Get squashed endpoint for docker import working

This commit is contained in:
Joseph Schorr 2014-09-16 11:53:54 -04:00
parent e3c52fa0eb
commit 9344839295
3 changed files with 4 additions and 136 deletions

View file

@ -44,7 +44,7 @@ application.register_blueprint(callback, url_prefix='/oauth2')
application.register_blueprint(index, url_prefix='/v1')
application.register_blueprint(tags, url_prefix='/v1')
application.register_blueprint(registry, url_prefix='/v1')
application.register_blueprint(verbs, url_prefix='/v1/repositories')
application.register_blueprint(verbs, url_prefix='/verbs/v1/repositories')
application.register_blueprint(api_bp, url_prefix='/api')
application.register_blueprint(webhooks, url_prefix='/webhooks')
application.register_blueprint(realtime, url_prefix='/realtime')

View file

@ -11,7 +11,8 @@ from auth.permissions import ReadRepositoryPermission
from data import model
from endpoints.registry import set_cache_headers
from util.dockerimportformat import build_docker_import_stream
from util.gzipwrap import GzipWrap
from util.streamlayerformat import StreamLayerMerger
from werkzeug.wsgi import wrap_file
@ -46,14 +47,6 @@ def get_squashed_tag(namespace, repository, tag, headers):
ancestry_data = store.get_content(repo_image.storage.locations, store.image_ancestry_path(uuid))
full_image_list = json.loads(ancestry_data)
# Load the JSON for the image.
json_data = store.get_content(repo_image.storage.locations, store.image_json_path(uuid))
layer_json = json.loads(json_data)
def get_next_image():
for current_image_id in full_image_list:
yield model.get_repo_image(namespace, repository, current_image_id)
def get_next_layer():
for current_image_id in full_image_list:
current_image_entry = model.get_repo_image(namespace, repository, current_image_id)
@ -64,9 +57,7 @@ def get_squashed_tag(namespace, repository, tag, headers):
logger.debug('Returning image layer %s: %s' % (current_image_id, current_image_path))
yield current_image_stream
stream = build_docker_import_stream(namespace, repository, tag, synthetic_image_id,
layer_json, get_next_image, get_next_layer)
stream = GzipWrap(StreamLayerMerger(get_next_layer).get_generator())
return app.response_class(wrap_file(request.environ, stream, 1024 * 16),
mimetype='application/octet-stream',
direct_passthrough=True)

View file

@ -1,123 +0,0 @@
from util.gzipwrap import GzipWrap
from util.streamlayerformat import StreamLayerMerger
import copy
import json
import tarfile
from itertools import chain, islice
class some_magic_adaptor(object):
def __init__(self, src):
self.src = chain.from_iterable(src)
def read(self, n):
return "".join(islice(self.src, None, n))
def build_docker_import_stream(namespace, repository, tag, synthetic_image_id,
layer_json, get_image_iterator, get_layer_iterator):
""" Builds and streams a synthetic .tar.gz that represents a squashed version
of the given layers, in `docker import` V1 format.
"""
return some_magic_adaptor(_import_format_generator(namespace, repository, tag,
synthetic_image_id, layer_json,
get_image_iterator, get_layer_iterator))
def _import_format_generator(namespace, repository, tag, synthetic_image_id,
layer_json, get_image_iterator, get_layer_iterator):
# Docker import V1 Format (.tar):
# repositories - JSON file containing a repo -> tag -> image map
# {image ID folder}:
# json - The layer JSON
# layer.tar - The TARed contents of the layer
# VERSION - The docker import version: '1.0'
layer_merger = StreamLayerMerger(get_layer_iterator)
# Yield the repositories file:
synthetic_layer_info = {}
synthetic_layer_info[tag + '.squash'] = synthetic_image_id
repositories = {}
repositories[namespace + '/' + repository] = synthetic_layer_info
yield _tar_file('repositories', json.dumps(repositories))
# Yield the image ID folder.
yield _tar_folder(synthetic_image_id)
# Yield the JSON layer data.
layer_json = _build_layer_json(layer_json, synthetic_image_id)
yield _tar_file(synthetic_image_id + '/json', json.dumps(layer_json))
# Yield the VERSION file.
yield _tar_file(synthetic_image_id + '/VERSION', '1.0')
# Yield the merged layer data's header.
estimated_file_size = 0
for image in get_image_iterator():
estimated_file_size += image.storage.uncompressed_size or 0
yield _tar_file_header(synthetic_image_id + '/layer.tar', estimated_file_size)
# Yield the contents of the merged layer.
yielded_size = 0
for entry in layer_merger.get_generator():
yield entry
yielded_size += len(entry)
# If the yielded size is less than the estimated size (which is likely), fill the rest with
# zeros.
if yielded_size < estimated_file_size:
yield '\0' * (estimated_file_size - yielded_size)
print estimated_file_size
print yielded_size
# Yield any file padding to 512 bytes that is necessary.
yield _tar_file_padding(estimated_file_size)
# Last two records are empty in TAR spec.
yield '\0' * 512
yield '\0' * 512
def _build_layer_json(layer_json, synthetic_image_id):
updated_json = copy.deepcopy(layer_json)
updated_json['id'] = synthetic_image_id
if 'parent' in updated_json:
del updated_json['parent']
if 'config' in updated_json and 'Image' in updated_json['config']:
updated_json['config']['Image'] = synthetic_image_id
if 'container_config' in updated_json and 'Image' in updated_json['container_config']:
updated_json['container_config']['Image'] = synthetic_image_id
return updated_json
def _tar_file(name, contents):
length = len(contents)
tar_data = _tar_file_header(name, length)
tar_data += contents
tar_data += _tar_file_padding(length)
return tar_data
def _tar_file_padding(length):
if length % 512 != 0:
return '\0' * (512 - (length % 512))
def _tar_file_header(name, file_size):
info = tarfile.TarInfo(name=name)
info.type = tarfile.REGTYPE
info.size = file_size
return info.tobuf()
def _tar_folder(name):
info = tarfile.TarInfo(name=name)
info.type = tarfile.DIRTYPE
return info.tobuf()