glance/glance/store/s3.py

735 lines
28 KiB
Python

# Copyright 2010 OpenStack Foundation
# All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""Storage backend for S3 or Storage Servers that follow the S3 Protocol"""
import hashlib
import httplib
import math
import re
import tempfile
import boto.exception
import eventlet
from oslo.config import cfg
import six
import six.moves.urllib.parse as urlparse
from glance.common import exception
from glance.common import utils
from glance import i18n
import glance.openstack.common.log as logging
from glance.openstack.common import units
import glance.store
import glance.store.base
import glance.store.location
LOG = logging.getLogger(__name__)
_LE = i18n._LE
_LI = i18n._LI
DEFAULT_LARGE_OBJECT_SIZE = 100 # 100M
DEFAULT_LARGE_OBJECT_CHUNK_SIZE = 10 # 10M
DEFAULT_LARGE_OBJECT_MIN_CHUNK_SIZE = 5 # 5M
DEFAULT_THREAD_POOLS = 10 # 10 pools
s3_opts = [
cfg.StrOpt('s3_store_host',
help=_('The host where the S3 server is listening.')),
cfg.StrOpt('s3_store_access_key', secret=True,
help=_('The S3 query token access key.')),
cfg.StrOpt('s3_store_secret_key', secret=True,
help=_('The S3 query token secret key.')),
cfg.StrOpt('s3_store_bucket',
help=_('The S3 bucket to be used to store the Glance data.')),
cfg.StrOpt('s3_store_object_buffer_dir',
help=_('The local directory where uploads will be staged '
'before they are transferred into S3.')),
cfg.BoolOpt('s3_store_create_bucket_on_put', default=False,
help=_('A boolean to determine if the S3 bucket should be '
'created on upload if it does not exist or if '
'an error should be returned to the user.')),
cfg.StrOpt('s3_store_bucket_url_format', default='subdomain',
help=_('The S3 calling format used to determine the bucket. '
'Either subdomain or path can be used.')),
cfg.IntOpt('s3_store_large_object_size',
default=DEFAULT_LARGE_OBJECT_SIZE,
help=_('What size, in MB, should S3 start chunking image files '
'and do a multipart upload in S3.')),
cfg.IntOpt('s3_store_large_object_chunk_size',
default=DEFAULT_LARGE_OBJECT_CHUNK_SIZE,
help=_('What multipart upload part size, in MB, should S3 use '
'when uploading parts. The size must be greater than or '
'equal to 5M.')),
cfg.IntOpt('s3_store_thread_pools', default=DEFAULT_THREAD_POOLS,
help=_('The number of thread pools to perform a multipart '
'upload in S3.')),
]
CONF = cfg.CONF
CONF.register_opts(s3_opts)
class UploadPart:
"""
The class for the upload part
"""
def __init__(self, mpu, fp, partnum, chunks):
self.mpu = mpu
self.partnum = partnum
self.fp = fp
self.size = 0
self.chunks = chunks
self.etag = {} # partnum -> etag
self.success = True
def run_upload(part):
"""
Upload the upload part into S3 and set returned etag and size
to its part info.
"""
pnum = part.partnum
bsize = part.chunks
LOG.info(_LI("Uploading upload part in S3 partnum=%(pnum)d, "
"size=%(bsize)d, key=%(key)s, UploadId=%(UploadId)s") %
{'pnum': pnum,
'bsize': bsize,
'key': part.mpu.key_name,
'UploadId': part.mpu.id})
try:
key = part.mpu.upload_part_from_file(part.fp,
part_num=part.partnum,
size=bsize)
part.etag[part.partnum] = key.etag
part.size = key.size
except boto.exception.BotoServerError as e:
status = e.status
reason = e.reason
LOG.error(_LE("Failed to upload part in S3 partnum=%(pnum)d, "
"size=%(bsize)d, status=%(status)d, "
"reason=%(reason)s") %
{'pnum': pnum,
'bsize': bsize,
'status': status,
'reason': reason})
part.success = False
except Exception as e:
LOG.error(_LE("Failed to upload part in S3 partnum=%(pnum)d, "
"size=%(bsize)d due to internal error: %(err)s") %
{'pnum': pnum,
'bsize': bsize,
'err': utils.exception_to_str(e)})
part.success = False
finally:
part.fp.close()
class StoreLocation(glance.store.location.StoreLocation):
"""
Class describing an S3 URI. An S3 URI can look like any of
the following:
s3://accesskey:secretkey@s3.amazonaws.com/bucket/key-id
s3+http://accesskey:secretkey@s3.amazonaws.com/bucket/key-id
s3+https://accesskey:secretkey@s3.amazonaws.com/bucket/key-id
The s3+https:// URIs indicate there is an HTTPS s3service URL
"""
def process_specs(self):
self.scheme = self.specs.get('scheme', 's3')
self.accesskey = self.specs.get('accesskey')
self.secretkey = self.specs.get('secretkey')
s3_host = self.specs.get('s3serviceurl')
self.bucket = self.specs.get('bucket')
self.key = self.specs.get('key')
if s3_host.startswith('https://'):
self.scheme = 's3+https'
s3_host = s3_host[8:].strip('/')
elif s3_host.startswith('http://'):
s3_host = s3_host[7:].strip('/')
self.s3serviceurl = s3_host.strip('/')
def _get_credstring(self):
if self.accesskey:
return '%s:%s@' % (self.accesskey, self.secretkey)
return ''
def get_uri(self):
return "%s://%s%s/%s/%s" % (
self.scheme,
self._get_credstring(),
self.s3serviceurl,
self.bucket,
self.key)
def parse_uri(self, uri):
"""
Parse URLs. This method fixes an issue where credentials specified
in the URL are interpreted differently in Python 2.6.1+ than prior
versions of Python.
Note that an Amazon AWS secret key can contain the forward slash,
which is entirely retarded, and breaks urlparse miserably.
This function works around that issue.
"""
# Make sure that URIs that contain multiple schemes, such as:
# s3://accesskey:secretkey@https://s3.amazonaws.com/bucket/key-id
# are immediately rejected.
if uri.count('://') != 1:
reason = _("URI cannot contain more than one occurrence "
"of a scheme. If you have specified a URI like "
"s3://accesskey:secretkey@"
"https://s3.amazonaws.com/bucket/key-id"
", you need to change it to use the "
"s3+https:// scheme, like so: "
"s3+https://accesskey:secretkey@"
"s3.amazonaws.com/bucket/key-id")
LOG.info(_LI("Invalid store uri: %s") % reason)
raise exception.BadStoreUri(message=reason)
pieces = urlparse.urlparse(uri)
assert pieces.scheme in ('s3', 's3+http', 's3+https')
self.scheme = pieces.scheme
path = pieces.path.strip('/')
netloc = pieces.netloc.strip('/')
entire_path = (netloc + '/' + path).strip('/')
if '@' in uri:
creds, path = entire_path.split('@')
cred_parts = creds.split(':')
try:
access_key = cred_parts[0]
secret_key = cred_parts[1]
# NOTE(jaypipes): Need to encode to UTF-8 here because of a
# bug in the HMAC library that boto uses.
# See: http://bugs.python.org/issue5285
# See: http://trac.edgewall.org/ticket/8083
access_key = access_key.encode('utf-8')
secret_key = secret_key.encode('utf-8')
self.accesskey = access_key
self.secretkey = secret_key
except IndexError:
reason = _("Badly formed S3 credentials")
LOG.info(reason)
raise exception.BadStoreUri(message=reason)
else:
self.accesskey = None
path = entire_path
try:
path_parts = path.split('/')
self.key = path_parts.pop()
self.bucket = path_parts.pop()
if path_parts:
self.s3serviceurl = '/'.join(path_parts).strip('/')
else:
reason = _("Badly formed S3 URI. Missing s3 service URL.")
raise exception.BadStoreUri(message=reason)
except IndexError:
reason = _("Badly formed S3 URI")
LOG.info(reason)
raise exception.BadStoreUri(message=reason)
class ChunkedFile(object):
"""
We send this back to the Glance API server as
something that can iterate over a ``boto.s3.key.Key``
"""
def __init__(self, fp):
self.fp = fp
def __iter__(self):
"""Return an iterator over the image file"""
try:
if self.fp:
while True:
chunk = self.fp.read(Store.READ_CHUNKSIZE)
if chunk:
yield chunk
else:
break
finally:
self.close()
def getvalue(self):
"""Return entire string value... used in testing."""
data = ""
self.len = 0
for chunk in self:
read_bytes = len(chunk)
data = data + chunk
self.len = self.len + read_bytes
return data
def close(self):
"""Close the internal file pointer."""
if self.fp:
self.fp.close()
self.fp = None
class Store(glance.store.base.Store):
"""An implementation of the s3 adapter."""
READ_CHUNKSIZE = 64 * units.Ki
WRITE_CHUNKSIZE = READ_CHUNKSIZE
EXAMPLE_URL = "s3://<ACCESS_KEY>:<SECRET_KEY>@<S3_URL>/<BUCKET>/<OBJ>"
def get_schemes(self):
return ('s3', 's3+http', 's3+https')
def configure_add(self):
"""
Configure the Store to use the stored configuration options
Any store that needs special configuration should implement
this method. If the store was not able to successfully configure
itself, it should raise `exception.BadStoreConfiguration`
"""
self.s3_host = self._option_get('s3_store_host')
access_key = self._option_get('s3_store_access_key')
secret_key = self._option_get('s3_store_secret_key')
# NOTE(jaypipes): Need to encode to UTF-8 here because of a
# bug in the HMAC library that boto uses.
# See: http://bugs.python.org/issue5285
# See: http://trac.edgewall.org/ticket/8083
self.access_key = access_key.encode('utf-8')
self.secret_key = secret_key.encode('utf-8')
self.bucket = self._option_get('s3_store_bucket')
self.scheme = 's3'
if self.s3_host.startswith('https://'):
self.scheme = 's3+https'
self.full_s3_host = self.s3_host
elif self.s3_host.startswith('http://'):
self.full_s3_host = self.s3_host
else: # Defaults http
self.full_s3_host = 'http://' + self.s3_host
self.s3_store_object_buffer_dir = CONF.s3_store_object_buffer_dir
_s3_obj_size = CONF.s3_store_large_object_size
self.s3_store_large_object_size = _s3_obj_size * units.Mi
_s3_ck_size = CONF.s3_store_large_object_chunk_size
_s3_ck_min = DEFAULT_LARGE_OBJECT_MIN_CHUNK_SIZE
if _s3_ck_size < _s3_ck_min:
reason = (_("s3_store_large_object_chunk_size must be at "
"least %(_s3_ck_min)d MB. "
"You configured it as %(_s3_ck_size)d MB") %
{'_s3_ck_min': _s3_ck_min,
'_s3_ck_size': _s3_ck_size})
LOG.error(reason)
raise exception.BadStoreConfiguration(store_name="s3",
reason=reason)
self.s3_store_large_object_chunk_size = _s3_ck_size * units.Mi
if CONF.s3_store_thread_pools <= 0:
reason = (_("s3_store_thread_pools must be a positive "
"integer. %s") % CONF.s3_store_thread_pools)
LOG.error(reason)
raise exception.BadStoreConfiguration(store_name="s3",
reason=reason)
def _option_get(self, param):
result = getattr(CONF, param)
if not result:
reason = ("Could not find %(param)s in configuration "
"options." % {'param': param})
LOG.debug(reason)
raise exception.BadStoreConfiguration(store_name="s3",
reason=reason)
return result
def get(self, location):
"""
Takes a `glance.store.location.Location` object that indicates
where to find the image file, and returns a tuple of generator
(for reading the image file) and image_size
:param location `glance.store.location.Location` object, supplied
from glance.store.location.get_location_from_uri()
:raises `glance.exception.NotFound` if image does not exist
"""
key = self._retrieve_key(location)
key.BufferSize = self.READ_CHUNKSIZE
class ChunkedIndexable(glance.store.Indexable):
def another(self):
return (self.wrapped.fp.read(self.READ_CHUNKSIZE)
if self.wrapped.fp else None)
return (ChunkedIndexable(ChunkedFile(key), key.size), key.size)
def get_size(self, location):
"""
Takes a `glance.store.location.Location` object that indicates
where to find the image file, and returns the image_size (or 0
if unavailable)
:param location `glance.store.location.Location` object, supplied
from glance.store.location.get_location_from_uri()
"""
try:
key = self._retrieve_key(location)
return key.size
except Exception:
return 0
def _retrieve_key(self, location):
loc = location.store_location
from boto.s3.connection import S3Connection
s3_conn = S3Connection(loc.accesskey, loc.secretkey,
host=loc.s3serviceurl,
is_secure=(loc.scheme == 's3+https'),
calling_format=get_calling_format())
bucket_obj = get_bucket(s3_conn, loc.bucket)
key = get_key(bucket_obj, loc.key)
msg = ("Retrieved image object from S3 using (s3_host=%(s3_host)s, "
"access_key=%(accesskey)s, bucket=%(bucket)s, "
"key=%(obj_name)s)" % ({'s3_host': loc.s3serviceurl,
'accesskey': loc.accesskey,
'bucket': loc.bucket,
'obj_name': loc.key}))
LOG.debug(msg)
return key
def add(self, image_id, image_file, image_size):
"""
Stores an image file with supplied identifier to the backend
storage system and returns a tuple containing information
about the stored image.
:param image_id: The opaque image identifier
:param image_file: The image data to write, as a file-like object
:param image_size: The size of the image data to write, in bytes
:retval tuple of URL in backing store, bytes written, checksum
and a dictionary with storage system specific information
:raises `glance.common.exception.Duplicate` if the image already
existed
S3 writes the image data using the scheme:
s3://<ACCESS_KEY>:<SECRET_KEY>@<S3_URL>/<BUCKET>/<OBJ>
where:
<USER> = ``s3_store_user``
<KEY> = ``s3_store_key``
<S3_HOST> = ``s3_store_host``
<BUCKET> = ``s3_store_bucket``
<ID> = The id of the image being added
"""
from boto.s3.connection import S3Connection
loc = StoreLocation({'scheme': self.scheme,
'bucket': self.bucket,
'key': image_id,
's3serviceurl': self.full_s3_host,
'accesskey': self.access_key,
'secretkey': self.secret_key})
s3_conn = S3Connection(loc.accesskey, loc.secretkey,
host=loc.s3serviceurl,
is_secure=(loc.scheme == 's3+https'),
calling_format=get_calling_format())
create_bucket_if_missing(self.bucket, s3_conn)
bucket_obj = get_bucket(s3_conn, self.bucket)
obj_name = str(image_id)
def _sanitize(uri):
return re.sub('//.*:.*@',
'//s3_store_secret_key:s3_store_access_key@',
uri)
key = bucket_obj.get_key(obj_name)
if key and key.exists():
raise exception.Duplicate(_("S3 already has an image at "
"location %s") %
_sanitize(loc.get_uri()))
msg = ("Adding image object to S3 using (s3_host=%(s3_host)s, "
"access_key=%(access_key)s, bucket=%(bucket)s, "
"key=%(obj_name)s)" % ({'s3_host': self.s3_host,
'access_key': self.access_key,
'bucket': self.bucket,
'obj_name': obj_name}))
LOG.debug(msg)
LOG.debug("Uploading an image file to S3 for %s" %
_sanitize(loc.get_uri()))
if image_size < self.s3_store_large_object_size:
key = bucket_obj.new_key(obj_name)
# We need to wrap image_file, which is a reference to the
# webob.Request.body_file, with a seekable file-like object,
# otherwise the call to set_contents_from_file() will die
# with an error about Input object has no method 'seek'. We
# might want to call webob.Request.make_body_seekable(), but
# unfortunately, that method copies the entire image into
# memory and results in LP Bug #818292 occurring. So, here
# we write temporary file in as memory-efficient manner as
# possible and then supply the temporary file to S3. We also
# take this opportunity to calculate the image checksum while
# writing the tempfile, so we don't need to call key.compute_md5()
msg = ("Writing request body file to temporary file "
"for %s") % _sanitize(loc.get_uri())
LOG.debug(msg)
tmpdir = self.s3_store_object_buffer_dir
temp_file = tempfile.NamedTemporaryFile(dir=tmpdir)
checksum = hashlib.md5()
for chunk in utils.chunkreadable(image_file, self.WRITE_CHUNKSIZE):
checksum.update(chunk)
temp_file.write(chunk)
temp_file.flush()
msg = ("Uploading temporary file to S3 "
"for %s") % _sanitize(loc.get_uri())
LOG.debug(msg)
# OK, now upload the data into the key
key.set_contents_from_file(open(temp_file.name, 'rb'),
replace=False)
size = key.size
checksum_hex = checksum.hexdigest()
LOG.debug("Wrote %(size)d bytes to S3 key named %(obj_name)s "
"with checksum %(checksum_hex)s" %
{'size': size,
'obj_name': obj_name,
'checksum_hex': checksum_hex})
return (loc.get_uri(), size, checksum_hex, {})
else:
checksum = hashlib.md5()
parts = int(math.ceil(float(image_size) /
float(self.s3_store_large_object_chunk_size)))
threads = parts
pool_size = CONF.s3_store_thread_pools
pool = eventlet.greenpool.GreenPool(size=pool_size)
mpu = bucket_obj.initiate_multipart_upload(obj_name)
LOG.debug("Multipart initiate key=%(obj_name)s, "
"UploadId=%(UploadId)s" %
{'obj_name': obj_name,
'UploadId': mpu.id})
cstart = 0
plist = []
it = utils.chunkreadable(image_file,
self.s3_store_large_object_chunk_size)
for p in range(threads):
chunk = next(it)
clen = len(chunk)
checksum.update(chunk)
fp = six.BytesIO(chunk)
fp.seek(0)
part = UploadPart(mpu, fp, cstart + 1, clen)
pool.spawn_n(run_upload, part)
plist.append(part)
cstart += 1
pedict = {}
total_size = 0
pool.waitall()
for part in plist:
pedict.update(part.etag)
total_size += part.size
success = True
for part in plist:
if not part.success:
success = False
if success:
# Complete
xml = get_mpu_xml(pedict)
bucket_obj.complete_multipart_upload(obj_name,
mpu.id,
xml)
checksum_hex = checksum.hexdigest()
LOG.info(_LI("Multipart complete key=%(obj_name)s "
"UploadId=%(UploadId)s "
"Wrote %(total_size)d bytes to S3 key"
"named %(obj_name)s "
"with checksum %(checksum_hex)s") %
{'obj_name': obj_name,
'UploadId': mpu.id,
'total_size': total_size,
'obj_name': obj_name,
'checksum_hex': checksum_hex})
return (loc.get_uri(), total_size, checksum_hex, {})
else:
# Abort
bucket_obj.cancel_multipart_upload(obj_name, mpu.id)
LOG.error(_LE("Some parts failed to upload to S3. "
"Aborted the object key=%(obj_name)s") %
{'obj_name': obj_name})
msg = (_("Failed to add image object to S3. "
"key=%(obj_name)s") % {'obj_name': obj_name})
raise glance.store.BackendException(msg)
def delete(self, location):
"""
Takes a `glance.store.location.Location` object that indicates
where to find the image file to delete
:location `glance.store.location.Location` object, supplied
from glance.store.location.get_location_from_uri()
:raises NotFound if image does not exist
"""
loc = location.store_location
from boto.s3.connection import S3Connection
s3_conn = S3Connection(loc.accesskey, loc.secretkey,
host=loc.s3serviceurl,
is_secure=(loc.scheme == 's3+https'),
calling_format=get_calling_format())
bucket_obj = get_bucket(s3_conn, loc.bucket)
# Close the key when we're through.
key = get_key(bucket_obj, loc.key)
msg = ("Deleting image object from S3 using (s3_host=%(s3_host)s, "
"access_key=%(accesskey)s, bucket=%(bucket)s, "
"key=%(obj_name)s)" % ({'s3_host': loc.s3serviceurl,
'accesskey': loc.accesskey,
'bucket': loc.bucket,
'obj_name': loc.key}))
LOG.debug(msg)
return key.delete()
def get_bucket(conn, bucket_id):
"""
Get a bucket from an s3 connection
:param conn: The ``boto.s3.connection.S3Connection``
:param bucket_id: ID of the bucket to fetch
:raises ``glance.exception.NotFound`` if bucket is not found.
"""
bucket = conn.get_bucket(bucket_id)
if not bucket:
msg = "Could not find bucket with ID %s" % bucket_id
LOG.debug(msg)
raise exception.NotFound(msg)
return bucket
def get_s3_location(s3_host):
from boto.s3.connection import Location
locations = {
's3.amazonaws.com': Location.DEFAULT,
's3-eu-west-1.amazonaws.com': Location.EU,
's3-us-west-1.amazonaws.com': Location.USWest,
's3-ap-southeast-1.amazonaws.com': Location.APSoutheast,
's3-ap-northeast-1.amazonaws.com': Location.APNortheast,
}
# strip off scheme and port if present
key = re.sub('^(https?://)?(?P<host>[^:]+)(:[0-9]+)?$',
'\g<host>',
s3_host)
return locations.get(key, Location.DEFAULT)
def create_bucket_if_missing(bucket, s3_conn):
"""
Creates a missing bucket in S3 if the
``s3_store_create_bucket_on_put`` option is set.
:param bucket: Name of bucket to create
:param s3_conn: Connection to S3
"""
from boto.exception import S3ResponseError
try:
s3_conn.get_bucket(bucket)
except S3ResponseError as e:
if e.status == httplib.NOT_FOUND:
if CONF.s3_store_create_bucket_on_put:
location = get_s3_location(CONF.s3_store_host)
try:
s3_conn.create_bucket(bucket, location=location)
except S3ResponseError as e:
msg = (_("Failed to add bucket to S3.\n"
"Got error from S3: %s") %
utils.exception_to_str(e))
raise glance.store.BackendException(msg)
else:
msg = (_("The bucket %(bucket)s does not exist in "
"S3. Please set the "
"s3_store_create_bucket_on_put option "
"to add bucket to S3 automatically.")
% {'bucket': bucket})
raise glance.store.BackendException(msg)
def get_key(bucket, obj):
"""
Get a key from a bucket
:param bucket: The ``boto.s3.Bucket``
:param obj: Object to get the key for
:raises ``glance.exception.NotFound`` if key is not found.
"""
key = bucket.get_key(obj)
if not key or not key.exists():
msg = ("Could not find key %(obj)s in bucket %(bucket)s" %
{'obj': obj, 'bucket': bucket})
LOG.debug(msg)
raise exception.NotFound(msg)
return key
def get_calling_format(bucket_format=None):
import boto.s3.connection
if bucket_format is None:
bucket_format = CONF.s3_store_bucket_url_format
if bucket_format.lower() == 'path':
return boto.s3.connection.OrdinaryCallingFormat()
else:
return boto.s3.connection.SubdomainCallingFormat()
def get_mpu_xml(pedict):
xml = '<CompleteMultipartUpload>\n'
for pnum, etag in pedict.iteritems():
xml += ' <Part>\n'
xml += ' <PartNumber>%d</PartNumber>\n' % pnum
xml += ' <ETag>%s</ETag>\n' % etag
xml += ' </Part>\n'
xml += '</CompleteMultipartUpload>'
return xml