is in the file + + if args.check_link: + + assert verify_link(str_user_data), "Documentation link missing from user data" # Verify autoinstall schema - user_data.seek(stream_pos) - data = yaml.safe_load(user_data) + try: + ai_user_data: dict[str, Any] = parse_autoinstall( + str_user_data, args.expect_cloudconfig + ) + except Exception as exc: + print(f"FAILURE: {exc}") + return 1 + + if args.legacy: + legacy_verify(ai_user_data, args.json_schema) + print(SUCCESS_MSG) + return 0 - jsonschema.validate(get_autoinstall_data(data), - json.load(args["json_schema"])) + return asyncio.run(_async_main(ai_user_data, args)) if __name__ == "__main__": - main() + sys.exit(main())