This repository has been archived on 2020-03-24. You can view files and clone it, but cannot push or open issues or pull requests.
quay/endpoints/registry.py

541 lines
18 KiB
Python
Raw Normal View History

2013-09-25 21:50:03 +00:00
import logging
import json
2013-09-25 21:50:03 +00:00
from flask import (make_response, request, session, Response, redirect,
Blueprint, abort as flask_abort)
2013-09-25 21:50:03 +00:00
from functools import wraps
from datetime import datetime
from time import time
from app import storage as store, image_diff_queue
from auth.auth import process_auth, extract_namespace_repo_from_session
from util import checksums, changes
from util.http import abort, exact_abort
from auth.permissions import (ReadRepositoryPermission,
ModifyRepositoryPermission)
from data import model
2013-09-25 21:50:03 +00:00
registry = Blueprint('registry', __name__)
2013-09-25 21:50:03 +00:00
logger = logging.getLogger(__name__)
profile = logging.getLogger('application.profiler')
2013-09-25 21:50:03 +00:00
class SocketReader(object):
2013-10-15 20:12:53 +00:00
def __init__(self, fp):
self._fp = fp
self.handlers = []
def add_handler(self, handler):
self.handlers.append(handler)
def read(self, n=-1):
buf = self._fp.read(n)
if not buf:
return ''
for handler in self.handlers:
handler(buf)
return buf
2013-09-25 21:50:03 +00:00
2014-08-12 06:06:44 +00:00
def tell(self):
raise IOError('Stream is not seekable.')
2013-09-25 21:50:03 +00:00
def image_is_uploading(repo_image):
if repo_image is None:
return False
return repo_image.storage.uploading
def set_uploading_flag(repo_image, is_image_uploading):
repo_image.storage.uploading = is_image_uploading
repo_image.storage.save()
2013-09-25 21:50:03 +00:00
def require_completion(f):
"""This make sure that the image push correctly finished."""
@wraps(f)
def wrapper(namespace, repository, *args, **kwargs):
image_id = kwargs['image_id']
repo_image = model.get_repo_image(namespace, repository, image_id)
if image_is_uploading(repo_image):
abort(400, 'Image %(image_id)s is being uploaded, retry later',
issue='upload-in-progress', image_id=kwargs['image_id'])
return f(namespace, repository, *args, **kwargs)
2013-09-25 21:50:03 +00:00
return wrapper
def set_cache_headers(f):
"""Returns HTTP headers suitable for caching."""
@wraps(f)
def wrapper(*args, **kwargs):
# Set TTL to 1 year by default
ttl = 31536000
expires = datetime.fromtimestamp(int(time()) + ttl)
expires = expires.strftime('%a, %d %b %Y %H:%M:%S GMT')
headers = {
'Cache-Control': 'public, max-age={0}'.format(ttl),
'Expires': expires,
'Last-Modified': 'Thu, 01 Jan 1970 00:00:00 GMT',
}
if 'If-Modified-Since' in request.headers:
response = make_response('Not modified', 304)
response.headers.extend(headers)
return response
kwargs['headers'] = headers
# Prevent the Cookie to be sent when the object is cacheable
session.modified = False
return f(*args, **kwargs)
return wrapper
@registry.route('/images/<image_id>/layer', methods=['HEAD'])
2013-09-25 21:50:03 +00:00
@process_auth
@extract_namespace_repo_from_session
2013-09-25 21:50:03 +00:00
@require_completion
@set_cache_headers
def head_image_layer(namespace, repository, image_id, headers):
permission = ReadRepositoryPermission(namespace, repository)
profile.debug('Checking repo permissions')
2013-09-28 04:05:32 +00:00
if permission.can() or model.repository_is_public(namespace, repository):
profile.debug('Looking up repo image')
repo_image = model.get_repo_image(namespace, repository, image_id)
if not repo_image:
profile.debug('Image not found')
abort(404, 'Image %(image_id)s not found', issue='unknown-image',
image_id=image_id)
extra_headers = {}
# Add the Accept-Ranges header if the storage engine supports resumable
# downloads.
if store.get_supports_resumable_downloads(repo_image.storage.locations):
profile.debug('Storage supports resumable downloads')
2014-07-07 22:19:52 +00:00
extra_headers['Accept-Ranges'] = 'bytes'
resp = make_response('')
2014-07-07 22:19:52 +00:00
resp.headers.extend(headers)
resp.headers.extend(extra_headers)
return resp
2014-07-07 22:19:52 +00:00
abort(403)
@registry.route('/images/<image_id>/layer', methods=['GET'])
@process_auth
@extract_namespace_repo_from_session
@require_completion
@set_cache_headers
def get_image_layer(namespace, repository, image_id, headers):
permission = ReadRepositoryPermission(namespace, repository)
profile.debug('Checking repo permissions')
if permission.can() or model.repository_is_public(namespace, repository):
profile.debug('Looking up repo image')
repo_image = model.get_repo_image(namespace, repository, image_id)
profile.debug('Looking up the layer path')
try:
path = store.image_layer_path(repo_image.storage.uuid)
profile.debug('Looking up the direct download URL')
direct_download_url = store.get_direct_download_url(repo_image.storage.locations, path)
if direct_download_url:
profile.debug('Returning direct download URL')
resp = redirect(direct_download_url)
return resp
profile.debug('Streaming layer data')
2014-07-07 22:19:52 +00:00
return Response(store.stream_read(repo_image.storage.locations, path), headers=headers)
except (IOError, AttributeError):
profile.debug('Image not found')
abort(404, 'Image %(image_id)s not found', issue='unknown-image',
image_id=image_id)
2014-07-07 22:19:52 +00:00
2013-09-28 04:05:32 +00:00
abort(403)
2013-09-25 21:50:03 +00:00
@registry.route('/images/<image_id>/layer', methods=['PUT'])
2013-09-25 21:50:03 +00:00
@process_auth
@extract_namespace_repo_from_session
def put_image_layer(namespace, repository, image_id):
profile.debug('Checking repo permissions')
permission = ModifyRepositoryPermission(namespace, repository)
if not permission.can():
abort(403)
profile.debug('Retrieving image')
repo_image = model.get_repo_image(namespace, repository, image_id)
2013-09-25 21:50:03 +00:00
try:
profile.debug('Retrieving image data')
uuid = repo_image.storage.uuid
json_data = store.get_content(repo_image.storage.locations, store.image_json_path(uuid))
except (IOError, AttributeError):
abort(404, 'Image %(image_id)s not found', issue='unknown-image',
image_id=image_id)
profile.debug('Retrieving image path info')
layer_path = store.image_layer_path(uuid)
if (store.exists(repo_image.storage.locations, layer_path) and not
image_is_uploading(repo_image)):
exact_abort(409, 'Image already exists')
profile.debug('Storing layer data')
2013-09-25 21:50:03 +00:00
input_stream = request.stream
if request.headers.get('transfer-encoding') == 'chunked':
# Careful, might work only with WSGI servers supporting chunked
# encoding (Gunicorn)
input_stream = request.environ['wsgi.input']
2013-09-25 21:50:03 +00:00
# compute checksums
csums = []
sr = SocketReader(input_stream)
2013-11-11 21:41:33 +00:00
tmp, store_hndlr = store.temp_store_handler()
2013-09-25 21:50:03 +00:00
sr.add_handler(store_hndlr)
h, sum_hndlr = checksums.simple_checksum_handler(json_data)
sr.add_handler(sum_hndlr)
store.stream_write(repo_image.storage.locations, layer_path, sr)
2013-09-25 21:50:03 +00:00
csums.append('sha256:{0}'.format(h.hexdigest()))
2013-09-25 21:50:03 +00:00
try:
image_size = tmp.tell()
# Save the size of the image.
model.set_image_size(image_id, namespace, repository, image_size)
2013-09-25 21:50:03 +00:00
tmp.seek(0)
csums.append(checksums.compute_tarsum(tmp, json_data))
tmp.close()
except (IOError, checksums.TarError) as e:
logger.debug('put_image_layer: Error when computing tarsum '
2013-10-08 15:29:42 +00:00
'{0}'.format(e))
if repo_image.storage.checksum is None:
2013-09-25 21:50:03 +00:00
# We don't have a checksum stored yet, that's fine skipping the check.
# Not removing the mark though, image is not downloadable yet.
session['checksum'] = csums
return make_response('true', 200)
checksum = repo_image.storage.checksum
2013-09-25 21:50:03 +00:00
# We check if the checksums provided matches one the one we computed
if checksum not in csums:
logger.warning('put_image_layer: Wrong checksum')
abort(400, 'Checksum mismatch; ignoring the layer for image %(image_id)s',
issue='checksum-mismatch', image_id=image_id)
2013-09-25 21:50:03 +00:00
# Checksum is ok, we remove the marker
set_uploading_flag(repo_image, False)
# The layer is ready for download, send a job to the work queue to
# process it.
profile.debug('Adding layer to diff queue')
image_diff_queue.put([namespace, repository, image_id], json.dumps({
'namespace': namespace,
'repository': repository,
'image_id': image_id,
}))
2013-09-25 21:50:03 +00:00
return make_response('true', 200)
@registry.route('/images/<image_id>/checksum', methods=['PUT'])
2013-09-25 21:50:03 +00:00
@process_auth
@extract_namespace_repo_from_session
def put_image_checksum(namespace, repository, image_id):
profile.debug('Checking repo permissions')
permission = ModifyRepositoryPermission(namespace, repository)
if not permission.can():
abort(403)
checksum = request.headers.get('X-Docker-Checksum')
2013-09-25 21:50:03 +00:00
if not checksum:
abort(400, "Missing checksum for image %(image_id)s", issue='missing-checksum',
image_id=image_id)
2013-09-25 21:50:03 +00:00
if not session.get('checksum'):
2014-05-08 15:50:38 +00:00
abort(400, 'Checksum not found in Cookie for image %(image_id)s',
issue='missing-checksum-cookie', image_id=image_id)
profile.debug('Looking up repo image')
repo_image = model.get_repo_image(namespace, repository, image_id)
uuid = repo_image.storage.uuid
profile.debug('Looking up repo layer data')
if not store.exists(repo_image.storage.locations, store.image_json_path(uuid)):
abort(404, 'Image not found: %(image_id)s', issue='unknown-image', image_id=image_id)
profile.debug('Marking image path')
if not image_is_uploading(repo_image):
abort(409, 'Cannot set checksum for image %(image_id)s',
issue='image-write-error', image_id=image_id)
profile.debug('Storing image checksum')
err = store_checksum(repo_image.storage, checksum)
2013-09-25 21:50:03 +00:00
if err:
abort(400, err)
2013-09-25 21:50:03 +00:00
if checksum not in session.get('checksum', []):
logger.debug('session checksums: %s' % session.get('checksum', []))
logger.debug('client supplied checksum: %s' % checksum)
logger.debug('put_image_checksum: Wrong checksum')
abort(400, 'Checksum mismatch for image: %(image_id)s',
issue='checksum-mismatch', image_id=image_id)
2013-09-25 21:50:03 +00:00
# Checksum is ok, we remove the marker
set_uploading_flag(repo_image, False)
# The layer is ready for download, send a job to the work queue to
# process it.
profile.debug('Adding layer to diff queue')
image_diff_queue.put([namespace, repository, image_id], json.dumps({
'namespace': namespace,
'repository': repository,
'image_id': image_id,
}))
2013-09-25 21:50:03 +00:00
return make_response('true', 200)
@registry.route('/images/<image_id>/json', methods=['GET'])
2013-09-25 21:50:03 +00:00
@process_auth
@extract_namespace_repo_from_session
2013-09-25 21:50:03 +00:00
@require_completion
@set_cache_headers
def get_image_json(namespace, repository, image_id, headers):
profile.debug('Checking repo permissions')
permission = ReadRepositoryPermission(namespace, repository)
2013-10-08 15:29:42 +00:00
if not permission.can() and not model.repository_is_public(namespace,
repository):
abort(403)
profile.debug('Looking up repo image')
repo_image = model.get_repo_image(namespace, repository, image_id)
profile.debug('Looking up repo layer data')
2013-09-25 21:50:03 +00:00
try:
uuid = repo_image.storage.uuid
data = store.get_content(repo_image.storage.locations, store.image_json_path(uuid))
except (IOError, AttributeError):
flask_abort(404)
profile.debug('Looking up repo layer size')
size = repo_image.storage.image_size
headers['X-Docker-Size'] = str(size)
2013-09-25 21:50:03 +00:00
response = make_response(data, 200)
response.headers.extend(headers)
return response
@registry.route('/images/<image_id>/ancestry', methods=['GET'])
2013-09-25 21:50:03 +00:00
@process_auth
@extract_namespace_repo_from_session
2013-09-25 21:50:03 +00:00
@require_completion
@set_cache_headers
def get_image_ancestry(namespace, repository, image_id, headers):
profile.debug('Checking repo permissions')
permission = ReadRepositoryPermission(namespace, repository)
2013-10-08 15:29:42 +00:00
if not permission.can() and not model.repository_is_public(namespace,
repository):
abort(403)
profile.debug('Looking up repo image')
repo_image = model.get_repo_image(namespace, repository, image_id)
profile.debug('Looking up image data')
2013-09-25 21:50:03 +00:00
try:
uuid = repo_image.storage.uuid
data = store.get_content(repo_image.storage.locations, store.image_ancestry_path(uuid))
except (IOError, AttributeError):
abort(404, 'Image %(image_id)s not found', issue='unknown-image',
image_id=image_id)
profile.debug('Converting to <-> from JSON')
response = make_response(json.dumps(json.loads(data)), 200)
2013-09-25 21:50:03 +00:00
response.headers.extend(headers)
profile.debug('Done')
2013-09-25 21:50:03 +00:00
return response
def generate_ancestry(image_id, uuid, locations, parent_id=None, parent_uuid=None,
parent_locations=None):
2013-09-25 21:50:03 +00:00
if not parent_id:
store.put_content(locations, store.image_ancestry_path(uuid), json.dumps([image_id]))
2013-09-25 21:50:03 +00:00
return
2014-08-18 17:35:03 +00:00
data = store.get_content(parent_locations, store.image_ancestry_path(parent_uuid))
2013-09-25 21:50:03 +00:00
data = json.loads(data)
data.insert(0, image_id)
store.put_content(locations, store.image_ancestry_path(uuid), json.dumps(data))
2013-09-25 21:50:03 +00:00
def store_checksum(image_storage, checksum):
2013-09-25 21:50:03 +00:00
checksum_parts = checksum.split(':')
if len(checksum_parts) != 2:
return 'Invalid checksum format'
2013-09-25 21:50:03 +00:00
# We store the checksum
image_storage.checksum = checksum
image_storage.save()
2013-09-25 21:50:03 +00:00
@registry.route('/images/<image_id>/json', methods=['PUT'])
2013-09-25 21:50:03 +00:00
@process_auth
@extract_namespace_repo_from_session
def put_image_json(namespace, repository, image_id):
profile.debug('Checking repo permissions')
permission = ModifyRepositoryPermission(namespace, repository)
if not permission.can():
abort(403)
profile.debug('Parsing image JSON')
2013-09-25 21:50:03 +00:00
try:
data = json.loads(request.data.decode('utf8'))
except ValueError:
2013-09-25 21:50:03 +00:00
pass
2013-09-25 21:50:03 +00:00
if not data or not isinstance(data, dict):
abort(400, 'Invalid JSON for image: %(image_id)s\nJSON: %(json)s',
issue='invalid-request', image_id=image_id, json=request.data)
2013-09-25 21:50:03 +00:00
if 'id' not in data:
abort(400, 'Missing key `id` in JSON for image: %(image_id)s',
issue='invalid-request', image_id=image_id)
profile.debug('Looking up repo image')
repo_image = model.get_repo_image(namespace, repository, image_id)
if not repo_image:
profile.debug('Image not found')
abort(404, 'Image %(image_id)s not found', issue='unknown-image',
image_id=image_id)
uuid = repo_image.storage.uuid
2013-09-25 21:50:03 +00:00
if image_id != data['id']:
abort(400, 'JSON data contains invalid id for image: %(image_id)s',
issue='invalid-request', image_id=image_id)
2013-09-25 21:50:03 +00:00
parent_id = data.get('parent')
parent_image = None
if parent_id:
profile.debug('Looking up parent image')
parent_image = model.get_repo_image(namespace, repository, parent_id)
parent_uuid = parent_image and parent_image.storage.uuid
parent_locations = parent_image and parent_image.storage.locations
if parent_id:
profile.debug('Looking up parent image data')
if (parent_id and not
store.exists(parent_locations, store.image_json_path(parent_uuid))):
abort(400, 'Image %(image_id)s depends on non existing parent image %(parent_id)s',
issue='invalid-request', image_id=image_id, parent_id=parent_id)
profile.debug('Looking up image storage paths')
json_path = store.image_json_path(uuid)
profile.debug('Checking if image already exists')
if (store.exists(repo_image.storage.locations, json_path) and not
image_is_uploading(repo_image)):
exact_abort(409, 'Image already exists')
set_uploading_flag(repo_image, True)
# We cleanup any old checksum in case it's a retry after a fail
profile.debug('Cleanup old checksum')
repo_image.storage.checksum = None
repo_image.storage.save()
2013-09-25 21:50:03 +00:00
# If we reach that point, it means that this is a new image or a retry
# on a failed push
# save the metadata
command_list = data.get('container_config', {}).get('Cmd', None)
command = json.dumps(command_list) if command_list else None
profile.debug('Setting image metadata')
model.set_image_metadata(image_id, namespace, repository,
data.get('created'), data.get('comment'), command,
parent_image)
profile.debug('Putting json path')
store.put_content(repo_image.storage.locations, json_path, request.data)
profile.debug('Generating image ancestry')
2014-08-18 17:35:03 +00:00
try:
generate_ancestry(image_id, uuid, repo_image.storage.locations, parent_id, parent_uuid,
parent_locations)
except IOError as ioe:
profile.debug('Error when generating ancestry: %s' % ioe.message)
abort(404)
profile.debug('Done')
2013-09-25 21:50:03 +00:00
return make_response('true', 200)
def process_image_changes(namespace, repository, image_id):
logger.debug('Generating diffs for image: %s' % image_id)
repo_image = model.get_repo_image(namespace, repository, image_id)
2014-06-30 17:51:42 +00:00
if not repo_image:
logger.warning('No image for id: %s', image_id)
return None, None
uuid = repo_image.storage.uuid
image_diffs_path = store.image_file_diffs_path(uuid)
image_trie_path = store.image_file_trie_path(uuid)
if store.exists(repo_image.storage.locations, image_diffs_path):
logger.debug('Diffs already exist for image: %s' % image_id)
2014-06-30 17:51:42 +00:00
return image_trie_path, repo_image.storage.locations
image = model.get_image_by_id(namespace, repository, image_id)
parents = model.get_parent_images(namespace, repository, image)
# Compute the diffs and fs for the parent first if necessary
parent_trie_path = None
if parents:
parent_trie_path, parent_locations = process_image_changes(namespace, repository,
parents[-1].docker_image_id)
# Read in the collapsed layer state of the filesystem for the parent
parent_trie = changes.empty_fs()
if parent_trie_path:
parent_trie_bytes = store.get_content(parent_locations, parent_trie_path)
parent_trie.frombytes(parent_trie_bytes)
# Read in the file entries from the layer tar file
layer_path = store.image_layer_path(uuid)
with store.stream_read_file(image.storage.locations, layer_path) as layer_tar_stream:
removed_files = set()
layer_files = changes.files_and_dirs_from_tar(layer_tar_stream,
removed_files)
new_metadata = changes.compute_new_diffs_and_fs(parent_trie, layer_files,
removed_files)
(new_trie, added, changed, removed) = new_metadata
# Write out the new trie
store.put_content(image.storage.locations, image_trie_path, new_trie.tobytes())
# Write out the diffs
diffs = {}
sections = ('added', 'changed', 'removed')
for section, source_trie in zip(sections, new_metadata[1:]):
diffs[section] = list(source_trie)
2013-10-18 21:05:51 +00:00
diffs[section].sort()
store.put_content(image.storage.locations, image_diffs_path, json.dumps(diffs, indent=2))
return image_trie_path, image.storage.locations