quay/storage/cloud.py

import cStringIO as StringIO
import os
import logging
import copy

from boto.exception import S3ResponseError
import boto.s3.connection
import boto.s3.multipart
import boto.gs.connection
import boto.s3.key
import boto.gs.key

from io import BufferedIOBase
from uuid import uuid4
from collections import namedtuple

from util.registry import filelike
from storage.basestorage import BaseStorageV2


logger = logging.getLogger(__name__)

_PartUploadMetadata = namedtuple('_PartUploadMetadata', ['path', 'offset', 'length'])
_CHUNKS_KEY = 'chunks'


class StreamReadKeyAsFile(BufferedIOBase):
  def __init__(self, key):
    self._key = key

  def read(self, amt=None):
    if self.closed:
      return None

    resp = self._key.read(amt)
    return resp

  def readable(self):
    return True

  @property
  def closed(self):
    return self._key.closed

  def close(self):
    self._key.close(fast=True)


class _CloudStorage(BaseStorageV2):
  def __init__(self, metric_queue, connection_class, key_class, connect_kwargs, upload_params,
               storage_path, bucket_name, access_key=None, secret_key=None):
    super(_CloudStorage, self).__init__()

    self.automatic_chunk_size = 5 * 1024 * 1024

    self._initialized = False
    self._bucket_name = bucket_name
    self._access_key = access_key
    self._secret_key = secret_key
    self._root_path = storage_path
    self._connection_class = connection_class
    self._key_class = key_class
    self._upload_params = upload_params
    self._connect_kwargs = connect_kwargs
    self._cloud_conn = None
    self._cloud_bucket = None
    self._metric_queue = metric_queue

  def _initialize_cloud_conn(self):
    if not self._initialized:
      self._cloud_conn = self._connection_class(self._access_key, self._secret_key,
                                                **self._connect_kwargs)
      self._cloud_bucket = self._cloud_conn.get_bucket(self._bucket_name)
      self._initialized = True

  def _debug_key(self, key):
    """Used for debugging only."""
    orig_meth = key.bucket.connection.make_request

    def new_meth(*args, **kwargs):
      print '#' * 16
      print args
      print kwargs
      print '#' * 16
      return orig_meth(*args, **kwargs)
    key.bucket.connection.make_request = new_meth

  def _init_path(self, path=None):
    path = os.path.join(self._root_path, path) if path else self._root_path
    if path and path[0] == '/':
      return path[1:]
    return path

  def get_cloud_conn(self):
    self._initialize_cloud_conn()
    return self._cloud_conn

  def get_cloud_bucket(self):
    return self._cloud_bucket

  def get_content(self, path):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    if not key.exists():
      raise IOError('No such key: \'{0}\''.format(path))
    return key.get_contents_as_string()

  def put_content(self, path, content):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    key.set_contents_from_string(content, **self._upload_params)
    return path

  def get_supports_resumable_downloads(self):
    return True

  def get_direct_download_url(self, path, expires_in=60, requires_cors=False, head=False):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    k = self._key_class(self._cloud_bucket, path)
    if head:
      return k.generate_url(expires_in, 'HEAD')
    return k.generate_url(expires_in)

  def get_direct_upload_url(self, path, mime_type, requires_cors=True):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    url = key.generate_url(300, 'PUT', headers={'Content-Type': mime_type}, encrypt_key=True)
    return url

  def stream_read(self, path):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    if not key.exists():
      raise IOError('No such key: \'{0}\''.format(path))
    while True:
      buf = key.read(self.buffer_size)
      if not buf:
        break
      yield buf

  def stream_read_file(self, path):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    if not key.exists():
      raise IOError('No such key: \'{0}\''.format(path))
    return StreamReadKeyAsFile(key)

  def __initiate_multipart_upload(self, path, content_type, content_encoding):
    # Minimum size of upload part size on S3 is 5MB
    self._initialize_cloud_conn()
    path = self._init_path(path)

    metadata = {}
    if content_type is not None:
      metadata['Content-Type'] = content_type

    if content_encoding is not None:
      metadata['Content-Encoding'] = content_encoding

    if self._metric_queue is not None:
      self._metric_queue.put_deprecated('MultipartUploadStart', 1)
      self._metric_queue.multipart_upload_start.Inc()

    return self._cloud_bucket.initiate_multipart_upload(path, metadata=metadata,
                                                        **self._upload_params)

  def stream_write(self, path, fp, content_type=None, content_encoding=None):
    self._stream_write_internal(path, fp, content_type, content_encoding)

  def _stream_write_internal(self, path, fp, content_type=None, content_encoding=None,
                             cancel_on_error=True, size=filelike.READ_UNTIL_END):
    write_error = None
    mp = self.__initiate_multipart_upload(path, content_type, content_encoding)

    # We are going to reuse this but be VERY careful to only read the number of bytes written to it
    buf = StringIO.StringIO()

    num_part = 1
    total_bytes_written = 0
    while size == filelike.READ_UNTIL_END or total_bytes_written < size:
      bytes_to_copy = self.automatic_chunk_size
      if size != filelike.READ_UNTIL_END:
        # We never want to ask for more bytes than our caller has indicated to copy
        bytes_to_copy = min(bytes_to_copy, size - total_bytes_written)

      buf.seek(0)
      try:
        # Stage the bytes into the buffer for use with the multipart upload file API
        bytes_staged = self.stream_write_to_fp(fp, buf, bytes_to_copy)
        if bytes_staged == 0:
          break

        buf.seek(0)
        mp.upload_part_from_file(buf, num_part, size=bytes_staged)
        total_bytes_written += bytes_staged
        num_part += 1
      except IOError as e:
        logger.warn('Error when writing to stream in stream_write_internal at path %s: %s', path, e)
        write_error = e

        if self._metric_queue is not None:
          self._metric_queue.put_deprecated('MultipartUploadFailure', 1)
          self._metric_queue.multipart_upload_end.Inc(labelvalues=['failure'])

        if cancel_on_error:
          mp.cancel_upload()
          return 0, write_error
        else:
          break

    if total_bytes_written > 0:
      if self._metric_queue is not None:
        self._metric_queue.put_deprecated('MultipartUploadSuccess', 1)
        self._metric_queue.multipart_upload_end.Inc(labelvalues=['success'])

      mp.complete_upload()

    return total_bytes_written, write_error

  def exists(self, path):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    return key.exists()

  def remove(self, path):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    if key.exists():
      # It's a file
      key.delete()
      return
    # We assume it's a directory
    if not path.endswith('/'):
      path += '/'
    for key in self._cloud_bucket.list(prefix=path):
      key.delete()

  def get_checksum(self, path):
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)
    k = self._cloud_bucket.lookup(key)
    if k is None:
      raise IOError('No such key: \'{0}\''.format(path))

    return k.etag[1:-1][:7]

  def copy_to(self, destination, path):
    self._initialize_cloud_conn()

    # First try to copy directly via boto, but only if the storages are the
    # same type, with the same access information.
    if (self.__class__ == destination.__class__ and
        self._access_key and self._secret_key and
        self._access_key == destination._access_key and
        self._secret_key == destination._secret_key):
      logger.debug('Copying file from %s to %s via a direct boto copy', self._cloud_bucket,
                   destination._cloud_bucket)

      source_path = self._init_path(path)
      source_key = self._key_class(self._cloud_bucket, source_path)

      destination._initialize_cloud_conn()
      dest_path = destination._init_path(path)
      source_key.copy(destination._cloud_bucket, dest_path)
      return

    # Fallback to a slower, default copy.
    logger.debug('Copying file from %s to %s via a streamed copy', self._cloud_bucket,
                 destination)
    with self.stream_read_file(path) as fp:
      destination.stream_write(path, fp)

  def _rel_upload_path(self, uuid):
    return 'uploads/{0}'.format(uuid)

  def initiate_chunked_upload(self):
    self._initialize_cloud_conn()
    random_uuid = str(uuid4())

    metadata = {
      _CHUNKS_KEY: [],
    }

    return random_uuid, metadata

  def stream_upload_chunk(self, uuid, offset, length, in_fp, storage_metadata, content_type=None):
    self._initialize_cloud_conn()

    # We are going to upload each chunk to a separate key
    chunk_path = self._rel_upload_path(str(uuid4()))
    bytes_written, write_error = self._stream_write_internal(chunk_path, in_fp,
                                                             cancel_on_error=False, size=length,
                                                             content_type=content_type)

    new_metadata = copy.deepcopy(storage_metadata)

    # We are only going to track keys to which data was confirmed written
    if bytes_written > 0:
      new_metadata[_CHUNKS_KEY].append(_PartUploadMetadata(chunk_path, offset, bytes_written))

    return bytes_written, new_metadata, write_error

  def _chunk_generator(self, chunk_list):
    for chunk in chunk_list:
      yield filelike.StreamSlice(self.stream_read_file(chunk.path), 0, chunk.length)

  @staticmethod
  def _chunk_list_from_metadata(storage_metadata):
    return [_PartUploadMetadata(*chunk_args) for chunk_args in storage_metadata[_CHUNKS_KEY]]

  def _client_side_chunk_join(self, final_path, chunk_list):
    # If there's only one chunk, just "move" (copy and delete) the key and call it a day.
    if len(chunk_list) == 1:
      chunk_path = self._init_path(chunk_list[0].path)
      abs_final_path = self._init_path(final_path)

      # Let the copy raise an exception if it fails.
      self._cloud_bucket.copy_key(abs_final_path, self._bucket_name, chunk_path)

      # Attempt to clean up the old chunk.
      try:
        self._cloud_bucket.delete_key(chunk_path)
      except IOError:
        # We failed to delete a chunk. This sucks, but we shouldn't fail the push.
        msg = 'Failed to clean up chunk %s for move of %s'
        logger.exception(msg, chunk_path, abs_final_path)
    else:
      # Concatenate and write all the chunks as one key.
      concatenated = filelike.FilelikeStreamConcat(self._chunk_generator(chunk_list))
      self.stream_write(final_path, concatenated)

      # Attempt to clean up all the chunks.
      for chunk in chunk_list:
        try:
          self._cloud_bucket.delete_key(self._init_path(chunk.path))
        except IOError:
          # We failed to delete a chunk. This sucks, but we shouldn't fail the push.
          msg = 'Failed to clean up chunk %s for reupload of %s'
          logger.exception(msg, chunk.path, final_path)

  def complete_chunked_upload(self, uuid, final_path, storage_metadata, force_client_side=False):
    self._initialize_cloud_conn()
    chunk_list = self._chunk_list_from_metadata(storage_metadata)

    # Here is where things get interesting: we are going to try to assemble this server side
    # In order to be a candidate all parts (after offsets have been computed) must be at least 5MB
    server_side_assembly = False
    if not force_client_side:
      server_side_assembly = True
      for chunk_offset, chunk in enumerate(chunk_list):
        # If the chunk is both too small, and not the last chunk, we rule out server side assembly
        if chunk.length < self.automatic_chunk_size and (chunk_offset + 1) < len(chunk_list):
          server_side_assembly = False
          break

    if server_side_assembly:
      logger.debug('Performing server side assembly of multi-part upload for: %s', final_path)
      try:
        # Awesome, we can do this completely server side, now we have to start a new multipart
        # upload and use copy_part_from_key to set all of the chunks.
        mpu = self.__initiate_multipart_upload(final_path, content_type=None, content_encoding=None)

        for chunk_offset, chunk in enumerate(chunk_list):
          abs_chunk_path = self._init_path(chunk.path)
          part_num = chunk_offset + 1
          chunk_end_offset_inclusive = chunk.length - 1
          mpu.copy_part_from_key(self.get_cloud_bucket().name, abs_chunk_path, part_num,
                                 start=0, end=chunk_end_offset_inclusive)

        # Note: Sometimes Amazon S3 simply raises an internal error when trying to complete a
        # multipart upload. The recommendation is to simply try calling complete_upload again.
        for remaining_retries in range(3, -1, -1):
          try:
            mpu.complete_upload()
            break
          except S3ResponseError as s3re:
            if remaining_retries and s3re.status == 200 and s3re.error_code == 'InternalError':
              # Weird internal error case. Retry.
              continue

            # Otherwise, raise it.
            logger.exception('Exception trying to complete multipart upload for: %s', final_path)
            raise s3re


      except IOError as ioe:
        # Something bad happened, log it and then give up
        msg = 'Exception when attempting server-side assembly for: %s'
        logger.exception(msg, final_path)
        mpu.cancel_upload()
        raise ioe

    else:
      # We are going to turn all of the server side objects into a single file-like stream, and
      # pass that to stream_write to chunk and upload the final object.
      self._client_side_chunk_join(final_path, chunk_list)


  def cancel_chunked_upload(self, uuid, storage_metadata):
    self._initialize_cloud_conn()

    # We have to go through and delete all of the uploaded chunks
    for chunk in self._chunk_list_from_metadata(storage_metadata):
      self.remove(chunk.path)


class S3Storage(_CloudStorage):
  def __init__(self, metric_queue, storage_path, s3_bucket, s3_access_key=None, s3_secret_key=None,
               host=None):
    upload_params = {
      'encrypt_key': True,
    }
    connect_kwargs = {}
    if host:
      if host.startswith('http:') or host.startswith('https:'):
        raise ValueError('host name must not start with http:// or https://')

      connect_kwargs['host'] = host
    super(S3Storage, self).__init__(metric_queue, boto.s3.connection.S3Connection, boto.s3.key.Key,
                                    connect_kwargs, upload_params, storage_path, s3_bucket,
                                    access_key=s3_access_key or None,
                                    secret_key=s3_secret_key or None)

  def setup(self):
    self.get_cloud_bucket().set_cors_xml("""<?xml version="1.0" encoding="UTF-8"?>
      <CORSConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/">
          <CORSRule>
              <AllowedOrigin>*</AllowedOrigin>
              <AllowedMethod>GET</AllowedMethod>
              <MaxAgeSeconds>3000</MaxAgeSeconds>
              <AllowedHeader>Authorization</AllowedHeader>
          </CORSRule>
          <CORSRule>
              <AllowedOrigin>*</AllowedOrigin>
              <AllowedMethod>PUT</AllowedMethod>
              <MaxAgeSeconds>3000</MaxAgeSeconds>
              <AllowedHeader>Content-Type</AllowedHeader>
              <AllowedHeader>x-amz-acl</AllowedHeader>
              <AllowedHeader>origin</AllowedHeader>
          </CORSRule>
      </CORSConfiguration>""")

class GoogleCloudStorage(_CloudStorage):
  def __init__(self, metric_queue, storage_path, access_key, secret_key, bucket_name):
    upload_params = {}
    connect_kwargs = {}
    super(GoogleCloudStorage, self).__init__(metric_queue, boto.gs.connection.GSConnection,
                                             boto.gs.key.Key, connect_kwargs, upload_params,
                                             storage_path, bucket_name, access_key, secret_key)

  def setup(self):
    self.get_cloud_bucket().set_cors_xml("""<?xml version="1.0" encoding="UTF-8"?>
      <CorsConfig>
        <Cors>
          <Origins>
            <Origin>*</Origin>
          </Origins>
          <Methods>
            <Method>GET</Method>
            <Method>PUT</Method>
          </Methods>
          <ResponseHeaders>
            <ResponseHeader>Content-Type</ResponseHeader>
          </ResponseHeaders>
          <MaxAgeSec>3000</MaxAgeSec>
        </Cors>
      </CorsConfig>""")

  def _stream_write_internal(self, path, fp, content_type=None, content_encoding=None,
                             cancel_on_error=True, size=filelike.READ_UNTIL_END):
    # Minimum size of upload part size on S3 is 5MB
    self._initialize_cloud_conn()
    path = self._init_path(path)
    key = self._key_class(self._cloud_bucket, path)

    if content_type is not None:
      key.set_metadata('Content-Type', content_type)

    if content_encoding is not None:
      key.set_metadata('Content-Encoding', content_encoding)

    if size != filelike.READ_UNTIL_END:
      fp = filelike.StreamSlice(fp, 0, size)

    # TODO figure out how to handle cancel_on_error=False
    try:
      key.set_contents_from_stream(fp)
    except IOError as ex:
      return 0, ex

    return key.size, None

  def complete_chunked_upload(self, uuid, final_path, storage_metadata):
    self._initialize_cloud_conn()

    # Boto does not support GCS's multipart upload API because it differs from S3, so
    # we are forced to join it all locally and then reupload.
    # See https://github.com/boto/boto/issues/3355
    chunk_list = self._chunk_list_from_metadata(storage_metadata)
    self._client_side_chunk_join(final_path, chunk_list)


class RadosGWStorage(_CloudStorage):
  def __init__(self, metric_queue, hostname, is_secure, storage_path, access_key, secret_key,
               bucket_name):
    upload_params = {}
    connect_kwargs = {
      'host': hostname,
      'is_secure': is_secure,
      'calling_format': boto.s3.connection.OrdinaryCallingFormat(),
    }
    super(RadosGWStorage, self).__init__(metric_queue, boto.s3.connection.S3Connection,
                                         boto.s3.key.Key, connect_kwargs, upload_params,
                                         storage_path, bucket_name, access_key, secret_key)

  # TODO remove when radosgw supports cors: http://tracker.ceph.com/issues/8718#change-38624
  def get_direct_download_url(self, path, expires_in=60, requires_cors=False):
    if requires_cors:
      return None

    return super(RadosGWStorage, self).get_direct_download_url(path, expires_in, requires_cors)

  # TODO remove when radosgw supports cors: http://tracker.ceph.com/issues/8718#change-38624
  def get_direct_upload_url(self, path, mime_type, requires_cors=True):
    if requires_cors:
      return None

    return super(RadosGWStorage, self).get_direct_upload_url(path, mime_type, requires_cors)

  def complete_chunked_upload(self, uuid, final_path, storage_metadata):
    self._initialize_cloud_conn()

    # RadosGW does not support multipart copying from keys, so we are forced to join
    # it all locally and then reupload.
    # See https://github.com/ceph/ceph/pull/5139
    chunk_list = self._chunk_list_from_metadata(storage_metadata)
    self._client_side_chunk_join(final_path, chunk_list)