diff --git a/swh/objstorage/api/server.py b/swh/objstorage/api/server.py index 7201af2..3a00106 100644 --- a/swh/objstorage/api/server.py +++ b/swh/objstorage/api/server.py @@ -1,204 +1,204 @@ # Copyright (C) 2015-2022 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import contextlib import functools import logging import os from flask import request from swh.core.api import RPCServerApp from swh.core.api import encode_data_server as encode_data from swh.core.api import error_handler from swh.core.config import read as config_read from swh.core.statsd import statsd from swh.model import hashutil from swh.objstorage.exc import Error, ObjNotFoundError from swh.objstorage.factory import get_objstorage as get_swhobjstorage from swh.objstorage.interface import ObjStorageInterface from swh.objstorage.objstorage import DEFAULT_LIMIT def timed(f): @functools.wraps(f) def w(*a, **kw): with statsd.timed( "swh_objstorage_request_duration_seconds", tags={"endpoint": f.__name__} ): return f(*a, **kw) return w @contextlib.contextmanager def timed_context(f_name): with statsd.timed( "swh_objstorage_request_duration_seconds", tags={"endpoint": f_name} ): yield def get_objstorage(): global objstorage if objstorage is None: objstorage = get_swhobjstorage(**app.config["objstorage"]) return objstorage class ObjStorageServerApp(RPCServerApp): client_exception_classes = (ObjNotFoundError, Error) method_decorators = [timed] def pre_add(self, kw): """Called before the 'add' method.""" statsd.increment( "swh_objstorage_in_bytes_total", len(kw["content"]), tags={"endpoint": "add_bytes"}, ) def post_get(self, ret, kw): """Called after the 'get' method.""" statsd.increment( "swh_objstorage_out_bytes_total", len(ret), tags={"endpoint": "get_bytes"} ) app = ObjStorageServerApp( - __name__, backend_class=ObjStorageInterface, backend_factory=get_objstorage, + __name__, + backend_class=ObjStorageInterface, + backend_factory=get_objstorage, ) objstorage = None @app.errorhandler(Error) def argument_error_handler(exception): return error_handler(exception, encode_data, status_code=400) @app.errorhandler(Exception) def my_error_handler(exception): return error_handler(exception, encode_data) @app.route("/") @timed def index(): return "SWH Objstorage API server" # Streaming methods @app.route("/content/get_stream/") def get_stream(hex_id): obj_id = hashutil.hash_to_bytes(hex_id) def generate(): with timed_context("get_stream"): yield from get_objstorage().get_stream(obj_id, 2 << 20) return app.response_class(generate()) @app.route("/content") def list_content(): last_obj_id = request.args.get("last_obj_id") if last_obj_id: last_obj_id = bytes.fromhex(last_obj_id) limit = int(request.args.get("limit", DEFAULT_LIMIT)) def generate(): with timed_context("get_stream"): yield from get_objstorage().list_content(last_obj_id, limit=limit) return app.response_class(generate()) api_cfg = None def load_and_check_config(config_file): """Check the minimal configuration is set to run the api or raise an error explanation. Args: config_file (str): Path to the configuration file to load Raises: Error if the setup is not as expected Returns: configuration as a dict """ if not config_file: raise EnvironmentError("Configuration file must be defined") if not os.path.exists(config_file): raise FileNotFoundError("Configuration file %s does not exist" % (config_file,)) cfg = config_read(config_file) return validate_config(cfg) def validate_config(cfg): """Check the minimal configuration is set to run the api or raise an explanatory error. Args: cfg (dict): Loaded configuration. Raises: Error if the setup is not as expected Returns: configuration as a dict """ if "objstorage" not in cfg: raise KeyError("Invalid configuration; missing objstorage config entry") missing_keys = [] vcfg = cfg["objstorage"] if "cls" not in vcfg: raise KeyError("Invalid configuration; missing cls config entry") cls = vcfg["cls"] if cls == "pathslicing": # Backwards-compatibility: either get the deprecated `args` from the # objstorage config, or use the full config itself to check for keys args = vcfg.get("args", vcfg) for key in ("root", "slicing"): v = args.get(key) if v is None: missing_keys.append(key) if missing_keys: raise KeyError( "Invalid configuration; missing %s config entry" % (", ".join(missing_keys),) ) return cfg def make_app_from_configfile(): - """Load configuration and then build application to run - - """ + """Load configuration and then build application to run""" global api_cfg if not api_cfg: config_path = os.environ.get("SWH_CONFIG_FILENAME") api_cfg = load_and_check_config(config_path) app.config.update(api_cfg) handler = logging.StreamHandler() app.logger.addHandler(handler) return app if __name__ == "__main__": print("Deprecated. Use swh-objstorage") diff --git a/swh/objstorage/backends/azure.py b/swh/objstorage/backends/azure.py index 029dab3..1cab6a2 100644 --- a/swh/objstorage/backends/azure.py +++ b/swh/objstorage/backends/azure.py @@ -1,433 +1,419 @@ # Copyright (C) 2016-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import asyncio import contextlib import datetime from itertools import product import string from typing import Dict, Optional, Union import warnings from azure.core.exceptions import ResourceExistsError, ResourceNotFoundError from azure.storage.blob import ( ContainerClient, ContainerSasPermissions, generate_container_sas, ) from azure.storage.blob.aio import ContainerClient as AsyncContainerClient from swh.model import hashutil from swh.objstorage.exc import Error, ObjNotFoundError from swh.objstorage.objstorage import ( ObjStorage, compressors, compute_hash, decompressors, ) from swh.objstorage.utils import call_async def get_container_url( account_name: str, account_key: str, container_name: str, access_policy: str = "read_only", expiry: datetime.timedelta = datetime.timedelta(days=365), **kwargs, ) -> str: """Get the full url, for the given container on the given account, with a Shared Access Signature granting the specified access policy. Args: account_name: name of the storage account for which to generate the URL account_key: shared account key of the storage account used to generate the SAS container_name: name of the container for which to grant access in the storage account access_policy: one of ``read_only``, ``append_only``, ``full`` expiry: the interval in the future with which the signature will expire Returns: the full URL of the container, with the shared access signature. """ access_policies = { "read_only": ContainerSasPermissions( read=True, list=True, delete=False, write=False ), "append_only": ContainerSasPermissions( read=True, list=True, delete=False, write=True ), "full": ContainerSasPermissions(read=True, list=True, delete=True, write=True), } current_time = datetime.datetime.utcnow() signature = generate_container_sas( account_name, container_name, account_key=account_key, permission=access_policies[access_policy], start=current_time + datetime.timedelta(minutes=-1), expiry=current_time + expiry, ) return f"https://{account_name}.blob.core.windows.net/{container_name}?{signature}" class AzureCloudObjStorage(ObjStorage): """ObjStorage backend for Azure blob storage accounts. Args: container_url: the URL of the container in which the objects are stored. account_name: (deprecated) the name of the storage account under which objects are stored api_secret_key: (deprecated) the shared account key container_name: (deprecated) the name of the container under which objects are stored compression: the compression algorithm used to compress objects in storage Notes: The container url should contain the credentials via a "Shared Access Signature". The :func:`get_container_url` helper can be used to generate such a URL from the account's access keys. The ``account_name``, ``api_secret_key`` and ``container_name`` arguments are deprecated. """ def __init__( self, container_url: Optional[str] = None, account_name: Optional[str] = None, api_secret_key: Optional[str] = None, container_name: Optional[str] = None, compression="gzip", **kwargs, ): if container_url is None: if account_name is None or api_secret_key is None or container_name is None: raise ValueError( "AzureCloudObjStorage must have a container_url or all three " "account_name, api_secret_key and container_name" ) else: warnings.warn( "The Azure objstorage account secret key parameters are " "deprecated, please use container URLs instead.", DeprecationWarning, ) container_url = get_container_url( account_name=account_name, account_key=api_secret_key, container_name=container_name, access_policy="full", ) super().__init__(**kwargs) self.container_url = container_url self.compression = compression def get_container_client(self, hex_obj_id): """Get the container client for the container that contains the object with internal id hex_obj_id This is used to allow the PrefixedAzureCloudObjStorage to dispatch the client according to the prefix of the object id. """ return ContainerClient.from_container_url(self.container_url) @contextlib.asynccontextmanager async def get_async_container_clients(self): """Returns a collection of container clients, to be passed to ``get_async_blob_client``. Each container may not be used in more than one asyncio loop.""" client = AsyncContainerClient.from_container_url(self.container_url) async with client: yield {"": client} def get_blob_client(self, hex_obj_id): """Get the azure blob client for the given hex obj id""" container_client = self.get_container_client(hex_obj_id) return container_client.get_blob_client(blob=hex_obj_id) def get_async_blob_client(self, hex_obj_id, container_clients): """Get the azure blob client for the given hex obj id and a collection yielded by ``get_async_container_clients``.""" return container_clients[""].get_blob_client(blob=hex_obj_id) def get_all_container_clients(self): """Get all active block_blob_services""" yield self.get_container_client("") def _internal_id(self, obj_id): - """Internal id is the hex version in objstorage. - - """ + """Internal id is the hex version in objstorage.""" return hashutil.hash_to_hex(obj_id) def check_config(self, *, check_write): """Check the configuration for this object storage""" for container_client in self.get_all_container_clients(): props = container_client.get_container_properties() # FIXME: check_write is ignored here if not props: return False return True def __contains__(self, obj_id): - """Does the storage contains the obj_id. - - """ + """Does the storage contains the obj_id.""" hex_obj_id = self._internal_id(obj_id) client = self.get_blob_client(hex_obj_id) try: client.get_blob_properties() except ResourceNotFoundError: return False else: return True def __iter__(self): - """Iterate over the objects present in the storage. - - """ + """Iterate over the objects present in the storage.""" for client in self.get_all_container_clients(): for obj in client.list_blobs(): yield hashutil.hash_to_bytes(obj.name) def __len__(self): """Compute the number of objects in the current object storage. Returns: number of objects contained in the storage. """ return sum(1 for i in self) def add(self, content, obj_id=None, check_presence=True): - """Add an obj in storage if it's not there already. - - """ + """Add an obj in storage if it's not there already.""" if obj_id is None: # Checksum is missing, compute it on the fly. obj_id = compute_hash(content) if check_presence and obj_id in self: return obj_id hex_obj_id = self._internal_id(obj_id) # Send the compressed content compressor = compressors[self.compression]() data = compressor.compress(content) data += compressor.flush() client = self.get_blob_client(hex_obj_id) try: client.upload_blob(data=data, length=len(data)) except ResourceExistsError: # There's a race condition between check_presence and upload_blob, # that we can't get rid of as the azure api doesn't allow atomic # replaces or renaming a blob. As the restore operation explicitly # removes the blob, it should be safe to just ignore the error. pass return obj_id def restore(self, content, obj_id=None): - """Restore a content. - - """ + """Restore a content.""" if obj_id is None: # Checksum is missing, compute it on the fly. obj_id = compute_hash(content) if obj_id in self: self.delete(obj_id) return self.add(content, obj_id, check_presence=False) def get(self, obj_id): - """retrieve blob's content if found. - - """ + """retrieve blob's content if found.""" return call_async(self._get_async, obj_id) async def _get_async(self, obj_id, container_clients=None): """Coroutine implementing ``get(obj_id)`` using azure-storage-blob's asynchronous implementation. While ``get(obj_id)`` does not need asynchronicity, this is useful to ``get_batch(obj_ids)``, as it can run multiple ``_get_async`` tasks concurrently.""" if container_clients is None: # If the container_clients argument is not passed, create a new # collection of container_clients and restart the function with it. async with self.get_async_container_clients() as container_clients: return await self._get_async(obj_id, container_clients) hex_obj_id = self._internal_id(obj_id) client = self.get_async_blob_client(hex_obj_id, container_clients) try: download = await client.download_blob() except ResourceNotFoundError: raise ObjNotFoundError(obj_id) from None else: data = await download.content_as_bytes() decompressor = decompressors[self.compression]() ret = decompressor.decompress(data) if decompressor.unused_data: raise Error("Corrupt object %s: trailing data found" % hex_obj_id) return ret async def _get_async_or_none(self, obj_id, container_clients): """Like ``get_async(obj_id)``, but returns None instead of raising ResourceNotFoundError. Used by ``get_batch`` so other blobs can be returned even if one is missing.""" try: return await self._get_async(obj_id, container_clients) except ObjNotFoundError: return None async def _get_batch_async(self, obj_ids): async with self.get_async_container_clients() as container_clients: return await asyncio.gather( *[ self._get_async_or_none(obj_id, container_clients) for obj_id in obj_ids ] ) def get_batch(self, obj_ids): """Retrieve objects' raw content in bulk from storage, concurrently.""" return call_async(self._get_batch_async, obj_ids) def check(self, obj_id): - """Check the content integrity. - - """ + """Check the content integrity.""" obj_content = self.get(obj_id) content_obj_id = compute_hash(obj_content) if content_obj_id != obj_id: raise Error(obj_id) def delete(self, obj_id): """Delete an object.""" super().delete(obj_id) # Check delete permission hex_obj_id = self._internal_id(obj_id) client = self.get_blob_client(hex_obj_id) try: client.delete_blob() except ResourceNotFoundError: raise ObjNotFoundError(obj_id) from None return True class PrefixedAzureCloudObjStorage(AzureCloudObjStorage): """ObjStorage with azure capabilities, striped by prefix. accounts is a dict containing entries of the form: : """ def __init__( self, accounts: Dict[str, Union[str, Dict[str, str]]], compression="gzip", **kwargs, ): # shortcut AzureCloudObjStorage __init__ ObjStorage.__init__(self, **kwargs) self.compression = compression # Definition sanity check prefix_lengths = set(len(prefix) for prefix in accounts) if not len(prefix_lengths) == 1: raise ValueError( "Inconsistent prefixes, found lengths %s" % ", ".join(str(lst) for lst in sorted(prefix_lengths)) ) self.prefix_len = prefix_lengths.pop() expected_prefixes = set( "".join(letters) for letters in product( set(string.hexdigits.lower()), repeat=self.prefix_len ) ) missing_prefixes = expected_prefixes - set(accounts) if missing_prefixes: raise ValueError( "Missing prefixes %s" % ", ".join(sorted(missing_prefixes)) ) do_warning = False self.container_urls = {} for prefix, container_url in accounts.items(): if isinstance(container_url, dict): do_warning = True container_url = get_container_url( account_name=container_url["account_name"], account_key=container_url["api_secret_key"], container_name=container_url["container_name"], access_policy="full", ) self.container_urls[prefix] = container_url if do_warning: warnings.warn( "The Azure objstorage account secret key parameters are " "deprecated, please use container URLs instead.", DeprecationWarning, ) def get_container_client(self, hex_obj_id): """Get the block_blob_service and container that contains the object with internal id hex_obj_id """ prefix = hex_obj_id[: self.prefix_len] return ContainerClient.from_container_url(self.container_urls[prefix]) @contextlib.asynccontextmanager async def get_async_container_clients(self): # This is equivalent to: # client1 = AsyncContainerClient.from_container_url(url1) # ... # client16 = AsyncContainerClient.from_container_url(url16) # async with client1, ..., client16: # yield {prefix1: client1, ..., prefix16: client16} clients = { prefix: AsyncContainerClient.from_container_url(url) for (prefix, url) in self.container_urls.items() } async with contextlib.AsyncExitStack() as stack: for client in clients.values(): await stack.enter_async_context(client) yield clients def get_async_blob_client(self, hex_obj_id, container_clients): """Get the azure blob client for the given hex obj id and a collection yielded by ``get_async_container_clients``.""" prefix = hex_obj_id[: self.prefix_len] return container_clients[prefix].get_blob_client(blob=hex_obj_id) def get_all_container_clients(self): """Get all active container clients""" # iterate on items() to sort blob services; # needed to be able to paginate in the list_content() method yield from ( self.get_container_client(prefix) for prefix in sorted(self.container_urls) ) diff --git a/swh/objstorage/backends/generator.py b/swh/objstorage/backends/generator.py index ac42203..3a0cca9 100644 --- a/swh/objstorage/backends/generator.py +++ b/swh/objstorage/backends/generator.py @@ -1,227 +1,226 @@ import functools import io from itertools import count, islice, repeat import logging import random from swh.objstorage.objstorage import DEFAULT_CHUNK_SIZE, DEFAULT_LIMIT, ObjStorage logger = logging.getLogger(__name__) class Randomizer: def __init__(self): self.size = 0 self.read(1024) # create a not-so-small initial buffer def read(self, size): if size > self.size: with open("/dev/urandom", "rb") as fobj: self.data = fobj.read(2 * size) self.size = len(self.data) # pick a random subset of our existing buffer idx = random.randint(0, self.size - size - 1) return self.data[idx : idx + size] def gen_sizes(): """generates numbers according to the rought distribution of file size in the SWH archive """ # these are the histogram bounds of the pg content.length column bounds = [ 0, 2, 72, 119, 165, 208, 256, 300, 345, 383, 429, 474, 521, 572, 618, 676, 726, 779, 830, 879, 931, 992, 1054, 1119, 1183, 1244, 1302, 1370, 1437, 1504, 1576, 1652, 1725, 1806, 1883, 1968, 2045, 2133, 2236, 2338, 2433, 2552, 2659, 2774, 2905, 3049, 3190, 3322, 3489, 3667, 3834, 4013, 4217, 4361, 4562, 4779, 5008, 5233, 5502, 5788, 6088, 6396, 6728, 7094, 7457, 7835, 8244, 8758, 9233, 9757, 10313, 10981, 11693, 12391, 13237, 14048, 14932, 15846, 16842, 18051, 19487, 20949, 22595, 24337, 26590, 28840, 31604, 34653, 37982, 41964, 46260, 51808, 58561, 66584, 78645, 95743, 122883, 167016, 236108, 421057, 1047367, 55056238, ] nbounds = len(bounds) for i in count(): idx = random.randint(1, nbounds - 1) lower = bounds[idx - 1] upper = bounds[idx] yield random.randint(lower, upper - 1) def gen_random_content(total=None, filesize=None): """generates random (file) content which sizes roughly follows the SWH archive file size distribution (by default). Args: total (int): the total number of objects to generate. Infinite if unset. filesize (int): generate objects with fixed size instead of random ones. """ randomizer = Randomizer() if filesize: gen = repeat(filesize) else: gen = gen_sizes() if total: gen = islice(gen, total) for objsize in gen: yield randomizer.read(objsize) class RandomGeneratorObjStorage(ObjStorage): - """A stupid read-only storage that generates blobs for testing purpose. - """ + """A stupid read-only storage that generates blobs for testing purpose.""" def __init__(self, filesize=None, total=None, **kwargs): super().__init__() if filesize: filesize = int(filesize) self.filesize = filesize if total: total = int(total) self.total = total self._content_generator = None @property def content_generator(self): if self._content_generator is None: self._content_generator = gen_random_content(self.total, self.filesize) return self._content_generator def check_config(self, *, check_write): return True def __contains__(self, obj_id, *args, **kwargs): return False def __iter__(self): i = 1 while True: j = yield (b"%d" % i) if self.total and i >= self.total: logger.debug("DONE") break if j is not None: i = j else: i += 1 def get(self, obj_id, *args, **kwargs): return next(self.content_generator) def add(self, content, obj_id=None, check_presence=True, *args, **kwargs): pass def check(self, obj_id, *args, **kwargs): return True def delete(self, obj_id, *args, **kwargs): return True def get_stream(self, obj_id, chunk_size=DEFAULT_CHUNK_SIZE): data = io.BytesIO(next(self.content_generator)) reader = functools.partial(data.read, chunk_size) yield from iter(reader, b"") def list_content(self, last_obj_id=None, limit=DEFAULT_LIMIT): it = iter(self) if last_obj_id: next(it) it.send(int(last_obj_id)) return islice(it, limit) diff --git a/swh/objstorage/backends/libcloud.py b/swh/objstorage/backends/libcloud.py index 6b50e78..a908d8d 100644 --- a/swh/objstorage/backends/libcloud.py +++ b/swh/objstorage/backends/libcloud.py @@ -1,259 +1,255 @@ # Copyright (C) 2016-2017 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc from collections import OrderedDict from collections.abc import Iterator from typing import Optional from urllib.parse import urlencode from libcloud.storage import providers import libcloud.storage.drivers.s3 from libcloud.storage.types import ObjectDoesNotExistError, Provider from swh.model import hashutil from swh.objstorage.exc import Error, ObjNotFoundError from swh.objstorage.objstorage import ( ObjStorage, compressors, compute_hash, decompressors, ) def patch_libcloud_s3_urlencode(): """Patches libcloud's S3 backend to properly sign queries. Recent versions of libcloud are not affected (they use signature V4), but 1.5.0 (the one in Debian 9) is.""" def s3_urlencode(params): """Like urllib.parse.urlencode, but sorts the parameters first. This is required to properly compute the request signature, see https://docs.aws.amazon.com/AmazonS3/latest/dev/RESTAuthentication.html#ConstructingTheCanonicalizedResourceElement """ # noqa return urlencode(OrderedDict(sorted(params.items()))) libcloud.storage.drivers.s3.urlencode = s3_urlencode patch_libcloud_s3_urlencode() class CloudObjStorage(ObjStorage, metaclass=abc.ABCMeta): """Abstract ObjStorage that connect to a cloud using Libcloud Implementations of this class must redefine the _get_provider method to make it return a driver provider (i.e. object that supports `get_driver` method) which return a LibCloud driver (see https://libcloud.readthedocs.io/en/latest/storage/api.html). Args: container_name: Name of the base container path_prefix: prefix to prepend to object paths in the container, separated with a slash compression: compression algorithm to use for objects kwargs: extra arguments are passed through to the LibCloud driver """ def __init__( self, container_name: str, compression: Optional[str] = None, path_prefix: Optional[str] = None, **kwargs, ): super().__init__(**kwargs) self.driver = self._get_driver(**kwargs) self.container_name = container_name self.container = self.driver.get_container(container_name=container_name) self.compression = compression self.path_prefix = None if path_prefix: self.path_prefix = path_prefix.rstrip("/") + "/" def _get_driver(self, **kwargs): """Initialize a driver to communicate with the cloud Kwargs: arguments passed to the StorageDriver class, typically key: key to connect to the API. secret: secret key for authentication. secure: (bool) support HTTPS host: (str) port: (int) api_version: (str) region: (str) Returns: a Libcloud driver to a cloud storage. """ # Get the driver class from its description. cls = providers.get_driver(self._get_provider()) # Initialize the driver. return cls(**kwargs) @abc.abstractmethod def _get_provider(self): """Get a libcloud driver provider This method must be overridden by subclasses to specify which of the native libcloud driver the current storage should connect to. Alternatively, provider for a custom driver may be returned, in which case the provider will have to support `get_driver` method. """ raise NotImplementedError( "%s must implement `get_provider` method" % type(self) ) def check_config(self, *, check_write): """Check the configuration for this object storage""" # FIXME: hopefully this blew up during instantiation return True def __contains__(self, obj_id): try: self._get_object(obj_id) except ObjNotFoundError: return False else: return True def __iter__(self): - """ Iterate over the objects present in the storage + """Iterate over the objects present in the storage Warning: Iteration over the contents of a cloud-based object storage may have bad efficiency: due to the very high amount of objects in it and the fact that it is remote, get all the contents of the current object storage may result in a lot of network requests. You almost certainly don't want to use this method in production. """ for obj in self.driver.iterate_container_objects(self.container): name = obj.name if self.path_prefix and not name.startswith(self.path_prefix): continue if self.path_prefix: name = name[len(self.path_prefix) :] yield hashutil.hash_to_bytes(name) def __len__(self): """Compute the number of objects in the current object storage. Warning: this currently uses `__iter__`, its warning about bad performance applies. Returns: number of objects contained in the storage. """ return sum(1 for i in self) def add(self, content, obj_id=None, check_presence=True): if obj_id is None: # Checksum is missing, compute it on the fly. obj_id = compute_hash(content) if check_presence and obj_id in self: return obj_id self._put_object(content, obj_id) return obj_id def restore(self, content, obj_id=None): return self.add(content, obj_id, check_presence=False) def get(self, obj_id): obj = b"".join(self._get_object(obj_id).as_stream()) d = decompressors[self.compression]() ret = d.decompress(obj) if d.unused_data: hex_obj_id = hashutil.hash_to_hex(obj_id) raise Error("Corrupt object %s: trailing data found" % hex_obj_id) return ret def check(self, obj_id): # Check that the file exists, as _get_object raises ObjNotFoundError self._get_object(obj_id) # Check the content integrity obj_content = self.get(obj_id) content_obj_id = compute_hash(obj_content) if content_obj_id != obj_id: raise Error(obj_id) def delete(self, obj_id): super().delete(obj_id) # Check delete permission obj = self._get_object(obj_id) return self.driver.delete_object(obj) def _object_path(self, obj_id): """Get the full path to an object""" hex_obj_id = hashutil.hash_to_hex(obj_id) if self.path_prefix: return self.path_prefix + hex_obj_id else: return hex_obj_id def _get_object(self, obj_id): """Get a Libcloud wrapper for an object pointer. This wrapper does not retrieve the content of the object directly. """ object_path = self._object_path(obj_id) try: return self.driver.get_object(self.container_name, object_path) except ObjectDoesNotExistError: raise ObjNotFoundError(obj_id) def _compressor(self, data): comp = compressors[self.compression]() for chunk in data: cchunk = comp.compress(chunk) if cchunk: yield cchunk trail = comp.flush() if trail: yield trail def _put_object(self, content, obj_id): """Create an object in the cloud storage. Created object will contain the content and be referenced by the given id. """ object_path = self._object_path(obj_id) if not isinstance(content, Iterator): content = (content,) self.driver.upload_object_via_stream( self._compressor(content), self.container, object_path ) class AwsCloudObjStorage(CloudObjStorage): - """ Amazon's S3 Cloud-based object storage - - """ + """Amazon's S3 Cloud-based object storage""" def _get_provider(self): return Provider.S3 class OpenStackCloudObjStorage(CloudObjStorage): - """ OpenStack Swift Cloud based object storage - - """ + """OpenStack Swift Cloud based object storage""" def _get_provider(self): return Provider.OPENSTACK_SWIFT diff --git a/swh/objstorage/backends/pathslicing.py b/swh/objstorage/backends/pathslicing.py index 65de5e0..a49cfa5 100644 --- a/swh/objstorage/backends/pathslicing.py +++ b/swh/objstorage/backends/pathslicing.py @@ -1,445 +1,449 @@ # Copyright (C) 2015-2019 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from collections.abc import Iterator from contextlib import contextmanager from itertools import islice import os import random import tempfile from typing import List from swh.model import hashutil from swh.objstorage.exc import Error, ObjNotFoundError from swh.objstorage.objstorage import ( DEFAULT_CHUNK_SIZE, DEFAULT_LIMIT, ID_HASH_ALGO, ID_HEXDIGEST_LENGTH, ObjStorage, compressors, compute_hash, decompressors, ) BUFSIZ = 1048576 DIR_MODE = 0o755 FILE_MODE = 0o644 class PathSlicer: """Helper class to compute a path based on a hash. Used to compute a directory path based on the object hash according to a given slicing. Each slicing correspond to a directory that is named according to the hash of its content. For instance a file with SHA1 34973274ccef6ab4dfaaf86599792fa9c3fe4689 will have the following computed path: - 0:2/2:4/4:6 : 34/97/32/34973274ccef6ab4dfaaf86599792fa9c3fe4689 - 0:1/0:5/ : 3/34973/34973274ccef6ab4dfaaf86599792fa9c3fe4689 Args: root (str): path to the root directory of the storage on the disk. slicing (str): the slicing configuration. """ def __init__(self, root: str, slicing: str): self.root = root # Make a list of tuples where each tuple contains the beginning # and the end of each slicing. try: self.bounds = [ slice(*(int(x) if x else None for x in sbounds.split(":"))) for sbounds in slicing.split("/") if sbounds ] except TypeError: raise ValueError( "Invalid slicing declaration; " "it should be a of the form ':[/:]..." ) def check_config(self): """Check the slicing configuration is valid. Raises: ValueError: if the slicing configuration is invalid. """ if len(self): max_char = max( max(bound.start or 0, bound.stop or 0) for bound in self.bounds ) if ID_HEXDIGEST_LENGTH < max_char: raise ValueError( "Algorithm %s has too short hash for slicing to char %d" % (ID_HASH_ALGO, max_char) ) def get_directory(self, hex_obj_id: str) -> str: - """ Compute the storage directory of an object. + """Compute the storage directory of an object. See also: PathSlicer::get_path Args: hex_obj_id: object id as hexlified string. Returns: Absolute path (including root) to the directory that contains the given object id. """ return os.path.join(self.root, *self.get_slices(hex_obj_id)) def get_path(self, hex_obj_id: str) -> str: - """ Compute the full path to an object into the current storage. + """Compute the full path to an object into the current storage. See also: PathSlicer::get_directory Args: hex_obj_id(str): object id as hexlified string. Returns: Absolute path (including root) to the object corresponding to the given object id. """ return os.path.join(self.get_directory(hex_obj_id), hex_obj_id) def get_slices(self, hex_obj_id: str) -> List[str]: """Compute the path elements for the given hash. Args: hex_obj_id(str): object id as hexlified string. Returns: Relative path to the actual object corresponding to the given id as a list. """ assert len(hex_obj_id) == ID_HEXDIGEST_LENGTH return [hex_obj_id[bound] for bound in self.bounds] def __len__(self) -> int: """Number of slices of the slicer""" return len(self.bounds) class PathSlicingObjStorage(ObjStorage): """Implementation of the ObjStorage API based on the hash of the content. On disk, an object storage is a directory tree containing files named after their object IDs. An object ID is a checksum of its content, depending on the value of the ID_HASH_ALGO constant (see swh.model.hashutil for its meaning). To avoid directories that contain too many files, the object storage has a given slicing. Each slicing correspond to a directory that is named according to the hash of its content. So for instance a file with SHA1 34973274ccef6ab4dfaaf86599792fa9c3fe4689 will be stored in the given object storages : - 0:2/2:4/4:6 : 34/97/32/34973274ccef6ab4dfaaf86599792fa9c3fe4689 - 0:1/0:5/ : 3/34973/34973274ccef6ab4dfaaf86599792fa9c3fe4689 The files in the storage are stored in gzipped compressed format. Args: root (str): path to the root directory of the storage on the disk. slicing (str): string that indicates the slicing to perform on the hash of the content to know the path where it should be stored (see the documentation of the PathSlicer class). """ def __init__(self, root, slicing, compression="gzip", **kwargs): super().__init__(**kwargs) self.root = root self.slicer = PathSlicer(root, slicing) self.use_fdatasync = hasattr(os, "fdatasync") self.compression = compression self.check_config(check_write=False) def check_config(self, *, check_write): """Check whether this object storage is properly configured""" self.slicer.check_config() if not os.path.isdir(self.root): raise ValueError( 'PathSlicingObjStorage root "%s" is not a directory' % self.root ) if check_write: if not os.access(self.root, os.W_OK): raise PermissionError( 'PathSlicingObjStorage root "%s" is not writable' % self.root ) if self.compression not in compressors: raise ValueError( 'Unknown compression algorithm "%s" for ' "PathSlicingObjStorage" % self.compression ) return True def __contains__(self, obj_id): hex_obj_id = hashutil.hash_to_hex(obj_id) return os.path.isfile(self.slicer.get_path(hex_obj_id)) def __iter__(self): """Iterate over the object identifiers currently available in the storage. Warning: with the current implementation of the object storage, this method will walk the filesystem to list objects, meaning that listing all objects will be very slow for large storages. You almost certainly don't want to use this method in production. Return: Iterator over object IDs """ def obj_iterator(): # XXX hackish: it does not verify that the depth of found files # matches the slicing depth of the storage for root, _dirs, files in os.walk(self.root): _dirs.sort() for f in sorted(files): yield bytes.fromhex(f) return obj_iterator() def __len__(self): """Compute the number of objects available in the storage. Warning: this currently uses `__iter__`, its warning about bad performances applies Return: number of objects contained in the storage """ return sum(1 for i in self) def add(self, content, obj_id=None, check_presence=True): if obj_id is None: obj_id = compute_hash(content) if check_presence and obj_id in self: # If the object is already present, return immediately. return obj_id hex_obj_id = hashutil.hash_to_hex(obj_id) if not isinstance(content, Iterator): content = [content] compressor = compressors[self.compression]() with self._write_obj_file(hex_obj_id) as f: for chunk in content: f.write(compressor.compress(chunk)) f.write(compressor.flush()) return obj_id def get(self, obj_id): if obj_id not in self: raise ObjNotFoundError(obj_id) # Open the file and return its content as bytes hex_obj_id = hashutil.hash_to_hex(obj_id) d = decompressors[self.compression]() with open(self.slicer.get_path(hex_obj_id), "rb") as f: out = d.decompress(f.read()) if d.unused_data: - raise Error("Corrupt object %s: trailing data found" % hex_obj_id,) + raise Error( + "Corrupt object %s: trailing data found" % hex_obj_id, + ) return out def check(self, obj_id): try: data = self.get(obj_id) except OSError: hex_obj_id = hashutil.hash_to_hex(obj_id) - raise Error("Corrupt object %s: not a proper compressed file" % hex_obj_id,) + raise Error( + "Corrupt object %s: not a proper compressed file" % hex_obj_id, + ) checksums = hashutil.MultiHash.from_data( data, hash_names=[ID_HASH_ALGO] ).digest() actual_obj_id = checksums[ID_HASH_ALGO] hex_obj_id = hashutil.hash_to_hex(obj_id) if hex_obj_id != hashutil.hash_to_hex(actual_obj_id): raise Error( "Corrupt object %s should have id %s" % (hashutil.hash_to_hex(obj_id), hashutil.hash_to_hex(actual_obj_id)) ) def delete(self, obj_id): super().delete(obj_id) # Check delete permission if obj_id not in self: raise ObjNotFoundError(obj_id) hex_obj_id = hashutil.hash_to_hex(obj_id) try: os.remove(self.slicer.get_path(hex_obj_id)) except FileNotFoundError: raise ObjNotFoundError(obj_id) return True # Management methods def get_random(self, batch_size): def get_random_content(self, batch_size): - """ Get a batch of content inside a single directory. + """Get a batch of content inside a single directory. Returns: a tuple (batch size, batch). """ dirs = [] for level in range(len(self.slicer)): path = os.path.join(self.root, *dirs) dir_list = next(os.walk(path))[1] if "tmp" in dir_list: dir_list.remove("tmp") dirs.append(random.choice(dir_list)) path = os.path.join(self.root, *dirs) content_list = next(os.walk(path))[2] length = min(batch_size, len(content_list)) return ( length, map(hashutil.hash_to_bytes, random.sample(content_list, length)), ) while batch_size: length, it = get_random_content(self, batch_size) batch_size = batch_size - length yield from it # Streaming methods @contextmanager def chunk_writer(self, obj_id): hex_obj_id = hashutil.hash_to_hex(obj_id) compressor = compressors[self.compression]() with self._write_obj_file(hex_obj_id) as f: yield lambda c: f.write(compressor.compress(c)) f.write(compressor.flush()) def add_stream(self, content_iter, obj_id, check_presence=True): """Add a new object to the object storage using streaming. This function is identical to add() except it takes a generator that yields the chunked content instead of the whole content at once. Args: content (bytes): chunked generator that yields the object's raw content to add in storage. obj_id (bytes): object identifier check_presence (bool): indicate if the presence of the content should be verified before adding the file. Returns: the id (bytes) of the object into the storage. """ if check_presence and obj_id in self: return obj_id with self.chunk_writer(obj_id) as writer: for chunk in content_iter: writer(chunk) return obj_id def get_stream(self, obj_id, chunk_size=DEFAULT_CHUNK_SIZE): if obj_id not in self: raise ObjNotFoundError(obj_id) hex_obj_id = hashutil.hash_to_hex(obj_id) decompressor = decompressors[self.compression]() with open(self.slicer.get_path(hex_obj_id), "rb") as f: while True: raw = f.read(chunk_size) if not raw: break r = decompressor.decompress(raw) if not r: continue yield r def list_content(self, last_obj_id=None, limit=DEFAULT_LIMIT): if last_obj_id: it = self.iter_from(last_obj_id) else: it = iter(self) return islice(it, limit) def iter_from(self, obj_id, n_leaf=False): hex_obj_id = hashutil.hash_to_hex(obj_id) slices = self.slicer.get_slices(hex_obj_id) rlen = len(self.root.split("/")) i = 0 for root, dirs, files in os.walk(self.root): if not dirs: i += 1 level = len(root.split("/")) - rlen dirs.sort() if dirs and root == os.path.join(self.root, *slices[:level]): cslice = slices[level] for d in dirs[:]: if d < cslice: dirs.remove(d) for f in sorted(files): if f > hex_obj_id: yield bytes.fromhex(f) if n_leaf: yield i @contextmanager def _write_obj_file(self, hex_obj_id): - """ Context manager for writing object files to the object storage. + """Context manager for writing object files to the object storage. During writing, data are written to a temporary file, which is atomically renamed to the right file name after closing. Usage sample: with objstorage._write_obj_file(hex_obj_id): f.write(obj_data) Yields: a file-like object open for writing bytes. """ # Get the final paths and create the directory if absent. dir = self.slicer.get_directory(hex_obj_id) if not os.path.isdir(dir): os.makedirs(dir, DIR_MODE, exist_ok=True) path = os.path.join(dir, hex_obj_id) # Create a temporary file. (tmp, tmp_path) = tempfile.mkstemp(suffix=".tmp", prefix="hex_obj_id.", dir=dir) # Open the file and yield it for writing. tmp_f = os.fdopen(tmp, "wb") yield tmp_f # Make sure the contents of the temporary file are written to disk tmp_f.flush() if self.use_fdatasync: os.fdatasync(tmp) else: os.fsync(tmp) # Then close the temporary file and move it to the right path. tmp_f.close() os.chmod(tmp_path, FILE_MODE) os.rename(tmp_path, path) diff --git a/swh/objstorage/backends/seaweedfs/objstorage.py b/swh/objstorage/backends/seaweedfs/objstorage.py index fa1efb1..24e14e2 100644 --- a/swh/objstorage/backends/seaweedfs/objstorage.py +++ b/swh/objstorage/backends/seaweedfs/objstorage.py @@ -1,156 +1,156 @@ # Copyright (C) 2019-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import io from itertools import islice import logging import os from swh.model import hashutil from swh.objstorage.exc import Error, ObjNotFoundError from swh.objstorage.objstorage import ( DEFAULT_LIMIT, ObjStorage, compressors, compute_hash, decompressors, ) from .http import HttpFiler LOGGER = logging.getLogger(__name__) class SeaweedFilerObjStorage(ObjStorage): """ObjStorage with seaweedfs abilities, using the Filer API. https://github.com/chrislusf/seaweedfs/wiki/Filer-Server-API """ def __init__(self, url, compression=None, **kwargs): super().__init__(**kwargs) self.wf = HttpFiler(url) self.compression = compression def check_config(self, *, check_write): """Check the configuration for this object storage""" # FIXME: hopefully this blew up during instantiation return True def __contains__(self, obj_id): return self.wf.exists(self._path(obj_id)) def __iter__(self): - """ Iterate over the objects present in the storage + """Iterate over the objects present in the storage Warning: Iteration over the contents of a cloud-based object storage may have bad efficiency: due to the very high amount of objects in it and the fact that it is remote, get all the contents of the current object storage may result in a lot of network requests. You almost certainly don't want to use this method in production. """ obj_id = last_obj_id = None while True: for obj_id in self.list_content(last_obj_id=last_obj_id): yield obj_id if last_obj_id == obj_id: break last_obj_id = obj_id def __len__(self): """Compute the number of objects in the current object storage. Warning: this currently uses `__iter__`, its warning about bad performance applies. Returns: number of objects contained in the storage. """ return sum(1 for i in self) def add(self, content, obj_id=None, check_presence=True): if obj_id is None: # Checksum is missing, compute it on the fly. obj_id = compute_hash(content) if check_presence and obj_id in self: return obj_id def compressor(data): comp = compressors[self.compression]() for chunk in data: yield comp.compress(chunk) yield comp.flush() if isinstance(content, bytes): content = [content] # XXX should handle streaming correctly... self.wf.put(io.BytesIO(b"".join(compressor(content))), self._path(obj_id)) return obj_id def restore(self, content, obj_id=None): return self.add(content, obj_id, check_presence=False) def get(self, obj_id): try: obj = self.wf.get(self._path(obj_id)) except Exception: raise ObjNotFoundError(obj_id) d = decompressors[self.compression]() ret = d.decompress(obj) if d.unused_data: hex_obj_id = hashutil.hash_to_hex(obj_id) raise Error("Corrupt object %s: trailing data found" % hex_obj_id) return ret def check(self, obj_id): # Check the content integrity obj_content = self.get(obj_id) content_obj_id = compute_hash(obj_content) if content_obj_id != obj_id: raise Error(obj_id) def delete(self, obj_id): super().delete(obj_id) # Check delete permission if obj_id not in self: raise ObjNotFoundError(obj_id) self.wf.delete(self._path(obj_id)) return True def list_content(self, last_obj_id=None, limit=DEFAULT_LIMIT): if last_obj_id: objid = hashutil.hash_to_hex(last_obj_id) lastfilename = objid else: lastfilename = None for fname in islice(self.wf.iterfiles(last_file_name=lastfilename), limit): bytehex = fname.rsplit("/", 1)[-1] yield hashutil.bytehex_to_hash(bytehex.encode()) # internal methods def _put_object(self, content, obj_id): """Create an object in the cloud storage. Created object will contain the content and be referenced by the given id. """ def compressor(data): comp = compressors[self.compression]() for chunk in data: yield comp.compress(chunk) yield comp.flush() if isinstance(content, bytes): content = [content] self.wf.put(io.BytesIO(b"".join(compressor(content))), self._path(obj_id)) def _path(self, obj_id): return os.path.join(self.wf.basepath, hashutil.hash_to_hex(obj_id)) diff --git a/swh/objstorage/backends/winery/objstorage.py b/swh/objstorage/backends/winery/objstorage.py index 4a9474a..ade9a2d 100644 --- a/swh/objstorage/backends/winery/objstorage.py +++ b/swh/objstorage/backends/winery/objstorage.py @@ -1,185 +1,192 @@ # Copyright (C) 2022 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import logging from multiprocessing import Process from swh.model import hashutil from swh.objstorage import exc from swh.objstorage.objstorage import ObjStorage from .roshard import ROShard from .rwshard import RWShard from .sharedbase import SharedBase from .stats import Stats logger = logging.getLogger(__name__) def compute_hash(content): algo = "sha256" - return hashutil.MultiHash.from_data(content, hash_names=[algo],).digest().get(algo) + return ( + hashutil.MultiHash.from_data( + content, + hash_names=[algo], + ) + .digest() + .get(algo) + ) class WineryObjStorage(ObjStorage): def __init__(self, **kwargs): super().__init__(**kwargs) if kwargs.get("readonly"): self.winery = WineryReader(**kwargs) else: self.winery = WineryWriter(**kwargs) def uninit(self): self.winery.uninit() def get(self, obj_id): return self.winery.get(obj_id) def check_config(self, *, check_write): return True def __contains__(self, obj_id): return obj_id in self.winery def add(self, content, obj_id=None, check_presence=True): return self.winery.add(content, obj_id, check_presence) def check(self, obj_id): return self.winery.check(obj_id) def delete(self, obj_id): raise PermissionError("Delete is not allowed.") class WineryBase: def __init__(self, **kwargs): self.args = kwargs self.init() def init(self): self.base = SharedBase(**self.args) def uninit(self): self.base.uninit() def __contains__(self, obj_id): return self.base.contains(obj_id) class WineryReader(WineryBase): def __init__(self, **kwargs): super().__init__(**kwargs) self.shards = {} def roshard(self, name): if name not in self.shards: shard = ROShard(name, **self.args) shard.load() self.shards[name] = shard return self.shards[name] def get(self, obj_id): shard_info = self.base.get(obj_id) if shard_info is None: raise exc.ObjNotFoundError(obj_id) name, readonly = shard_info if readonly: shard = self.roshard(name) content = shard.get(obj_id) del shard else: shard = RWShard(name, **self.args) content = shard.get(obj_id) if content is None: raise exc.ObjNotFoundError(obj_id) return content def pack(shard, **kwargs): return Packer(shard, **kwargs).run() class Packer: def __init__(self, shard, **kwargs): self.stats = Stats(kwargs.get("output_dir")) self.args = kwargs self.shard = shard self.init() def init(self): self.rw = RWShard(self.shard, **self.args) self.ro = ROShard(self.shard, **self.args) def uninit(self): del self.ro self.rw.uninit() def run(self): self.ro.create(self.rw.count()) for obj_id, content in self.rw.all(): self.ro.add(content, obj_id) if self.stats.stats_active: self.stats.stats_read(obj_id, content) self.stats.stats_write(obj_id, content) self.ro.save() base = SharedBase(**self.args) base.shard_packing_ends(self.shard) base.uninit() self.rw.uninit() self.rw.drop() return True class WineryWriter(WineryReader): def __init__(self, **kwargs): super().__init__(**kwargs) self.packers = [] self.init() def init(self): super().init() self.shard = RWShard(self.base.whoami, **self.args) def uninit(self): self.shard.uninit() super().uninit() def add(self, content, obj_id=None, check_presence=True): if obj_id is None: obj_id = compute_hash(content) if check_presence and obj_id in self: return obj_id shard = self.base.add_phase_1(obj_id) if shard != self.base.id: # this object is the responsibility of another shard return obj_id self.shard.add(obj_id, content) self.base.add_phase_2(obj_id) if self.shard.is_full(): self.pack() return obj_id def check(self, obj_id): # load all shards packing == True and not locked (i.e. packer # was interrupted for whatever reason) run pack for each of them pass def pack(self): self.base.shard_packing_starts() p = Process(target=pack, args=(self.shard.name,), kwargs=self.args) self.uninit() p.start() self.packers.append(p) self.init() def __del__(self): for p in self.packers: p.kill() p.join() diff --git a/swh/objstorage/cli.py b/swh/objstorage/cli.py index 33c3933..c27a532 100644 --- a/swh/objstorage/cli.py +++ b/swh/objstorage/cli.py @@ -1,135 +1,135 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import logging # WARNING: do not import unnecessary things here to keep cli startup time under # control import os import time import click from swh.core.cli import CONTEXT_SETTINGS from swh.core.cli import swh as swh_cli_group @swh_cli_group.group(name="objstorage", context_settings=CONTEXT_SETTINGS) @click.option( "--config-file", "-C", default=None, - type=click.Path(exists=True, dir_okay=False,), + type=click.Path( + exists=True, + dir_okay=False, + ), help="Configuration file.", ) @click.pass_context def objstorage_cli_group(ctx, config_file): - """Software Heritage Objstorage tools. - """ + """Software Heritage Objstorage tools.""" from swh.core import config if not config_file: config_file = os.environ.get("SWH_CONFIG_FILENAME") if config_file: if not os.path.exists(config_file): raise ValueError("%s does not exist" % config_file) conf = config.read(config_file) else: conf = {} ctx.ensure_object(dict) ctx.obj["config"] = conf # for BW compat cli = objstorage_cli_group @objstorage_cli_group.command("rpc-serve") @click.option( "--host", default="0.0.0.0", metavar="IP", show_default=True, help="Host ip address to bind the server on", ) @click.option( "--port", "-p", default=5003, type=click.INT, metavar="PORT", show_default=True, help="Binding port of the server", ) @click.option( "--debug/--no-debug", default=True, help="Indicates if the server should run in debug mode", ) @click.pass_context def serve(ctx, host, port, debug): """Run a standalone objstorage server. This is not meant to be run on production systems. """ from swh.objstorage.api.server import app, validate_config if "log_level" in ctx.obj: logging.getLogger("werkzeug").setLevel(ctx.obj["log_level"]) validate_config(ctx.obj["config"]) app.config.update(ctx.obj["config"]) app.run(host, port=int(port), debug=bool(debug)) @objstorage_cli_group.command("import") @click.argument("directory", required=True, nargs=-1) @click.pass_context def import_directories(ctx, directory): - """Import a local directory in an existing objstorage. - """ + """Import a local directory in an existing objstorage.""" from swh.objstorage.factory import get_objstorage objstorage = get_objstorage(**ctx.obj["config"]["objstorage"]) nobj = 0 volume = 0 t0 = time.time() for dirname in directory: for root, _dirs, files in os.walk(dirname): for name in files: path = os.path.join(root, name) with open(path, "rb") as f: objstorage.add(f.read()) volume += os.stat(path).st_size nobj += 1 click.echo( "Imported %d files for a volume of %s bytes in %d seconds" % (nobj, volume, time.time() - t0) ) @objstorage_cli_group.command("fsck") @click.pass_context def fsck(ctx): - """Check the objstorage is not corrupted. - """ + """Check the objstorage is not corrupted.""" from swh.objstorage.factory import get_objstorage objstorage = get_objstorage(**ctx.obj["config"]["objstorage"]) for obj_id in objstorage: try: objstorage.check(obj_id) except objstorage.Error as err: logging.error(err) def main(): return cli(auto_envvar_prefix="SWH_OBJSTORAGE") if __name__ == "__main__": main() diff --git a/swh/objstorage/exc.py b/swh/objstorage/exc.py index 59bc9b1..a8abb94 100644 --- a/swh/objstorage/exc.py +++ b/swh/objstorage/exc.py @@ -1,41 +1,40 @@ # Copyright (C) 2015 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information class Error(Exception): def __str__(self): return "storage error on object: %s" % self.args class ObjNotFoundError(Error): def __str__(self): return "object not found: %s" % self.args class ObjStorageAPIError(Exception): - """ Specific internal exception of an object storage (mainly connection). - """ + """Specific internal exception of an object storage (mainly connection).""" def __str__(self): args = self.args return "An unexpected error occurred in the api backend: %s" % args class ReadOnlyObjStorage(Error): def __init__(self, method, *args): super().__init__(*args) self.method = method def __str__(self): return "This object storage is Read-Only: cannot use %s" % self.method class NonIterableObjStorage(Error): def __init__(self, method, *args): super().__init__(*args) self.method = method def __str__(self): return "This object storage is not iterable: cannot use %s" % self.method diff --git a/swh/objstorage/factory.py b/swh/objstorage/factory.py index 3ccaed0..dfc017b 100644 --- a/swh/objstorage/factory.py +++ b/swh/objstorage/factory.py @@ -1,129 +1,129 @@ # Copyright (C) 2016-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from typing import Callable, Dict, Union import warnings from swh.objstorage.api.client import RemoteObjStorage from swh.objstorage.backends.generator import RandomGeneratorObjStorage from swh.objstorage.backends.http import HTTPReadOnlyObjStorage from swh.objstorage.backends.in_memory import InMemoryObjStorage from swh.objstorage.backends.noop import NoopObjStorage from swh.objstorage.backends.pathslicing import PathSlicingObjStorage from swh.objstorage.backends.seaweedfs import SeaweedFilerObjStorage from swh.objstorage.multiplexer import MultiplexerObjStorage, StripingObjStorage from swh.objstorage.multiplexer.filter import add_filters from swh.objstorage.objstorage import ID_HEXDIGEST_LENGTH, ObjStorage # noqa __all__ = ["get_objstorage", "ObjStorage"] _STORAGE_CLASSES: Dict[str, Union[type, Callable[..., type]]] = { "pathslicing": PathSlicingObjStorage, "remote": RemoteObjStorage, "memory": InMemoryObjStorage, "seaweedfs": SeaweedFilerObjStorage, "random": RandomGeneratorObjStorage, "http": HTTPReadOnlyObjStorage, "noop": NoopObjStorage, } _STORAGE_CLASSES_MISSING = {} _STORAGE_CLASSES_DEPRECATED = {"weed": "seaweedfs"} try: from swh.objstorage.backends.azure import ( AzureCloudObjStorage, PrefixedAzureCloudObjStorage, ) _STORAGE_CLASSES["azure"] = AzureCloudObjStorage _STORAGE_CLASSES["azure-prefixed"] = PrefixedAzureCloudObjStorage except ImportError as e: _STORAGE_CLASSES_MISSING["azure"] = e.args[0] _STORAGE_CLASSES_MISSING["azure-prefixed"] = e.args[0] try: from swh.objstorage.backends.libcloud import ( AwsCloudObjStorage, OpenStackCloudObjStorage, ) _STORAGE_CLASSES["s3"] = AwsCloudObjStorage _STORAGE_CLASSES["swift"] = OpenStackCloudObjStorage except ImportError as e: _STORAGE_CLASSES_MISSING["s3"] = e.args[0] _STORAGE_CLASSES_MISSING["swift"] = e.args[0] try: from swh.objstorage.backends.winery import WineryObjStorage _STORAGE_CLASSES["winery"] = WineryObjStorage except ImportError as e: _STORAGE_CLASSES_MISSING["winery"] = e.args[0] def get_objstorage(cls: str, args=None, **kwargs): - """ Create an ObjStorage using the given implementation class. + """Create an ObjStorage using the given implementation class. Args: cls: objstorage class unique key contained in the _STORAGE_CLASSES dict. kwargs: arguments for the required class of objstorage that must match exactly the one in the `__init__` method of the class. Returns: subclass of ObjStorage that match the given `storage_class` argument. Raises: ValueError: if the given storage class is not a valid objstorage key. """ if cls in _STORAGE_CLASSES_DEPRECATED: warnings.warn( f"{cls} objstorage class is deprecated, " f"use {_STORAGE_CLASSES_DEPRECATED[cls]} class instead.", DeprecationWarning, ) cls = _STORAGE_CLASSES_DEPRECATED[cls] if cls in _STORAGE_CLASSES: if args is not None: warnings.warn( 'Explicit "args" key is deprecated for objstorage initialization, ' "use class arguments keys directly instead.", DeprecationWarning, ) # TODO: when removing this, drop the "args" backwards compatibility # from swh.objstorage.api.server configuration checker kwargs = args return _STORAGE_CLASSES[cls](**kwargs) else: raise ValueError( "Storage class {} is not available: {}".format( cls, _STORAGE_CLASSES_MISSING.get(cls, "unknown name") ) ) def _construct_filtered_objstorage(storage_conf, filters_conf): return add_filters(get_objstorage(**storage_conf), filters_conf) _STORAGE_CLASSES["filtered"] = _construct_filtered_objstorage def _construct_multiplexer_objstorage(objstorages): storages = [get_objstorage(**conf) for conf in objstorages] return MultiplexerObjStorage(storages) _STORAGE_CLASSES["multiplexer"] = _construct_multiplexer_objstorage def _construct_striping_objstorage(objstorages): storages = [get_objstorage(**conf) for conf in objstorages] return StripingObjStorage(storages) _STORAGE_CLASSES["striping"] = _construct_striping_objstorage diff --git a/swh/objstorage/interface.py b/swh/objstorage/interface.py index 23b37a8..b91c0b9 100644 --- a/swh/objstorage/interface.py +++ b/swh/objstorage/interface.py @@ -1,237 +1,237 @@ # Copyright (C) 2015-2022 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from typing import Dict from typing_extensions import Protocol, runtime_checkable from swh.core.api import remote_api_endpoint from swh.objstorage.objstorage import DEFAULT_CHUNK_SIZE, DEFAULT_LIMIT @runtime_checkable class ObjStorageInterface(Protocol): - """ High-level API to manipulate the Software Heritage object storage. + """High-level API to manipulate the Software Heritage object storage. Conceptually, the object storage offers the following methods: - check_config() check if the object storage is properly configured - __contains__() check if an object is present, by object id - add() add a new object, returning an object id - restore() same as add() but erase an already existed content - get() retrieve the content of an object, by object id - check() check the integrity of an object, by object id - delete() remove an object And some management methods: - get_random() get random object id of existing contents (used for the content integrity checker). Some of the methods have available streaming equivalents: - get_stream() same as get() but returns a chunked iterator Each implementation of this interface can have a different behavior and its own way to store the contents. """ @remote_api_endpoint("check_config") def check_config(self, *, check_write): """Check whether the object storage is properly configured. Args: check_write (bool): if True, check if writes to the object storage can succeed. Returns: True if the configuration check worked, an exception if it didn't. """ ... @remote_api_endpoint("content/contains") def __contains__(self, obj_id): """Indicate if the given object is present in the storage. Args: obj_id (bytes): object identifier. Returns: True if and only if the object is present in the current object storage. """ ... @remote_api_endpoint("content/add") def add(self, content, obj_id=None, check_presence=True): """Add a new object to the object storage. Args: content (bytes): object's raw content to add in storage. obj_id (bytes): checksum of [bytes] using [ID_HASH_ALGO] algorithm. When given, obj_id will be trusted to match the bytes. If missing, obj_id will be computed on the fly. check_presence (bool): indicate if the presence of the content should be verified before adding the file. Returns: the id (bytes) of the object into the storage. """ ... @remote_api_endpoint("content/add/batch") def add_batch(self, contents, check_presence=True) -> Dict: """Add a batch of new objects to the object storage. Args: contents: mapping from obj_id to object contents Returns: the summary of objects added to the storage (count of object, count of bytes object) """ ... def restore(self, content, obj_id=None): """Restore a content that have been corrupted. This function is identical to add but does not check if the object id is already in the file system. The default implementation provided by the current class is suitable for most cases. Args: content (bytes): object's raw content to add in storage obj_id (bytes): checksum of `bytes` as computed by ID_HASH_ALGO. When given, obj_id will be trusted to match bytes. If missing, obj_id will be computed on the fly. """ ... @remote_api_endpoint("content/get") def get(self, obj_id): """Retrieve the content of a given object. Args: obj_id (bytes): object id. Returns: the content of the requested object as bytes. Raises: ObjNotFoundError: if the requested object is missing. """ ... @remote_api_endpoint("content/get/batch") def get_batch(self, obj_ids): """Retrieve objects' raw content in bulk from storage. Note: This function does have a default implementation in ObjStorage that is suitable for most cases. For object storages that needs to do the minimal number of requests possible (ex: remote object storages), that method can be overridden to perform a more efficient operation. Args: obj_ids ([bytes]: list of object ids. Returns: list of resulting contents, or None if the content could not be retrieved. Do not raise any exception as a fail for one content will not cancel the whole request. """ ... @remote_api_endpoint("content/check") def check(self, obj_id): """Perform an integrity check for a given object. Verify that the file object is in place and that the content matches the object id. Args: obj_id (bytes): object identifier. Raises: ObjNotFoundError: if the requested object is missing. Error: if the request object is corrupted. """ ... @remote_api_endpoint("content/delete") def delete(self, obj_id): """Delete an object. Args: obj_id (bytes): object identifier. Raises: ObjNotFoundError: if the requested object is missing. """ ... # Management methods @remote_api_endpoint("content/get/random") def get_random(self, batch_size): """Get random ids of existing contents. This method is used in order to get random ids to perform content integrity verifications on random contents. Args: batch_size (int): Number of ids that will be given Yields: An iterable of ids (bytes) of contents that are in the current object storage. """ ... # Streaming methods def get_stream(self, obj_id, chunk_size=DEFAULT_CHUNK_SIZE): """Retrieve the content of a given object as a chunked iterator. Args: obj_id (bytes): object id. Returns: the content of the requested object as bytes. Raises: ObjNotFoundError: if the requested object is missing. """ ... def __iter__(self): ... def list_content(self, last_obj_id=None, limit=DEFAULT_LIMIT): """Generates known object ids. Args: last_obj_id (bytes): object id from which to iterate from (excluded). limit (int): max number of object ids to generate. Generates: obj_id (bytes): object ids. """ ... diff --git a/swh/objstorage/multiplexer/filter/__init__.py b/swh/objstorage/multiplexer/filter/__init__.py index a36e426..36fcd1f 100644 --- a/swh/objstorage/multiplexer/filter/__init__.py +++ b/swh/objstorage/multiplexer/filter/__init__.py @@ -1,93 +1,93 @@ from .id_filter import PrefixIdObjStorageFilter, RegexIdObjStorageFilter from .read_write_filter import ReadObjStorageFilter _FILTERS_CLASSES = { "readonly": ReadObjStorageFilter, "regex": RegexIdObjStorageFilter, "prefix": PrefixIdObjStorageFilter, } _FILTERS_PRIORITY = {"readonly": 0, "prefix": 1, "regex": 2} def read_only(): return {"type": "readonly"} def id_prefix(prefix): return {"type": "prefix", "prefix": prefix} def id_regex(regex): return {"type": "regex", "regex": regex} def _filter_priority(filter_type): """Get the priority of this filter. Priority is a value that indicates if the operation of the filter is time-consuming (smaller values means quick execution), or very likely to be almost always the same value (False being small, and True high). In case the filters are chained, they will be ordered in a way that small priorities (quick execution or instantly break the chain) are executed first. Default value is 1. Value 0 is recommended for storages that change behavior only by disabling some operations (making the method return None). """ return _FILTERS_PRIORITY.get(filter_type, 1) def add_filter(storage, filter_conf): """Add a filter to the given storage. Args: storage (swh.objstorage.ObjStorage): storage which will be filtered. filter_conf (dict): configuration of an ObjStorageFilter, given as a dictionary that contains the keys: - type: which represent the type of filter, one of the keys of _FILTERS_CLASSES - Every arguments that this type of filter requires. Returns: A filtered storage that perform only the valid operations. """ type = filter_conf["type"] args = {k: v for k, v in filter_conf.items() if k != "type"} filtered_storage = _FILTERS_CLASSES[type](storage=storage, **args) return filtered_storage def add_filters(storage, filter_confs): - """ Add multiple filters to the given storage. + """Add multiple filters to the given storage. (See filter.add_filter) Args: storage (swh.objstorage.ObjStorage): storage which will be filtered. filter_confs (list): any number of filter conf, as a dict with: - type: which represent the type of filter, one of the keys of FILTERS. - Every arguments that this type of filter require. Returns: A filtered storage that fulfill the requirement of all the given filters. """ # Reverse sorting in order to put the filter with biggest priority first. filter_confs.sort(key=lambda conf: _filter_priority(conf["type"]), reverse=True) # Add the bigest filter to the storage, and reduce it to accumulate filters # on top of it, until the smallest (fastest, see filter.filter_priority) is # added. for filter_conf in filter_confs: storage = add_filter(storage, filter_conf) return storage diff --git a/swh/objstorage/multiplexer/filter/filter.py b/swh/objstorage/multiplexer/filter/filter.py index 7363962..35ac628 100644 --- a/swh/objstorage/multiplexer/filter/filter.py +++ b/swh/objstorage/multiplexer/filter/filter.py @@ -1,77 +1,77 @@ # Copyright (C) 2015-2016 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from swh.objstorage.objstorage import ObjStorage class ObjStorageFilter(ObjStorage): """Base implementation of a filter that allow inputs on ObjStorage or not. This class copy the API of ...objstorage in order to filter the inputs of this class. If the operation is allowed, return the result of this operation applied to the destination implementation. Otherwise, just return without any operation. This class is an abstract base class for a classic read/write storage. Filters can inherit from it and only redefine some methods in order to change behavior. """ def __init__(self, storage): self.storage = storage def check_config(self, *, check_write): """Check the object storage for proper configuration. Args: check_write: check whether writes to the objstorage will succeed Returns: True if the storage is properly configured """ return self.storage.check_config(check_write=check_write) def __contains__(self, *args, **kwargs): return self.storage.__contains__(*args, **kwargs) def __iter__(self): - """ Iterates over the content of each storages + """Iterates over the content of each storages Warning: The `__iter__` methods frequently have bad performance. You almost certainly don't want to use this method in production as the wrapped storage may cause performance issues. """ return self.storage.__iter__() def __len__(self): - """ Compute the number of objects in the current object storage. + """Compute the number of objects in the current object storage. Warning: performance issue in `__iter__` also applies here. Returns: number of objects contained in the storage. """ return self.storage.__len__() def add(self, content, obj_id=None, check_presence=True, *args, **kwargs): return self.storage.add(content, obj_id, check_presence, *args, **kwargs) def restore(self, content, obj_id=None, *args, **kwargs): return self.storage.restore(content, obj_id, *args, **kwargs) def get(self, obj_id, *args, **kwargs): return self.storage.get(obj_id, *args, **kwargs) def check(self, obj_id, *args, **kwargs): return self.storage.check(obj_id, *args, **kwargs) def delete(self, obj_id, *args, **kwargs): return self.storage.delete(obj_id, *args, **kwargs) def get_random(self, batch_size, *args, **kwargs): return self.storage.get_random(batch_size, *args, **kwargs) diff --git a/swh/objstorage/multiplexer/filter/id_filter.py b/swh/objstorage/multiplexer/filter/id_filter.py index 47e6a91..5557b3e 100644 --- a/swh/objstorage/multiplexer/filter/id_filter.py +++ b/swh/objstorage/multiplexer/filter/id_filter.py @@ -1,92 +1,89 @@ # Copyright (C) 2015-2017 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc import re from swh.model import hashutil from swh.objstorage.exc import ObjNotFoundError from swh.objstorage.multiplexer.filter.filter import ObjStorageFilter from swh.objstorage.objstorage import compute_hash class IdObjStorageFilter(ObjStorageFilter, metaclass=abc.ABCMeta): - """ Filter that only allow operations if the object id match a requirement. + """Filter that only allow operations if the object id match a requirement. Even for read operations, check before if the id match the requirements. This may prevent for unnecessary disk access. """ @abc.abstractmethod def is_valid(self, obj_id): - """ Indicates if the given id is valid. - """ + """Indicates if the given id is valid.""" raise NotImplementedError( "Implementations of an IdObjStorageFilter " 'must have a "is_valid" method' ) def __contains__(self, obj_id, *args, **kwargs): if self.is_valid(obj_id): return self.storage.__contains__(*args, obj_id=obj_id, **kwargs) return False def __len__(self): return sum(1 for i in [id for id in self.storage if self.is_valid(id)]) def __iter__(self): yield from filter(lambda id: self.is_valid(id), iter(self.storage)) def add(self, content, obj_id=None, check_presence=True, *args, **kwargs): if obj_id is None: obj_id = compute_hash(content) if self.is_valid(obj_id): return self.storage.add(content, *args, obj_id=obj_id, **kwargs) def restore(self, content, obj_id=None, *args, **kwargs): if obj_id is None: obj_id = compute_hash(content) if self.is_valid(obj_id): return self.storage.restore(content, *args, obj_id=obj_id, **kwargs) def get(self, obj_id, *args, **kwargs): if self.is_valid(obj_id): return self.storage.get(*args, obj_id=obj_id, **kwargs) raise ObjNotFoundError(obj_id) def check(self, obj_id, *args, **kwargs): if self.is_valid(obj_id): return self.storage.check(*args, obj_id=obj_id, **kwargs) raise ObjNotFoundError(obj_id) def get_random(self, *args, **kwargs): yield from filter( lambda id: self.is_valid(id), self.storage.get_random(*args, **kwargs) ) class RegexIdObjStorageFilter(IdObjStorageFilter): - """ Filter that allow operations if the content's id as hex match a regex. - """ + """Filter that allow operations if the content's id as hex match a regex.""" def __init__(self, storage, regex): super().__init__(storage) self.regex = re.compile(regex) def is_valid(self, obj_id): hex_obj_id = hashutil.hash_to_hex(obj_id) return self.regex.match(hex_obj_id) is not None class PrefixIdObjStorageFilter(IdObjStorageFilter): - """ Filter that allow operations if the hexlified id have a given prefix. - """ + """Filter that allow operations if the hexlified id have a given prefix.""" def __init__(self, storage, prefix): super().__init__(storage) self.prefix = str(prefix) def is_valid(self, obj_id): hex_obj_id = hashutil.hash_to_hex(obj_id) return str(hex_obj_id).startswith(self.prefix) diff --git a/swh/objstorage/multiplexer/filter/read_write_filter.py b/swh/objstorage/multiplexer/filter/read_write_filter.py index 753ad67..281ddc3 100644 --- a/swh/objstorage/multiplexer/filter/read_write_filter.py +++ b/swh/objstorage/multiplexer/filter/read_write_filter.py @@ -1,25 +1,25 @@ # Copyright (C) 2015-2016 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from swh.objstorage.multiplexer.filter.filter import ObjStorageFilter class ReadObjStorageFilter(ObjStorageFilter): - """ Filter that disable write operation of the storage. + """Filter that disable write operation of the storage. Writes will always succeed without doing any actual write operations. """ def check_config(self, *, check_write): return self.storage.check_config(check_write=False) def add(self, *args, **kwargs): return def restore(self, *args, **kwargs): return def delete(self, *args, **kwargs): return True diff --git a/swh/objstorage/multiplexer/multiplexer_objstorage.py b/swh/objstorage/multiplexer/multiplexer_objstorage.py index ae86e06..73b2bd5 100644 --- a/swh/objstorage/multiplexer/multiplexer_objstorage.py +++ b/swh/objstorage/multiplexer/multiplexer_objstorage.py @@ -1,323 +1,330 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import queue import random import threading from typing import Dict from swh.objstorage.exc import ObjNotFoundError from swh.objstorage.objstorage import ObjStorage class ObjStorageThread(threading.Thread): def __init__(self, storage): super().__init__(daemon=True) self.storage = storage self.commands = queue.Queue() def run(self): while True: try: mailbox, command, args, kwargs = self.commands.get(True, 0.05) except queue.Empty: continue try: ret = getattr(self.storage, command)(*args, **kwargs) except Exception as exc: self.queue_result(mailbox, "exception", exc) else: self.queue_result(mailbox, "result", ret) def queue_command(self, command, *args, mailbox=None, **kwargs): """Enqueue a new command to be processed by the thread. Args: command (str): one of the method names for the underlying storage. mailbox (queue.Queue): explicit mailbox if the calling thread wants to override it. args, kwargs: arguments for the command. Returns: queue.Queue: The mailbox you can read the response from """ if not mailbox: mailbox = queue.Queue() self.commands.put((mailbox, command, args, kwargs)) return mailbox def queue_result(self, mailbox, result_type, result): """Enqueue a new result in the mailbox This also provides a reference to the storage, which can be useful when an exceptional condition arises. Args: mailbox (queue.Queue): the mailbox to which we need to enqueue the result result_type (str): one of 'result', 'exception' result: the result to pass back to the calling thread """ mailbox.put( - {"type": result_type, "result": result,} + { + "type": result_type, + "result": result, + } ) @staticmethod def get_result_from_mailbox(mailbox, *args, **kwargs): """Unpack the result from the mailbox. Args: mailbox (queue.Queue): A mailbox to unpack a result from args: positional arguments to :func:`mailbox.get` kwargs: keyword arguments to :func:`mailbox.get` Returns: the next result unpacked from the queue Raises: either the exception we got back from the underlying storage, or :exc:`queue.Empty` if :func:`mailbox.get` raises that. """ result = mailbox.get(*args, **kwargs) if result["type"] == "exception": raise result["result"] from None else: return result["result"] @staticmethod def collect_results(mailbox, num_results): """Collect num_results from the mailbox""" collected = 0 ret = [] while collected < num_results: try: ret.append( ObjStorageThread.get_result_from_mailbox(mailbox, True, 0.05) ) except queue.Empty: continue collected += 1 return ret def __getattr__(self, attr): def call(*args, **kwargs): mailbox = self.queue_command(attr, *args, **kwargs) return self.get_result_from_mailbox(mailbox) return call def __contains__(self, *args, **kwargs): mailbox = self.queue_command("__contains__", *args, **kwargs) return self.get_result_from_mailbox(mailbox) class MultiplexerObjStorage(ObjStorage): """Implementation of ObjStorage that distributes between multiple storages. The multiplexer object storage allows an input to be demultiplexed among multiple storages that will or will not accept it by themselves (see .filter package). As the ids can be different, no pre-computed ids should be submitted. Also, there are no guarantees that the returned ids can be used directly into the storages that the multiplexer manage. Use case examples follow. Example 1:: storage_v1 = filter.read_only(PathSlicingObjStorage('/dir1', '0:2/2:4/4:6')) storage_v2 = PathSlicingObjStorage('/dir2', '0:1/0:5') storage = MultiplexerObjStorage([storage_v1, storage_v2]) When using 'storage', all the new contents will only be added to the v2 storage, while it will be retrievable from both. Example 2:: storage_v1 = filter.id_regex( PathSlicingObjStorage('/dir1', '0:2/2:4/4:6'), r'[^012].*' ) storage_v2 = filter.if_regex( PathSlicingObjStorage('/dir2', '0:1/0:5'), r'[012]/*' ) storage = MultiplexerObjStorage([storage_v1, storage_v2]) When using this storage, the contents with a sha1 starting with 0, 1 or 2 will be redirected (read AND write) to the storage_v2, while the others will be redirected to the storage_v1. If a content starting with 0, 1 or 2 is present in the storage_v1, it would be ignored anyway. """ def __init__(self, storages, **kwargs): super().__init__(**kwargs) self.storages = storages self.storage_threads = [ObjStorageThread(storage) for storage in storages] for thread in self.storage_threads: thread.start() def wrap_call(self, threads, call, *args, **kwargs): threads = list(threads) mailbox = queue.Queue() for thread in threads: thread.queue_command(call, *args, mailbox=mailbox, **kwargs) return ObjStorageThread.collect_results(mailbox, len(threads)) def get_read_threads(self, obj_id=None): yield from self.storage_threads def get_write_threads(self, obj_id=None): yield from self.storage_threads def check_config(self, *, check_write): """Check whether the object storage is properly configured. Args: check_write (bool): if True, check if writes to the object storage can succeed. Returns: True if the configuration check worked, an exception if it didn't. """ return all( self.wrap_call( self.storage_threads, "check_config", check_write=check_write ) ) def __contains__(self, obj_id): """Indicate if the given object is present in the storage. Args: obj_id (bytes): object identifier. Returns: True if and only if the object is present in the current object storage. """ for storage in self.get_read_threads(obj_id): if obj_id in storage: return True return False def __iter__(self): def obj_iterator(): for storage in self.storages: yield from storage return obj_iterator() def add(self, content, obj_id=None, check_presence=True): - """ Add a new object to the object storage. + """Add a new object to the object storage. If the adding step works in all the storages that accept this content, this is a success. Otherwise, the full adding step is an error even if it succeed in some of the storages. Args: content: content of the object to be added to the storage. obj_id: checksum of [bytes] using [ID_HASH_ALGO] algorithm. When given, obj_id will be trusted to match the bytes. If missing, obj_id will be computed on the fly. check_presence: indicate if the presence of the content should be verified before adding the file. Returns: an id of the object into the storage. As the write-storages are always readable as well, any id will be valid to retrieve a content. """ results = self.wrap_call( self.get_write_threads(obj_id), "add", content, obj_id=obj_id, check_presence=check_presence, ) for result in results: if not result: continue return result def add_batch(self, contents, check_presence=True) -> Dict: - """Add a batch of new objects to the object storage. - - """ + """Add a batch of new objects to the object storage.""" write_threads = list(self.get_write_threads()) results = self.wrap_call( - write_threads, "add_batch", contents, check_presence=check_presence, + write_threads, + "add_batch", + contents, + check_presence=check_presence, ) summed = {"object:add": 0, "object:add:bytes": 0} for result in results: summed["object:add"] += result["object:add"] summed["object:add:bytes"] += result["object:add:bytes"] return { "object:add": summed["object:add"] // len(results), "object:add:bytes": summed["object:add:bytes"] // len(results), } def restore(self, content, obj_id=None): return self.wrap_call( - self.get_write_threads(obj_id), "restore", content, obj_id=obj_id, + self.get_write_threads(obj_id), + "restore", + content, + obj_id=obj_id, ).pop() def get(self, obj_id): for storage in self.get_read_threads(obj_id): try: return storage.get(obj_id) except ObjNotFoundError: continue # If no storage contains this content, raise the error raise ObjNotFoundError(obj_id) def check(self, obj_id): nb_present = 0 for storage in self.get_read_threads(obj_id): try: storage.check(obj_id) except ObjNotFoundError: continue else: nb_present += 1 # If there is an Error because of a corrupted file, then let it pass. # Raise the ObjNotFoundError only if the content couldn't be found in # all the storages. if nb_present == 0: raise ObjNotFoundError(obj_id) def delete(self, obj_id): super().delete(obj_id) # Check delete permission return all(self.wrap_call(self.get_write_threads(obj_id), "delete", obj_id)) def get_random(self, batch_size): storages_set = [storage for storage in self.storages if len(storage) > 0] if len(storages_set) <= 0: return [] while storages_set: storage = random.choice(storages_set) try: return storage.get_random(batch_size) except NotImplementedError: storages_set.remove(storage) # There is no storage that allow the get_random operation raise NotImplementedError( "There is no storage implementation into the multiplexer that " "support the 'get_random' operation" ) diff --git a/swh/objstorage/multiplexer/striping_objstorage.py b/swh/objstorage/multiplexer/striping_objstorage.py index 67e9538..3fa3ffa 100644 --- a/swh/objstorage/multiplexer/striping_objstorage.py +++ b/swh/objstorage/multiplexer/striping_objstorage.py @@ -1,75 +1,76 @@ # Copyright (C) 2018-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from collections import defaultdict import queue from typing import Dict from swh.objstorage.multiplexer.multiplexer_objstorage import ( MultiplexerObjStorage, ObjStorageThread, ) class StripingObjStorage(MultiplexerObjStorage): """Stripes objects across multiple objstorages This objstorage implementation will write objects to objstorages in a predictable way: it takes the modulo of the last 8 bytes of the object identifier with the number of object storages passed, which will yield an (almost) even distribution. Objects are read from all storages in turn until it succeeds. """ MOD_BYTES = 8 def __init__(self, storages, **kwargs): super().__init__(storages, **kwargs) self.num_storages = len(storages) def get_storage_index(self, obj_id): if obj_id is None: raise ValueError("StripingObjStorage always needs obj_id to be set") index = int.from_bytes(obj_id[: -self.MOD_BYTES], "big") return index % self.num_storages def get_write_threads(self, obj_id): idx = self.get_storage_index(obj_id) yield self.storage_threads[idx] def get_read_threads(self, obj_id=None): if obj_id: idx = self.get_storage_index(obj_id) else: idx = 0 for i in range(self.num_storages): yield self.storage_threads[(idx + i) % self.num_storages] def add_batch(self, contents, check_presence=True) -> Dict: - """Add a batch of new objects to the object storage. - - """ + """Add a batch of new objects to the object storage.""" content_by_storage_index: Dict[bytes, Dict] = defaultdict(dict) for obj_id, content in contents.items(): storage_index = self.get_storage_index(obj_id) content_by_storage_index[storage_index][obj_id] = content mailbox: queue.Queue[Dict] = queue.Queue() for storage_index, contents in content_by_storage_index.items(): self.storage_threads[storage_index].queue_command( - "add_batch", contents, check_presence=check_presence, mailbox=mailbox, + "add_batch", + contents, + check_presence=check_presence, + mailbox=mailbox, ) results = ObjStorageThread.collect_results( mailbox, len(content_by_storage_index) ) summed = {"object:add": 0, "object:add:bytes": 0} for result in results: summed["object:add"] += result["object:add"] summed["object:add:bytes"] += result["object:add:bytes"] return summed diff --git a/swh/objstorage/objstorage.py b/swh/objstorage/objstorage.py index c109086..d7e2883 100644 --- a/swh/objstorage/objstorage.py +++ b/swh/objstorage/objstorage.py @@ -1,153 +1,156 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import abc import bz2 from itertools import dropwhile, islice import lzma from typing import Dict import zlib from swh.model import hashutil from .exc import ObjNotFoundError ID_HASH_ALGO = "sha1" ID_HEXDIGEST_LENGTH = 40 """Size in bytes of the hash hexadecimal representation.""" ID_DIGEST_LENGTH = 20 """Size in bytes of the hash""" DEFAULT_CHUNK_SIZE = 2 * 1024 * 1024 """Size in bytes of the streaming chunks""" DEFAULT_LIMIT = 10000 """Default number of results of ``list_content``.""" def compute_hash(content): """Compute the content's hash. Args: content (bytes): The raw content to hash hash_name (str): Hash's name (default to ID_HASH_ALGO) Returns: The ID_HASH_ALGO for the content """ return ( - hashutil.MultiHash.from_data(content, hash_names=[ID_HASH_ALGO],) + hashutil.MultiHash.from_data( + content, + hash_names=[ID_HASH_ALGO], + ) .digest() .get(ID_HASH_ALGO) ) class NullCompressor: def compress(self, data): return data def flush(self): return b"" class NullDecompressor: def decompress(self, data): return data @property def unused_data(self): return b"" decompressors = { "bz2": bz2.BZ2Decompressor, "lzma": lzma.LZMADecompressor, "gzip": lambda: zlib.decompressobj(wbits=31), "zlib": zlib.decompressobj, "none": NullDecompressor, } compressors = { "bz2": bz2.BZ2Compressor, "lzma": lzma.LZMACompressor, "gzip": lambda: zlib.compressobj(wbits=31), "zlib": zlib.compressobj, "none": NullCompressor, } class ObjStorage(metaclass=abc.ABCMeta): def __init__(self, *, allow_delete=False, **kwargs): # A more complete permission system could be used in place of that if # it becomes needed self.allow_delete = allow_delete @abc.abstractmethod def check_config(self, *, check_write): pass @abc.abstractmethod def __contains__(self, obj_id): pass @abc.abstractmethod def add(self, content, obj_id=None, check_presence=True): pass def add_batch(self, contents, check_presence=True) -> Dict: summary = {"object:add": 0, "object:add:bytes": 0} for obj_id, content in contents.items(): if check_presence and obj_id in self: continue self.add(content, obj_id, check_presence=False) summary["object:add"] += 1 summary["object:add:bytes"] += len(content) return summary def restore(self, content, obj_id=None): # check_presence to false will erase the potential previous content. return self.add(content, obj_id, check_presence=False) @abc.abstractmethod def get(self, obj_id): pass def get_batch(self, obj_ids): for obj_id in obj_ids: try: yield self.get(obj_id) except ObjNotFoundError: yield None @abc.abstractmethod def check(self, obj_id): pass @abc.abstractmethod def delete(self, obj_id): if not self.allow_delete: raise PermissionError("Delete is not allowed.") # Management methods def get_random(self, batch_size): pass # Streaming methods def add_stream(self, content_iter, obj_id, check_presence=True): raise NotImplementedError def get_stream(self, obj_id, chunk_size=DEFAULT_CHUNK_SIZE): raise NotImplementedError def list_content(self, last_obj_id=None, limit=DEFAULT_LIMIT): it = iter(self) if last_obj_id: it = dropwhile(lambda x: x <= last_obj_id, it) return islice(it, limit) diff --git a/swh/objstorage/tests/test_multiplexer_filter.py b/swh/objstorage/tests/test_multiplexer_filter.py index 03142a2..e773253 100644 --- a/swh/objstorage/tests/test_multiplexer_filter.py +++ b/swh/objstorage/tests/test_multiplexer_filter.py @@ -1,332 +1,332 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import random import shutil from string import ascii_lowercase import tempfile import unittest from swh.model import hashutil from swh.objstorage.exc import Error, ObjNotFoundError from swh.objstorage.factory import get_objstorage from swh.objstorage.multiplexer.filter import id_prefix, id_regex, read_only from swh.objstorage.objstorage import compute_hash def get_random_content(): return bytes("".join(random.sample(ascii_lowercase, 10)), "utf8") class MixinTestReadFilter(unittest.TestCase): # Read only filter should not allow writing def setUp(self): super().setUp() self.tmpdir = tempfile.mkdtemp() pstorage = { "cls": "pathslicing", "root": self.tmpdir, "slicing": "0:5", } base_storage = get_objstorage(**pstorage) base_storage.id = compute_hash self.storage = get_objstorage( "filtered", storage_conf=pstorage, filters_conf=[read_only()] ) self.valid_content = b"pre-existing content" self.invalid_content = b"invalid_content" self.true_invalid_content = b"Anything that is not correct" self.absent_content = b"non-existent content" # Create a valid content. self.valid_id = base_storage.add(self.valid_content) # Create an invalid id and add a content with it. self.invalid_id = base_storage.id(self.true_invalid_content) base_storage.add(self.invalid_content, obj_id=self.invalid_id) # Compute an id for a non-existing content. self.absent_id = base_storage.id(self.absent_content) def tearDown(self): super().tearDown() shutil.rmtree(self.tmpdir) def test_can_contains(self): self.assertTrue(self.valid_id in self.storage) self.assertTrue(self.invalid_id in self.storage) self.assertFalse(self.absent_id in self.storage) def test_can_iter(self): self.assertIn(self.valid_id, iter(self.storage)) self.assertIn(self.invalid_id, iter(self.storage)) def test_can_len(self): self.assertEqual(2, len(self.storage)) def test_can_get(self): self.assertEqual(self.valid_content, self.storage.get(self.valid_id)) self.assertEqual(self.invalid_content, self.storage.get(self.invalid_id)) def test_can_check(self): with self.assertRaises(ObjNotFoundError): self.storage.check(self.absent_id) with self.assertRaises(Error): self.storage.check(self.invalid_id) self.storage.check(self.valid_id) def test_can_get_random(self): self.assertEqual(1, len(list(self.storage.get_random(1)))) self.assertEqual( len(list(self.storage)), len(set(self.storage.get_random(1000))) ) def test_cannot_add(self): new_id = self.storage.add(b"New content") result = self.storage.add(self.valid_content, self.valid_id) self.assertIsNone(new_id, self.storage) self.assertIsNone(result) def test_cannot_restore(self): result = self.storage.restore(self.valid_content, self.valid_id) self.assertIsNone(result) class MixinTestIdFilter: - """ Mixin class that tests the filters based on filter.IdFilter + """Mixin class that tests the filters based on filter.IdFilter Methods "make_valid", "make_invalid" and "filter_storage" must be implemented by subclasses. """ def setUp(self): super().setUp() # Use a hack here : as the mock uses the content as id, it is easy to # create contents that are filtered or not. self.prefix = "71" self.tmpdir = tempfile.mkdtemp() # Make the storage filtered self.sconf = { "cls": "pathslicing", "args": {"root": self.tmpdir, "slicing": "0:5"}, } storage = get_objstorage(**self.sconf) self.base_storage = storage self.storage = self.filter_storage(self.sconf) # Set the id calculators storage.id = compute_hash # Present content with valid id self.present_valid_content = self.ensure_valid(b"yroqdtotji") self.present_valid_id = storage.id(self.present_valid_content) # Present content with invalid id self.present_invalid_content = self.ensure_invalid(b"glxddlmmzb") self.present_invalid_id = storage.id(self.present_invalid_content) # Missing content with valid id self.missing_valid_content = self.ensure_valid(b"rmzkdclkez") self.missing_valid_id = storage.id(self.missing_valid_content) # Missing content with invalid id self.missing_invalid_content = self.ensure_invalid(b"hlejfuginh") self.missing_invalid_id = storage.id(self.missing_invalid_content) # Present corrupted content with valid id self.present_corrupted_valid_content = self.ensure_valid(b"cdsjwnpaij") self.true_present_corrupted_valid_content = self.ensure_valid(b"mgsdpawcrr") self.present_corrupted_valid_id = storage.id( self.true_present_corrupted_valid_content ) # Present corrupted content with invalid id self.present_corrupted_invalid_content = self.ensure_invalid(b"pspjljnrco") self.true_present_corrupted_invalid_content = self.ensure_invalid(b"rjocbnnbso") self.present_corrupted_invalid_id = storage.id( self.true_present_corrupted_invalid_content ) # Missing (potentially) corrupted content with valid id self.missing_corrupted_valid_content = self.ensure_valid(b"zxkokfgtou") self.true_missing_corrupted_valid_content = self.ensure_valid(b"royoncooqa") self.missing_corrupted_valid_id = storage.id( self.true_missing_corrupted_valid_content ) # Missing (potentially) corrupted content with invalid id self.missing_corrupted_invalid_content = self.ensure_invalid(b"hxaxnrmnyk") self.true_missing_corrupted_invalid_content = self.ensure_invalid(b"qhbolyuifr") self.missing_corrupted_invalid_id = storage.id( self.true_missing_corrupted_invalid_content ) # Add the content that are supposed to be present self.storage.add(self.present_valid_content) self.storage.add(self.present_invalid_content) self.storage.add( self.present_corrupted_valid_content, obj_id=self.present_corrupted_valid_id ) self.storage.add( self.present_corrupted_invalid_content, obj_id=self.present_corrupted_invalid_id, ) def tearDown(self): super().tearDown() shutil.rmtree(self.tmpdir) def filter_storage(self, sconf): raise NotImplementedError( "Id_filter test class must have a filter_storage method" ) def ensure_valid(self, content=None): if content is None: content = get_random_content() while not self.storage.is_valid(self.base_storage.id(content)): content = get_random_content() return content def ensure_invalid(self, content=None): if content is None: content = get_random_content() while self.storage.is_valid(self.base_storage.id(content)): content = get_random_content() return content def test_contains(self): # Both contents are present, but the invalid one should be ignored. self.assertTrue(self.present_valid_id in self.storage) self.assertFalse(self.present_invalid_id in self.storage) self.assertFalse(self.missing_valid_id in self.storage) self.assertFalse(self.missing_invalid_id in self.storage) self.assertTrue(self.present_corrupted_valid_id in self.storage) self.assertFalse(self.present_corrupted_invalid_id in self.storage) self.assertFalse(self.missing_corrupted_valid_id in self.storage) self.assertFalse(self.missing_corrupted_invalid_id in self.storage) def test_iter(self): self.assertIn(self.present_valid_id, iter(self.storage)) self.assertNotIn(self.present_invalid_id, iter(self.storage)) self.assertNotIn(self.missing_valid_id, iter(self.storage)) self.assertNotIn(self.missing_invalid_id, iter(self.storage)) self.assertIn(self.present_corrupted_valid_id, iter(self.storage)) self.assertNotIn(self.present_corrupted_invalid_id, iter(self.storage)) self.assertNotIn(self.missing_corrupted_valid_id, iter(self.storage)) self.assertNotIn(self.missing_corrupted_invalid_id, iter(self.storage)) def test_len(self): # Four contents are present, but only two should be valid. self.assertEqual(2, len(self.storage)) def test_get(self): self.assertEqual( self.present_valid_content, self.storage.get(self.present_valid_id) ) with self.assertRaises(ObjNotFoundError): self.storage.get(self.present_invalid_id) with self.assertRaises(ObjNotFoundError): self.storage.get(self.missing_valid_id) with self.assertRaises(ObjNotFoundError): self.storage.get(self.missing_invalid_id) self.assertEqual( self.present_corrupted_valid_content, self.storage.get(self.present_corrupted_valid_id), ) with self.assertRaises(ObjNotFoundError): self.storage.get(self.present_corrupted_invalid_id) with self.assertRaises(ObjNotFoundError): self.storage.get(self.missing_corrupted_valid_id) with self.assertRaises(ObjNotFoundError): self.storage.get(self.missing_corrupted_invalid_id) def test_check(self): self.storage.check(self.present_valid_id) with self.assertRaises(ObjNotFoundError): self.storage.check(self.present_invalid_id) with self.assertRaises(ObjNotFoundError): self.storage.check(self.missing_valid_id) with self.assertRaises(ObjNotFoundError): self.storage.check(self.missing_invalid_id) with self.assertRaises(Error): self.storage.check(self.present_corrupted_valid_id) with self.assertRaises(ObjNotFoundError): self.storage.check(self.present_corrupted_invalid_id) with self.assertRaises(ObjNotFoundError): self.storage.check(self.missing_corrupted_valid_id) with self.assertRaises(ObjNotFoundError): self.storage.check(self.missing_corrupted_invalid_id) def test_get_random(self): self.assertEqual(0, len(list(self.storage.get_random(0)))) random_content = list(self.storage.get_random(1000)) self.assertIn(self.present_valid_id, random_content) self.assertNotIn(self.present_invalid_id, random_content) self.assertNotIn(self.missing_valid_id, random_content) self.assertNotIn(self.missing_invalid_id, random_content) self.assertIn(self.present_corrupted_valid_id, random_content) self.assertNotIn(self.present_corrupted_invalid_id, random_content) self.assertNotIn(self.missing_corrupted_valid_id, random_content) self.assertNotIn(self.missing_corrupted_invalid_id, random_content) def test_add(self): # Add valid and invalid contents to the storage and check their # presence with the unfiltered storage. valid_content = self.ensure_valid(b"ulepsrjbgt") valid_id = self.base_storage.id(valid_content) invalid_content = self.ensure_invalid(b"znvghkjked") invalid_id = self.base_storage.id(invalid_content) self.storage.add(valid_content) self.storage.add(invalid_content) self.assertTrue(valid_id in self.base_storage) self.assertFalse(invalid_id in self.base_storage) def test_restore(self): # Add corrupted content to the storage and the try to restore it valid_content = self.ensure_valid(b"ulepsrjbgt") valid_id = self.base_storage.id(valid_content) corrupted_content = self.ensure_valid(b"ltjkjsloyb") corrupted_id = self.base_storage.id(corrupted_content) self.storage.add(corrupted_content, obj_id=valid_id) with self.assertRaises(ObjNotFoundError): self.storage.check(corrupted_id) with self.assertRaises(Error): self.storage.check(valid_id) self.storage.restore(valid_content) self.storage.check(valid_id) class TestPrefixFilter(MixinTestIdFilter, unittest.TestCase): def setUp(self): self.prefix = b"71" super().setUp() def ensure_valid(self, content): obj_id = compute_hash(content) hex_obj_id = hashutil.hash_to_hex(obj_id) self.assertTrue(hex_obj_id.startswith(self.prefix)) return content def ensure_invalid(self, content): obj_id = compute_hash(content) hex_obj_id = hashutil.hash_to_hex(obj_id) self.assertFalse(hex_obj_id.startswith(self.prefix)) return content def filter_storage(self, sconf): return get_objstorage( "filtered", {"storage_conf": sconf, "filters_conf": [id_prefix(self.prefix)]}, ) class TestRegexFilter(MixinTestIdFilter, unittest.TestCase): def setUp(self): self.regex = r"[a-f][0-9].*" super().setUp() def filter_storage(self, sconf): return get_objstorage( "filtered", {"storage_conf": sconf, "filters_conf": [id_regex(self.regex)]} ) diff --git a/swh/objstorage/tests/test_objstorage_api.py b/swh/objstorage/tests/test_objstorage_api.py index c6458cc..2bea41b 100644 --- a/swh/objstorage/tests/test_objstorage_api.py +++ b/swh/objstorage/tests/test_objstorage_api.py @@ -1,48 +1,47 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import shutil import tempfile import unittest import pytest from swh.core.api.tests.server_testing import ServerTestFixture from swh.objstorage.api.server import app from swh.objstorage.factory import get_objstorage from swh.objstorage.tests.objstorage_testing import ObjStorageTestFixture class TestRemoteObjStorage(ServerTestFixture, ObjStorageTestFixture, unittest.TestCase): - """ Test the remote archive API. - """ + """Test the remote archive API.""" def setUp(self): self.tmpdir = tempfile.mkdtemp() self.config = { "objstorage": { "cls": "pathslicing", "root": self.tmpdir, "slicing": "0:1/0:5", "allow_delete": True, }, "client_max_size": 8 * 1024 * 1024, } self.app = app super().setUp() self.storage = get_objstorage("remote", {"url": self.url()}) def tearDown(self): super().tearDown() shutil.rmtree(self.tmpdir) @pytest.mark.skip("makes no sense to test this for the remote api") def test_delete_not_allowed(self): pass @pytest.mark.skip("makes no sense to test this for the remote api") def test_delete_not_allowed_by_default(self): pass diff --git a/swh/objstorage/tests/test_objstorage_azure.py b/swh/objstorage/tests/test_objstorage_azure.py index 85e0d0a..dcc1f78 100644 --- a/swh/objstorage/tests/test_objstorage_azure.py +++ b/swh/objstorage/tests/test_objstorage_azure.py @@ -1,313 +1,317 @@ # Copyright (C) 2016-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import asyncio import base64 import collections from dataclasses import dataclass import unittest from unittest.mock import patch from urllib.parse import parse_qs, urlparse from azure.core.exceptions import ResourceExistsError, ResourceNotFoundError import pytest from swh.model.hashutil import hash_to_hex import swh.objstorage.backends.azure from swh.objstorage.exc import Error from swh.objstorage.factory import get_objstorage from swh.objstorage.objstorage import decompressors from .objstorage_testing import ObjStorageTestFixture @dataclass class MockListedObject: name: str class MockAsyncDownloadClient: def __init__(self, blob_data): self.blob_data = blob_data def content_as_bytes(self): future = asyncio.Future() future.set_result(self.blob_data) return future class MockDownloadClient: def __init__(self, blob_data): self.blob_data = blob_data def content_as_bytes(self): return self.blob_data def __await__(self): yield from () return MockAsyncDownloadClient(self.blob_data) class MockBlobClient: def __init__(self, container, blob): self.container = container self.blob = blob def get_blob_properties(self): if self.blob not in self.container.blobs: raise ResourceNotFoundError("Blob not found") return {"exists": True} def upload_blob(self, data, length=None): if self.blob in self.container.blobs: raise ResourceExistsError("Blob already exists") if length is not None and length != len(data): raise ValueError("Wrong length for blob data!") self.container.blobs[self.blob] = data def download_blob(self): if self.blob not in self.container.blobs: raise ResourceNotFoundError("Blob not found") return MockDownloadClient(self.container.blobs[self.blob]) def delete_blob(self): if self.blob not in self.container.blobs: raise ResourceNotFoundError("Blob not found") del self.container.blobs[self.blob] def get_MockContainerClient(): blobs = collections.defaultdict(dict) # {container_url: {blob_id: blob}} class MockContainerClient: def __init__(self, container_url): self.container_url = container_url self.blobs = blobs[self.container_url] @classmethod def from_container_url(cls, container_url): return cls(container_url) def get_container_properties(self): return {"exists": True} def get_blob_client(self, blob): return MockBlobClient(self, blob) def list_blobs(self): for obj in sorted(self.blobs): yield MockListedObject(obj) def delete_blob(self, blob): self.get_blob_client(blob.name).delete_blob() def __aenter__(self): return self def __await__(self): future = asyncio.Future() future.set_result(self) yield from future def __aexit__(self, *args): return self return MockContainerClient class TestAzureCloudObjStorage(ObjStorageTestFixture, unittest.TestCase): compression = "none" def setUp(self): super().setUp() ContainerClient = get_MockContainerClient() patcher = patch( "swh.objstorage.backends.azure.ContainerClient", ContainerClient ) patcher.start() self.addCleanup(patcher.stop) patcher = patch( "swh.objstorage.backends.azure.AsyncContainerClient", ContainerClient ) patcher.start() self.addCleanup(patcher.stop) self.storage = get_objstorage( "azure", { "container_url": "https://bogus-container-url.example", "compression": self.compression, }, ) def test_compression(self): content, obj_id = self.hash_content(b"test content is compressed") self.storage.add(content, obj_id=obj_id) internal_id = self.storage._internal_id(obj_id) blob_client = self.storage.get_blob_client(internal_id) raw_blob = blob_client.download_blob().content_as_bytes() d = decompressors[self.compression]() assert d.decompress(raw_blob) == content assert d.unused_data == b"" def test_trailing_data_on_stored_blob(self): content, obj_id = self.hash_content(b"test content without garbage") self.storage.add(content, obj_id=obj_id) internal_id = self.storage._internal_id(obj_id) blob_client = self.storage.get_blob_client(internal_id) raw_blob = blob_client.download_blob().content_as_bytes() new_data = raw_blob + b"trailing garbage" blob_client.delete_blob() blob_client.upload_blob(data=new_data, length=len(new_data)) if self.compression == "none": with self.assertRaises(Error) as e: self.storage.check(obj_id) else: with self.assertRaises(Error) as e: self.storage.get(obj_id) assert "trailing data" in e.exception.args[0] class TestAzureCloudObjStorageGzip(TestAzureCloudObjStorage): compression = "gzip" class TestAzureCloudObjStorageZlib(TestAzureCloudObjStorage): compression = "zlib" class TestAzureCloudObjStorageLzma(TestAzureCloudObjStorage): compression = "lzma" class TestAzureCloudObjStorageBz2(TestAzureCloudObjStorage): compression = "bz2" class TestPrefixedAzureCloudObjStorage(ObjStorageTestFixture, unittest.TestCase): def setUp(self): super().setUp() self.ContainerClient = get_MockContainerClient() patcher = patch( "swh.objstorage.backends.azure.ContainerClient", self.ContainerClient ) patcher.start() self.addCleanup(patcher.stop) patcher = patch( "swh.objstorage.backends.azure.AsyncContainerClient", self.ContainerClient ) patcher.start() self.addCleanup(patcher.stop) self.accounts = {} for prefix in "0123456789abcdef": self.accounts[prefix] = "https://bogus-container-url.example/" + prefix self.storage = get_objstorage("azure-prefixed", {"accounts": self.accounts}) def test_prefixedazure_instantiation_missing_prefixes(self): del self.accounts["d"] del self.accounts["e"] with self.assertRaisesRegex(ValueError, "Missing prefixes"): get_objstorage("azure-prefixed", {"accounts": self.accounts}) def test_prefixedazure_instantiation_inconsistent_prefixes(self): self.accounts["00"] = self.accounts["0"] with self.assertRaisesRegex(ValueError, "Inconsistent prefixes"): get_objstorage("azure-prefixed", {"accounts": self.accounts}) def test_prefixedazure_sharding_behavior(self): for i in range(100): content, obj_id = self.hash_content(b"test_content_%02d" % i) self.storage.add(content, obj_id=obj_id) hex_obj_id = hash_to_hex(obj_id) prefix = hex_obj_id[0] self.assertTrue( self.ContainerClient(self.storage.container_urls[prefix]) .get_blob_client(hex_obj_id) .get_blob_properties() ) def test_get_container_url(): # r=read, l=list, w=write, d=delete policy_map = { "read_only": "rl", "append_only": "rwl", "full": "rwdl", } for policy, expected in policy_map.items(): ret = swh.objstorage.backends.azure.get_container_url( account_name="account_name", account_key=base64.b64encode(b"account_key"), container_name="container_name", access_policy=policy, ) p = urlparse(ret) assert p.scheme == "https" assert p.netloc == "account_name.blob.core.windows.net" assert p.path == "/container_name" qs = parse_qs(p.query) # sp: permissions assert qs["sp"] == [expected] # sr: resource (c=container) assert qs["sr"] == ["c"] # st: start; se: expiry assert qs["st"][0] < qs["se"][0] def test_bwcompat_args(monkeypatch): monkeypatch.setattr( - swh.objstorage.backends.azure, "ContainerClient", get_MockContainerClient(), + swh.objstorage.backends.azure, + "ContainerClient", + get_MockContainerClient(), ) with pytest.deprecated_call(): objs = get_objstorage( "azure", { "account_name": "account_name", "api_secret_key": base64.b64encode(b"account_key"), "container_name": "container_name", }, ) assert objs is not None def test_bwcompat_args_prefixed(monkeypatch): monkeypatch.setattr( - swh.objstorage.backends.azure, "ContainerClient", get_MockContainerClient(), + swh.objstorage.backends.azure, + "ContainerClient", + get_MockContainerClient(), ) accounts = { prefix: { "account_name": f"account_name{prefix}", "api_secret_key": base64.b64encode(b"account_key"), "container_name": "container_name", } for prefix in "0123456789abcdef" } with pytest.deprecated_call(): objs = get_objstorage("azure-prefixed", {"accounts": accounts}) assert objs is not None diff --git a/swh/objstorage/tests/test_objstorage_cloud.py b/swh/objstorage/tests/test_objstorage_cloud.py index 656a7b7..4d86814 100644 --- a/swh/objstorage/tests/test_objstorage_cloud.py +++ b/swh/objstorage/tests/test_objstorage_cloud.py @@ -1,164 +1,164 @@ # Copyright (C) 2016 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from typing import Optional import unittest from libcloud.common.types import InvalidCredsError from libcloud.storage.types import ContainerDoesNotExistError, ObjectDoesNotExistError from swh.objstorage.backends.libcloud import CloudObjStorage from swh.objstorage.exc import Error from swh.objstorage.objstorage import decompressors from .objstorage_testing import ObjStorageTestFixture API_KEY = "API_KEY" API_SECRET_KEY = "API SECRET KEY" CONTAINER_NAME = "test_container" class MockLibcloudObject: - """ Libcloud object mock that replicates its API """ + """Libcloud object mock that replicates its API""" def __init__(self, name, content): self.name = name self.content = list(content) def as_stream(self): yield from iter(self.content) class MockLibcloudDriver: - """ Mock driver that replicates the used LibCloud API """ + """Mock driver that replicates the used LibCloud API""" def __init__(self, api_key, api_secret_key): self.containers = {CONTAINER_NAME: {}} # Storage is initialized self.api_key = api_key self.api_secret_key = api_secret_key def _check_credentials(self): # Private method may be known as another name in Libcloud but is used # to replicate libcloud behavior (i.e. check credential at each # request) if self.api_key != API_KEY or self.api_secret_key != API_SECRET_KEY: raise InvalidCredsError() def get_container(self, container_name): try: return self.containers[container_name] except KeyError: raise ContainerDoesNotExistError( container_name=container_name, driver=self, value=None ) def iterate_container_objects(self, container): self._check_credentials() yield from (v for k, v in sorted(container.items())) def get_object(self, container_name, obj_id): self._check_credentials() try: container = self.get_container(container_name) return container[obj_id] except KeyError: raise ObjectDoesNotExistError(object_name=obj_id, driver=self, value=None) def delete_object(self, obj): self._check_credentials() try: container = self.get_container(CONTAINER_NAME) container.pop(obj.name) return True except KeyError: raise ObjectDoesNotExistError(object_name=obj.name, driver=self, value=None) def upload_object_via_stream(self, content, container, obj_id): self._check_credentials() obj = MockLibcloudObject(obj_id, content) container[obj_id] = obj class MockCloudObjStorage(CloudObjStorage): - """ Cloud object storage that uses a mocked driver """ + """Cloud object storage that uses a mocked driver""" def _get_driver(self, **kwargs): return MockLibcloudDriver(**kwargs) def _get_provider(self): # Implement this for the abc requirement, but behavior is defined in # _get_driver. pass class TestCloudObjStorage(ObjStorageTestFixture, unittest.TestCase): compression = "none" path_prefix: Optional[str] = None def setUp(self): super().setUp() self.storage = MockCloudObjStorage( CONTAINER_NAME, api_key=API_KEY, api_secret_key=API_SECRET_KEY, compression=self.compression, path_prefix=self.path_prefix, ) def test_compression(self): content, obj_id = self.hash_content(b"add_get_w_id") self.storage.add(content, obj_id=obj_id) libcloud_object = self.storage._get_object(obj_id) raw_content = b"".join(libcloud_object.content) d = decompressors[self.compression]() assert d.decompress(raw_content) == content assert d.unused_data == b"" def test_trailing_data_on_stored_blob(self): content, obj_id = self.hash_content(b"test content without garbage") self.storage.add(content, obj_id=obj_id) libcloud_object = self.storage._get_object(obj_id) libcloud_object.content.append(b"trailing garbage") if self.compression == "none": with self.assertRaises(Error) as e: self.storage.check(obj_id) else: with self.assertRaises(Error) as e: self.storage.get(obj_id) assert "trailing data" in e.exception.args[0] class TestCloudObjStorageBz2(TestCloudObjStorage): compression = "bz2" class TestCloudObjStorageGzip(TestCloudObjStorage): compression = "gzip" class TestCloudObjStorageLzma(TestCloudObjStorage): compression = "lzma" class TestCloudObjStorageZlib(TestCloudObjStorage): compression = "zlib" class TestCloudObjStoragePrefix(TestCloudObjStorage): path_prefix = "contents" def test_path_prefix(self): content, obj_id = self.hash_content(b"test content") self.storage.add(content, obj_id=obj_id) container = self.storage.driver.containers[CONTAINER_NAME] object_path = self.storage._object_path(obj_id) assert object_path.startswith(self.path_prefix + "/") assert object_path in container diff --git a/swh/objstorage/tests/test_objstorage_instantiation.py b/swh/objstorage/tests/test_objstorage_instantiation.py index a1a5c88..6308718 100644 --- a/swh/objstorage/tests/test_objstorage_instantiation.py +++ b/swh/objstorage/tests/test_objstorage_instantiation.py @@ -1,40 +1,40 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import shutil import tempfile import unittest from swh.objstorage.api.client import RemoteObjStorage from swh.objstorage.backends.pathslicing import PathSlicingObjStorage from swh.objstorage.factory import get_objstorage class TestObjStorageInitialization(unittest.TestCase): - """ Test that the methods for ObjStorage initializations with + """Test that the methods for ObjStorage initializations with `get_objstorage` works properly. """ def setUp(self): self.path = tempfile.mkdtemp() self.path2 = tempfile.mkdtemp() # Server is launched at self.url() self.config = {"storage_base": self.path2, "storage_slicing": "0:1/0:5"} super().setUp() def tearDown(self): super().tearDown() shutil.rmtree(self.path) shutil.rmtree(self.path2) def test_pathslicing_objstorage(self): conf = {"cls": "pathslicing", "args": {"root": self.path, "slicing": "0:2/0:5"}} st = get_objstorage(**conf) self.assertTrue(isinstance(st, PathSlicingObjStorage)) def test_remote_objstorage(self): conf = {"cls": "remote", "args": {"url": "http://127.0.0.1:4242/"}} st = get_objstorage(**conf) self.assertTrue(isinstance(st, RemoteObjStorage)) diff --git a/swh/objstorage/tests/test_objstorage_seaweedfs.py b/swh/objstorage/tests/test_objstorage_seaweedfs.py index ef8e83c..4da8ccd 100644 --- a/swh/objstorage/tests/test_objstorage_seaweedfs.py +++ b/swh/objstorage/tests/test_objstorage_seaweedfs.py @@ -1,174 +1,177 @@ # Copyright (C) 2019 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from itertools import dropwhile, islice import json import os import unittest from urllib.parse import urlparse from requests.utils import get_encoding_from_headers import requests_mock from requests_mock.contrib import fixture from swh.objstorage.exc import Error from swh.objstorage.factory import get_objstorage from swh.objstorage.objstorage import decompressors from swh.objstorage.tests.objstorage_testing import ObjStorageTestFixture class FilerRequestsMock: """This is a requests_mock based mock for the seaweedfs Filer API It does not implement the whole API, only the parts required to make the WeedFiler (used by WeedObjStorage) work. It stores the files in a dict. """ MODE_DIR = 0o20000000771 MODE_FILE = 0o660 def __init__(self, baseurl): self.baseurl = baseurl self.basepath = urlparse(baseurl).path self.content = {} self.requests_mock = fixture.Fixture() self.requests_mock.setUp() self.requests_mock.register_uri( requests_mock.GET, requests_mock.ANY, content=self.get_cb ) self.requests_mock.register_uri( requests_mock.POST, requests_mock.ANY, content=self.post_cb ) self.requests_mock.register_uri( requests_mock.HEAD, requests_mock.ANY, content=self.head_cb ) self.requests_mock.register_uri( requests_mock.DELETE, requests_mock.ANY, content=self.delete_cb ) def relpath(self, path): if path.startswith(self.basepath): return os.path.relpath(path, self.basepath) def head_cb(self, request, context): relpath = self.relpath(request.path) if relpath == "." or relpath in self.content: return b"Found" # ok, found it context.status_code = 404 return b"Not Found" def get_cb(self, request, context): if self.head_cb(request, context) == b"Not Found": return relpath = self.relpath(request.path) if relpath == ".": if "limit" in request.qs: limit = int(request.qs["limit"][0]) assert limit > 0 else: limit = None items = sorted(self.content.items()) if items and "lastfilename" in request.qs: lastfilename = request.qs["lastfilename"][0] if lastfilename: # exclude all filenames up to lastfilename items = dropwhile(lambda kv: kv[0] <= lastfilename, items) if limit: # +1 to easily detect if there are more items = islice(items, limit + 1) entries = [ - {"FullPath": os.path.join(request.path, fname), "Mode": self.MODE_FILE,} + { + "FullPath": os.path.join(request.path, fname), + "Mode": self.MODE_FILE, + } for fname, obj in items ] thereismore = False if limit and len(entries) > limit: entries = entries[:limit] thereismore = True if entries: lastfilename = entries[-1]["FullPath"].split("/")[-1] else: lastfilename = None text = json.dumps( { "Path": request.path, "Limit": limit, "LastFileName": lastfilename, "ShouldDisplayLoadMore": thereismore, "Entries": entries, } ) encoding = get_encoding_from_headers(request.headers) or "utf-8" return text.encode(encoding) else: return self.content[relpath] def post_cb(self, request, context): from requests_toolbelt.multipart import decoder multipart_data = decoder.MultipartDecoder( request.body, request.headers["content-type"] ) part = multipart_data.parts[0] self.content[self.relpath(request.path)] = part.content def delete_cb(self, request, context): del self.content[self.relpath(request.path)] class TestWeedObjStorage(ObjStorageTestFixture, unittest.TestCase): compression = "none" url = "http://127.0.0.1/test/" def setUp(self): super().setUp() self.storage = get_objstorage( cls="seaweedfs", url=self.url, compression=self.compression ) self.mock = FilerRequestsMock(baseurl=self.url) def test_compression(self): content, obj_id = self.hash_content(b"test compression") self.storage.add(content, obj_id=obj_id) raw_content = self.storage.wf.get(self.storage._path(obj_id)) d = decompressors[self.compression]() assert d.decompress(raw_content) == content assert d.unused_data == b"" def test_trailing_data_on_stored_blob(self): content, obj_id = self.hash_content(b"test content without garbage") self.storage.add(content, obj_id=obj_id) self.mock.content[obj_id.hex()] += b"trailing garbage" if self.compression == "none": with self.assertRaises(Error) as e: self.storage.check(obj_id) else: with self.assertRaises(Error) as e: self.storage.get(obj_id) assert "trailing data" in e.exception.args[0] class TestWeedObjStorageWithCompression(TestWeedObjStorage): compression = "lzma" class TestWeedObjStorageWithSmallBatch(TestWeedObjStorage): def setUp(self): super().setUp() self.storage.wf.batchsize = 1 class TestWeedObjStorageWithNoPath(TestWeedObjStorage): url = "http://127.0.0.1/" diff --git a/swh/objstorage/tests/test_server.py b/swh/objstorage/tests/test_server.py index 6a59d32..5dbde1d 100644 --- a/swh/objstorage/tests/test_server.py +++ b/swh/objstorage/tests/test_server.py @@ -1,116 +1,126 @@ # Copyright (C) 2019-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import copy import pytest import yaml from swh.objstorage.api.server import load_and_check_config def prepare_config_file(tmpdir, content, name="config.yml"): """Prepare configuration file in `$tmpdir/name` with content `content`. Args: tmpdir (LocalPath): root directory content (str/dict): Content of the file either as string or as a dict. If a dict, converts the dict into a yaml string. name (str): configuration filename Returns path (str) of the configuration file prepared. """ config_path = tmpdir / name if isinstance(content, dict): # convert if needed content = yaml.dump(content) config_path.write_text(content, encoding="utf-8") # pytest on python3.5 does not support LocalPath manipulation, so # convert path to string return str(config_path) def test_load_and_check_config_no_configuration(): """Inexistent configuration files raises""" with pytest.raises(EnvironmentError, match="Configuration file must be defined"): load_and_check_config(None) config_path = "/indexer/inexistent/config.yml" with pytest.raises(FileNotFoundError, match=f"{config_path} does not exist"): load_and_check_config(config_path) def test_load_and_check_config_invalid_configuration_toplevel(tmpdir): """Invalid configuration raises""" config = {"something": "useless"} config_path = prepare_config_file(tmpdir, content=config) with pytest.raises(KeyError, match="missing objstorage config entry"): load_and_check_config(config_path) def test_load_and_check_config_invalid_configuration(tmpdir): """Invalid configuration raises""" config_path = prepare_config_file( tmpdir, content={"objstorage": {"something": "useless"}} ) with pytest.raises(KeyError, match="missing cls config entry"): load_and_check_config(config_path) def test_load_and_check_config_invalid_configuration_level2(tmpdir): """Invalid configuration at 2nd level raises""" config = { "objstorage": { "cls": "pathslicing", - "args": {"root": "root", "slicing": "slicing",}, + "args": { + "root": "root", + "slicing": "slicing", + }, "client_max_size": "10", } } for key in ("root", "slicing"): c = copy.deepcopy(config) c["objstorage"]["args"].pop(key) config_path = prepare_config_file(tmpdir, c) with pytest.raises(KeyError, match=f"missing {key} config entry"): load_and_check_config(config_path) @pytest.mark.parametrize( "config", [ pytest.param( { "objstorage": { "cls": "pathslicing", "args": {"root": "root", "slicing": "slicing"}, } }, id="pathslicing-bw-compat", ), pytest.param( { "objstorage": { "cls": "pathslicing", "root": "root", "slicing": "slicing", } }, id="pathslicing", ), pytest.param( {"client_max_size": "10", "objstorage": {"cls": "memory", "args": {}}}, id="empty-args-bw-compat", ), pytest.param( {"client_max_size": "10", "objstorage": {"cls": "memory"}}, id="empty-args" ), - pytest.param({"objstorage": {"cls": "noop",}}, id="noop",), + pytest.param( + { + "objstorage": { + "cls": "noop", + } + }, + id="noop", + ), ], ) def test_load_and_check_config(tmpdir, config): """pathslicing configuration fine loads ok""" config_path = prepare_config_file(tmpdir, config) cfg = load_and_check_config(config_path) assert cfg == config