##// END OF EJS Templates
chore: fixed merge conflicts
chore: fixed merge conflicts

File last commit:

r5516:3496180b default
r5546:50cf7822 merge v5.2.0 stable
Show More
filesystem_legacy.py
278 lines | 9.2 KiB | text/x-python | PythonLexer
feat(artifacts): new artifact storage engines allowing an s3 based uploads
r5516 # Copyright (C) 2016-2023 RhodeCode GmbH
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License, version 3
# (only), as published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#
# This program is dual-licensed. If you wish to learn more about the
# RhodeCode Enterprise Edition, including its added features, Support services,
# and proprietary license terms, please see https://rhodecode.com/licenses/
import errno
import os
import hashlib
import functools
import time
import logging
from .. import config_keys
from ..exceptions import FileOverSizeException
from ..backends.base import BaseFileStoreBackend, fsspec, BaseShard, ShardFileReader
from ....lib.ext_json import json
log = logging.getLogger(__name__)
class LegacyFileSystemShard(BaseShard):
# legacy ver
METADATA_VER = 'v2'
BACKEND_TYPE = config_keys.backend_legacy_filesystem
storage_type: str = 'dir_struct'
# legacy suffix
metadata_suffix: str = '.meta'
@classmethod
def _sub_store_from_filename(cls, filename):
return filename[:2]
@classmethod
def apply_counter(cls, counter, filename):
name_counted = '%d-%s' % (counter, filename)
return name_counted
@classmethod
def safe_make_dirs(cls, dir_path):
if not os.path.exists(dir_path):
try:
os.makedirs(dir_path)
except OSError as e:
if e.errno != errno.EEXIST:
raise
return
@classmethod
def resolve_name(cls, name, directory):
"""
Resolves a unique name and the correct path. If a filename
for that path already exists then a numeric prefix with values > 0 will be
added, for example test.jpg -> 1-test.jpg etc. initially file would have 0 prefix.
:param name: base name of file
:param directory: absolute directory path
"""
counter = 0
while True:
name_counted = cls.apply_counter(counter, name)
# sub_store prefix to optimize disk usage, e.g some_path/ab/final_file
sub_store: str = cls._sub_store_from_filename(name_counted)
sub_store_path: str = os.path.join(directory, sub_store)
cls.safe_make_dirs(sub_store_path)
path = os.path.join(sub_store_path, name_counted)
if not os.path.exists(path):
return name_counted, path
counter += 1
def __init__(self, index, directory, directory_folder, fs, **settings):
self._index: int = index
self._directory: str = directory
self._directory_folder: str = directory_folder
self.fs = fs
@property
def dir_struct(self) -> str:
"""Cache directory final path."""
return os.path.join(self._directory, '0-')
def _write_file(self, full_path, iterator, max_filesize, mode='wb'):
# ensure dir exists
destination, _ = os.path.split(full_path)
if not self.fs.exists(destination):
self.fs.makedirs(destination)
writer = self.fs.open(full_path, mode)
digest = hashlib.sha256()
oversize_cleanup = False
with writer:
size = 0
for chunk in iterator:
size += len(chunk)
digest.update(chunk)
writer.write(chunk)
if max_filesize and size > max_filesize:
# free up the copied file, and raise exc
oversize_cleanup = True
break
writer.flush()
# Get the file descriptor
fd = writer.fileno()
# Sync the file descriptor to disk, helps with NFS cases...
os.fsync(fd)
if oversize_cleanup:
self.fs.rm(full_path)
raise FileOverSizeException(f'given file is over size limit ({max_filesize}): {full_path}')
sha256 = digest.hexdigest()
log.debug('written new artifact under %s, sha256: %s', full_path, sha256)
return size, sha256
def _store(self, key: str, uid_key, value_reader, max_filesize: int | None = None, metadata: dict | None = None, **kwargs):
filename = key
uid_filename = uid_key
# NOTE:, also apply N- Counter...
uid_filename, full_path = self.resolve_name(uid_filename, self._directory)
# STORE METADATA
# TODO: make it compatible, and backward proof
_metadata = {
"version": self.METADATA_VER,
"filename": filename,
"filename_uid_path": full_path,
"filename_uid": uid_filename,
"sha256": "", # NOTE: filled in by reader iteration
"store_time": time.time(),
"size": 0
}
if metadata:
_metadata.update(metadata)
read_iterator = iter(functools.partial(value_reader.read, 2**22), b'')
size, sha256 = self._write_file(full_path, read_iterator, max_filesize)
_metadata['size'] = size
_metadata['sha256'] = sha256
# after storing the artifacts, we write the metadata present
_metadata_file, metadata_file_path = self.get_metadata_filename(uid_filename)
with self.fs.open(metadata_file_path, 'wb') as f:
f.write(json.dumps(_metadata))
return uid_filename, _metadata
def store_path(self, uid_filename):
"""
Returns absolute file path of the uid_filename
"""
prefix_dir = ''
if '/' in uid_filename:
prefix_dir, filename = uid_filename.split('/')
sub_store = self._sub_store_from_filename(filename)
else:
sub_store = self._sub_store_from_filename(uid_filename)
return os.path.join(self._directory, prefix_dir, sub_store, uid_filename)
def metadata_convert(self, uid_filename, metadata):
# NOTE: backward compat mode here... this is for file created PRE 5.2 system
if 'meta_ver' in metadata:
full_path = self.store_path(uid_filename)
metadata = {
"_converted": True,
"_org": metadata,
"version": self.METADATA_VER,
"store_type": self.BACKEND_TYPE,
"filename": metadata['filename'],
"filename_uid_path": full_path,
"filename_uid": uid_filename,
"sha256": metadata['sha256'],
"store_time": metadata['time'],
"size": metadata['size']
}
return metadata
def _fetch(self, key, presigned_url_expires: int = 0):
if key not in self:
log.exception(f'requested key={key} not found in {self}')
raise KeyError(key)
metadata = self.get_metadata(key)
file_path = metadata['filename_uid_path']
if presigned_url_expires and presigned_url_expires > 0:
metadata['url'] = self.fs.url(file_path, expires=presigned_url_expires)
return ShardFileReader(self.fs.open(file_path, 'rb')), metadata
def delete(self, key):
return self._delete(key)
def _delete(self, key):
if key not in self:
log.exception(f'requested key={key} not found in {self}')
raise KeyError(key)
metadata = self.get_metadata(key)
metadata_file, metadata_file_path = self.get_metadata_filename(key)
artifact_file_path = metadata['filename_uid_path']
self.fs.rm(artifact_file_path)
self.fs.rm(metadata_file_path)
def get_metadata_filename(self, uid_filename) -> tuple[str, str]:
metadata_file: str = f'{uid_filename}{self.metadata_suffix}'
uid_path_in_store = self.store_path(uid_filename)
metadata_file_path = f'{uid_path_in_store}{self.metadata_suffix}'
return metadata_file, metadata_file_path
class LegacyFileSystemBackend(BaseFileStoreBackend):
_shard_cls = LegacyFileSystemShard
def __init__(self, settings):
super().__init__(settings)
store_dir = self.get_conf(config_keys.legacy_filesystem_storage_path)
directory = os.path.expanduser(store_dir)
self._directory = directory
self._storage_path = directory # common path for all from BaseCache
log.debug('Initializing %s file_store instance', self)
fs = fsspec.filesystem('file')
if not fs.exists(self._directory):
fs.makedirs(self._directory, exist_ok=True)
# legacy system uses single shard
self._shards = tuple(
[
self._shard_cls(
index=0,
directory=directory,
directory_folder='',
fs=fs,
**settings,
)
]
)
@classmethod
def get_shard_index(cls, filename: str, num_shards) -> int:
# legacy filesystem doesn't use shards, and always uses single shard
return 0