Example #1
0
 def __iter__(self):
     """ Standard iterator function that returns the object's contents. """
     try:
         while True:
             if not self.segment_iter:
                 self._load_next_segment()
             while True:
                 with ChunkReadTimeout(self.controller.app.node_timeout):
                     try:
                         chunk = self.segment_iter.next()
                         break
                     except StopIteration:
                         self._load_next_segment()
             self.position += len(chunk)
             yield chunk
     except StopIteration:
         raise
     except (Exception, Timeout), err:
         if not getattr(err, 'swift_logged', False):
             self.controller.app.logger.exception(_(
                 'ERROR: While processing manifest '
                 '/%(acc)s/%(cont)s/%(obj)s'),
                 {'acc': self.controller.account_name,
                  'cont': self.controller.container_name,
                  'obj': self.controller.object_name})
             err.swift_logged = True
             self.response.status_int = HTTP_SERVICE_UNAVAILABLE
         raise
Example #2
0
    def _make_app_iter(self, node, source):
        """
        Returns an iterator over the contents of the source (via its read
        func).  There is also quite a bit of cleanup to ensure garbage
        collection works and the underlying socket of the source is closed.

        :param source: The httplib.Response object this iterator should read
                       from.
        :param node: The node the source is reading from, for logging purposes.
        """
        try:
            # Spawn reader to read from the source and place in the queue.
            # We then drop any reference to the source or node, for garbage
            # collection purposes.
            queue = Queue(1)
            spawn_n(self._make_app_iter_reader, node, source, queue,
                    self.app.logger.thread_locals)
            source = node = None
            while True:
                chunk = queue.get(timeout=self.app.node_timeout)
                if isinstance(chunk, bool):  # terminator
                    success = chunk
                    if not success:
                        raise Exception(
                            _('Failed to read all data'
                              ' from the source'))
                    break
                yield chunk
        except Empty:
            raise ChunkReadTimeout()
        except (GeneratorExit, Timeout):
            self.app.logger.warn(_('Client disconnected on read'))
        except Exception:
            self.app.logger.exception(_('Trying to send to client'))
            raise
Example #3
0
    def _read_metadata_footer(self, mime_documents_iter):
        try:
            with ChunkReadTimeout(self.client_timeout):
                footer_hdrs, footer_iter = next(mime_documents_iter)
        except ChunkReadTimeout:
            raise HTTPClientDisconnect()
        except StopIteration:
            raise HTTPBadRequest(body="couldn't find footer MIME doc")

        timeout_reader = self._make_timeout_reader(footer_iter)
        try:
            footer_body = ''.join(iter(timeout_reader, ''))
        except ChunkReadTimeout:
            raise HTTPClientDisconnect()

        footer_md5 = footer_hdrs.get('Content-MD5')
        if not footer_md5:
            raise HTTPBadRequest(body="no Content-MD5 in footer")
        if footer_md5 != md5(footer_body).hexdigest():
            raise HTTPUnprocessableEntity(body="footer MD5 mismatch")

        try:
            return HeaderKeyDict(json.loads(footer_body))
        except ValueError:
            raise HTTPBadRequest("invalid JSON for footer doc")
Example #4
0
    def _make_app_iter(self, node, source):
        """
        Returns an iterator over the contents of the source (via its read
        func).  There is also quite a bit of cleanup to ensure garbage
        collection works and the underlying socket of the source is closed.

        :param source: The httplib.Response object this iterator should read
                       from.
        :param node: The node the source is reading from, for logging purposes.
        """
        try:
            nchunks = 0
            while True:
                with ChunkReadTimeout(self.app.node_timeout):
                    chunk = source.read(self.app.object_chunk_size)
                    nchunks += 1
                if not chunk:
                    break
                with ChunkWriteTimeout(self.app.client_timeout):
                    yield chunk
                # This is for fairness; if the network is outpacing the CPU,
                # we'll always be able to read and write data without
                # encountering an EWOULDBLOCK, and so eventlet will not switch
                # greenthreads on its own. We do it manually so that clients
                # don't starve.
                #
                # The number 5 here was chosen by making stuff up. It's not
                # every single chunk, but it's not too big either, so it seemed
                # like it would probably be an okay choice.
                #
                # Note that we may trampoline to other greenthreads more often
                # than once every 5 chunks, depending on how blocking our
                # network IO is; the explicit sleep here simply provides a
                # lower bound on the rate of trampolining.
                if nchunks % 5 == 0:
                    sleep()
        except ChunkReadTimeout:
            self.exception_occurred(node, _('Object'),
                                    _('Trying to read during GET'))
            raise
        except ChunkWriteTimeout:
            self.app.logger.warn(
                _('Client did not read from proxy within %ss') %
                self.app.client_timeout)
            self.app.logger.increment('client_timeouts')
        except GeneratorExit:
            self.app.logger.warn(_('Client disconnected on read'))
        except Exception:
            self.app.logger.exception(_('Trying to send to client'))
            raise
        finally:
            # Close-out the connection as best as possible.
            if getattr(source, 'swift_conn', None):
                self.close_swift_conn(source)
Example #5
0
 def _read_put_commit_message(self, mime_documents_iter):
     rcvd_commit = False
     try:
         with ChunkReadTimeout(self.client_timeout):
             commit_hdrs, commit_iter = next(mime_documents_iter)
             if commit_hdrs.get('X-Document', None) == "put commit":
                 rcvd_commit = True
         drain(commit_iter, self.network_chunk_size, self.client_timeout)
     except ChunkReadTimeout:
         raise HTTPClientDisconnect()
     except StopIteration:
         raise HTTPBadRequest(body="couldn't find PUT commit MIME doc")
     return rcvd_commit
Example #6
0
 def response_iter():
     try:
         while True:
             with ChunkReadTimeout(self.client_timeout):
                 chunk = result.read(self.client_chunk_size)
             if not chunk:
                 break
             yield chunk
             response.bytes_transferred += len(chunk)
     except GeneratorExit:
         pass
     except (Exception, TimeoutError):
         raise
Example #7
0
def drain(file_like, read_size, timeout):
    """
    Read and discard any bytes from file_like.

    :param file_like: file-like object to read from
    :param read_size: how big a chunk to read at a time
    :param timeout: how long to wait for a read (use None for no timeout)

    :raises ChunkReadTimeout: if no chunk was read in time
    """

    while True:
        with ChunkReadTimeout(timeout):
            chunk = file_like.read(read_size)
            if not chunk:
                break
Example #8
0
    def _make_app_iter_reader(self, node, source, queue, logger_thread_locals):
        """
        Reads from the source and places data in the queue. It expects
        something else be reading from the queue and, if nothing does within
        self.app.client_timeout seconds, the process will be aborted.

        :param node: The node dict that the source is connected to, for
                     logging/error-limiting purposes.
        :param source: The httplib.Response object to read from.
        :param queue: The eventlet.queue.Queue to place read source data into.
        :param logger_thread_locals: The thread local values to be set on the
                                     self.app.logger to retain transaction
                                     logging information.
        """
        self.app.logger.thread_locals = logger_thread_locals
        success = True
        try:
            try:
                while True:
                    with ChunkReadTimeout(self.app.node_timeout):
                        chunk = source.read(self.app.object_chunk_size)
                    if not chunk:
                        break
                    queue.put(chunk, timeout=self.app.client_timeout)
            except Full:
                self.app.logger.warn(
                    _('Client did not read from queue within %ss') %
                    self.app.client_timeout)
                self.app.logger.increment('client_timeouts')
                success = False
            except (Exception, Timeout):
                self.exception_occurred(node, _('Object'),
                                        _('Trying to read during GET'))
                success = False
        finally:
            # Ensure the queue getter gets a terminator.
            queue.resize(2)
            queue.put(success)
            # Close-out the connection as best as possible.
            if getattr(source, 'swift_conn', None):
                self.close_swift_conn(source)
Example #9
0
    def PUT(self, request):
        """Handle HTTP PUT requests for the Swift Object Server."""
        device, partition, account, container, obj, policy = \
            get_name_and_placement(request, 5, 5, True)
        req_timestamp = valid_timestamp(request)
        error_response = check_object_creation(request, obj)
        if error_response:
            return error_response
        new_delete_at = int(request.headers.get('X-Delete-At') or 0)
        if new_delete_at and new_delete_at < time.time():
            return HTTPBadRequest(body='X-Delete-At in past', request=request,
                                  content_type='text/plain')
        try:
            fsize = request.message_length()
        except ValueError as e:
            return HTTPBadRequest(body=str(e), request=request,
                                  content_type='text/plain')

        # In case of multipart-MIME put, the proxy sends a chunked request,
        # but may let us know the real content length so we can verify that
        # we have enough disk space to hold the object.
        if fsize is None:
            fsize = request.headers.get('X-Backend-Obj-Content-Length')
            if fsize is not None:
                try:
                    fsize = int(fsize)
                except ValueError as e:
                    return HTTPBadRequest(body=str(e), request=request,
                                          content_type='text/plain')
        # SSYNC will include Frag-Index header for subrequests to primary
        # nodes; handoff nodes should 409 subrequests to over-write an
        # existing data fragment until they offloaded the existing fragment
        frag_index = request.headers.get('X-Backend-Ssync-Frag-Index')
        try:
            disk_file = self.get_diskfile(
                device, partition, account, container, obj,
                policy=policy, frag_index=frag_index)
        except DiskFileDeviceUnavailable:
            return HTTPInsufficientStorage(drive=device, request=request)
        try:
            orig_metadata = disk_file.read_metadata()
        except DiskFileXattrNotSupported:
            return HTTPInsufficientStorage(drive=device, request=request)
        except (DiskFileNotExist, DiskFileQuarantined):
            orig_metadata = {}

        # Checks for If-None-Match
        if request.if_none_match is not None and orig_metadata:
            if '*' in request.if_none_match:
                # File exists already so return 412
                return HTTPPreconditionFailed(request=request)
            if orig_metadata.get('ETag') in request.if_none_match:
                # The current ETag matches, so return 412
                return HTTPPreconditionFailed(request=request)

        orig_timestamp = Timestamp(orig_metadata.get('X-Timestamp', 0))
        if orig_timestamp >= req_timestamp:
            return HTTPConflict(
                request=request,
                headers={'X-Backend-Timestamp': orig_timestamp.internal})
        orig_delete_at = int(orig_metadata.get('X-Delete-At') or 0)
        upload_expiration = time.time() + self.max_upload_time
        etag = md5()
        elapsed_time = 0
        try:
            with disk_file.create(size=fsize) as writer:
                upload_size = 0

                # If the proxy wants to send us object metadata after the
                # object body, it sets some headers. We have to tell the
                # proxy, in the 100 Continue response, that we're able to
                # parse a multipart MIME document and extract the object and
                # metadata from it. If we don't, then the proxy won't
                # actually send the footer metadata.
                have_metadata_footer = False
                use_multiphase_commit = False
                mime_documents_iter = iter([])
                obj_input = request.environ['wsgi.input']

                hundred_continue_headers = []
                if config_true_value(
                        request.headers.get(
                            'X-Backend-Obj-Multiphase-Commit')):
                    use_multiphase_commit = True
                    hundred_continue_headers.append(
                        ('X-Obj-Multiphase-Commit', 'yes'))

                if config_true_value(
                        request.headers.get('X-Backend-Obj-Metadata-Footer')):
                    have_metadata_footer = True
                    hundred_continue_headers.append(
                        ('X-Obj-Metadata-Footer', 'yes'))

                if have_metadata_footer or use_multiphase_commit:
                    obj_input.set_hundred_continue_response_headers(
                        hundred_continue_headers)
                    mime_boundary = request.headers.get(
                        'X-Backend-Obj-Multipart-Mime-Boundary')
                    if not mime_boundary:
                        return HTTPBadRequest("no MIME boundary")

                    try:
                        with ChunkReadTimeout(self.client_timeout):
                            mime_documents_iter = iter_mime_headers_and_bodies(
                                request.environ['wsgi.input'],
                                mime_boundary, self.network_chunk_size)
                            _junk_hdrs, obj_input = next(mime_documents_iter)
                    except ChunkReadTimeout:
                        return HTTPRequestTimeout(request=request)

                timeout_reader = self._make_timeout_reader(obj_input)
                try:
                    for chunk in iter(timeout_reader, ''):
                        start_time = time.time()
                        if start_time > upload_expiration:
                            self.logger.increment('PUT.timeouts')
                            return HTTPRequestTimeout(request=request)
                        etag.update(chunk)
                        upload_size = writer.write(chunk)
                        elapsed_time += time.time() - start_time
                except ChunkReadTimeout:
                    return HTTPRequestTimeout(request=request)
                if upload_size:
                    self.logger.transfer_rate(
                        'PUT.' + device + '.timing', elapsed_time,
                        upload_size)
                if fsize is not None and fsize != upload_size:
                    return HTTPClientDisconnect(request=request)

                footer_meta = {}
                if have_metadata_footer:
                    footer_meta = self._read_metadata_footer(
                        mime_documents_iter)

                request_etag = (footer_meta.get('etag') or
                                request.headers.get('etag', '')).lower()
                etag = etag.hexdigest()
                if request_etag and request_etag != etag:
                    return HTTPUnprocessableEntity(request=request)
                metadata = {
                    'X-Timestamp': request.timestamp.internal,
                    'Content-Type': request.headers['content-type'],
                    'ETag': etag,
                    'Content-Length': str(upload_size),
                }
                metadata.update(val for val in request.headers.items()
                                if is_sys_or_user_meta('object', val[0]))
                metadata.update(val for val in footer_meta.items()
                                if is_sys_or_user_meta('object', val[0]))
                headers_to_copy = (
                    request.headers.get(
                        'X-Backend-Replication-Headers', '').split() +
                    list(self.allowed_headers))
                for header_key in headers_to_copy:
                    if header_key in request.headers:
                        header_caps = header_key.title()
                        metadata[header_caps] = request.headers[header_key]
                writer.put(metadata)

                # if the PUT requires a two-phase commit (a data and a commit
                # phase) send the proxy server another 100-continue response
                # to indicate that we are finished writing object data
                if use_multiphase_commit:
                    request.environ['wsgi.input'].\
                        send_hundred_continue_response()
                    if not self._read_put_commit_message(mime_documents_iter):
                        return HTTPServerError(request=request)
                    # got 2nd phase confirmation, write a timestamp.durable
                    # state file to indicate a successful PUT

                writer.commit(request.timestamp)

                # Drain any remaining MIME docs from the socket. There
                # shouldn't be any, but we must read the whole request body.
                try:
                    while True:
                        with ChunkReadTimeout(self.client_timeout):
                            _junk_hdrs, _junk_body = next(mime_documents_iter)
                        drain(_junk_body, self.network_chunk_size,
                              self.client_timeout)
                except ChunkReadTimeout:
                    raise HTTPClientDisconnect()
                except StopIteration:
                    pass

        except (DiskFileXattrNotSupported, DiskFileNoSpace):
            return HTTPInsufficientStorage(drive=device, request=request)
        if orig_delete_at != new_delete_at:
            if new_delete_at:
                self.delete_at_update(
                    'PUT', new_delete_at, account, container, obj, request,
                    device, policy)
            if orig_delete_at:
                self.delete_at_update(
                    'DELETE', orig_delete_at, account, container, obj,
                    request, device, policy)
        update_headers = HeaderKeyDict({
            'x-size': metadata['Content-Length'],
            'x-content-type': metadata['Content-Type'],
            'x-timestamp': metadata['X-Timestamp'],
            'x-etag': metadata['ETag']})
        # apply any container update header overrides sent with request
        self._check_container_override(update_headers, request.headers)
        self._check_container_override(update_headers, footer_meta)
        self.container_update(
            'PUT', account, container, obj, request,
            update_headers,
            device, policy)
        return HTTPCreated(request=request, etag=etag)
Example #10
0
    def __call__(self):
        """
        :return httplib.HTTP(S)Connection in success, and webob.exc.HTTPException in failure
        """
        if self.headers.has_key('content-length'):
            if int(self.headers['content-length']) >= MAX_FILE_SIZE:
                return HTTPRequestEntityTooLarge(request=self.req)

        parsed = urlparse(self.url)
        if self.proxy:
            proxy_parsed = urlparse(self.proxy)

        if self._proxy_request_check(parsed.path):
            host, port = self.split_netloc(proxy_parsed)
            path = self.url
        else:
            host, port = self.split_netloc(parsed)
            path = parsed.path
        self.headers['host'] = '%s:%s' % (host, port)

        if self.method == 'PUT' and len(parsed.path.split('/')) == 5:
            chunked = self.req.headers.get('transfer-encoding')
            reader = self.req.environ['wsgi.input'].read
            data_source = iter(lambda: reader(self.chunk_size), '')
            bytes_transferred = 0
            # pile = GreenPile()
            # pile.spawn(self._connect_server, host, port, self.method, path, self.headers, parsed.query)
            # conns = [conn for conn in pile if conn]
            # conn = conns[0]
            try:
                with ConnectionTimeout(self.conn_timeout):
                    conn = http_connect_raw(host,
                                            port,
                                            self.method,
                                            path,
                                            headers=self.headers,
                                            query_string=parsed.query)
                with ContextPool(1) as pool:
                    conn.failed = False
                    conn.queue = Queue(10)
                    pool.spawn(self._send_file, conn, path)
                    while True:
                        with ChunkReadTimeout(self.client_timeout):
                            try:
                                chunk = next(data_source)
                            except StopIteration:
                                if chunked:
                                    conn.queue.put('0\r\n\r\n')
                                break
                            except TypeError, err:
                                self.logger.info('Chunk Read Error: %s' % err)
                                break
                            except Exception, err:
                                self.logger.info('Chunk Read Error: %s' % err)
                                return HTTPServerError(request=self.req)
                        bytes_transferred += len(chunk)
                        if bytes_transferred > MAX_FILE_SIZE:
                            return HTTPRequestEntityTooLarge(request=self.req)
                        if not conn.failed:
                            conn.queue.put('%x\r\n%s\r\n' %
                                           (len(chunk),
                                            chunk) if chunked else chunk)
Example #11
0
    def PUT(self, req):
        """HTTP PUT request handler."""
        if req.if_none_match is not None and '*' not in req.if_none_match:
            # Sending an etag with if-none-match isn't currently supported
            return HTTPBadRequest(request=req,
                                  content_type='text/plain',
                                  body='If-None-Match only supports *')
        container_info = self.container_info(self.account_name,
                                             self.container_name, req)
        policy_index = req.headers.get('X-Backend-Storage-Policy-Index',
                                       container_info['storage_policy'])
        obj_ring = self.app.get_object_ring(policy_index)

        # pass the policy index to storage nodes via req header
        req.headers['X-Backend-Storage-Policy-Index'] = policy_index
        container_partition = container_info['partition']
        containers = container_info['nodes']
        req.acl = container_info['write_acl']
        req.environ['swift_sync_key'] = container_info['sync_key']
        object_versions = container_info['versions']
        if 'swift.authorize' in req.environ:
            aresp = req.environ['swift.authorize'](req)
            if aresp:
                return aresp

        if not containers:
            return HTTPNotFound(request=req)

        # Sometimes the 'content-type' header exists, but is set to None.
        content_type_manually_set = True
        detect_content_type = \
            config_true_value(req.headers.get('x-detect-content-type'))
        if detect_content_type or not req.headers.get('content-type'):
            guessed_type, _junk = mimetypes.guess_type(req.path_info)
            req.headers['Content-Type'] = guessed_type or \
                'application/octet-stream'
            if detect_content_type:
                req.headers.pop('x-detect-content-type')
            else:
                content_type_manually_set = False

        error_response = check_object_creation(req, self.object_name) or \
            check_content_type(req)
        if error_response:
            return error_response

        partition, nodes = obj_ring.get_nodes(self.account_name,
                                              self.container_name,
                                              self.object_name)
        ####################################  CHANGED_CODE  ############################################################
        # Change the nodes list to contain only one dictionary item instead of the original 3 returned by the ring.
        d = dict()
        # d[partition] = nodes[1:]
        # f.write(str(d)+"\n")
        # f.close()
        print("===Original Nodes===")
        print(nodes)
        temp_nodes = []
        flag = 0
        f = open("/home/hduser/swift/swift/proxy/controllers/spindowndevices",
                 "r")
        sdlist = f.read().split("\n")
        print("===Spun down devices===:", sdlist)
        f.close()

        upnodes = [item for item in nodes if item['device'] not in sdlist]
        downnodes = [item for item in nodes if item['device'] in sdlist]
        temp_nodes = upnodes
        if (len(downnodes) > 0):
            d = ast.literal_eval(
                open("/home/hduser/swift/swift/proxy/controllers/nodes.txt",
                     "r").read())
            # d_temp=pickle.load("/home/hduser/swift/proxy/controllers/nodes.p","rb")
            # print("===Current dict===:",d)
            for item in downnodes:
                if (partition in d):
                    d[partition].append(item)
                    # print("===Modified dict===:",d)
                else:
                    d[partition] = [item]
                    # print("===Modified dict===:",d)
        # pickle.dump(d,open("/home/hduser/nodes.p","wb"))
        # print("Before writing:",d)
        fo = open("/home/hduser/swift/swift/proxy/controllers/nodes.txt", "w")
        fo.write(str(d) + "\n")
        fo.close()
        # pickle.dump(d,open("/home/hduser/swift/swift/proxy/controllers/nodes.p","wb"))
        ## Old method, IGNORE
        # for item in nodes:
        #     device = item['device']
        #     if(device not in sdlist):
        #     # if(os.path.ismount("path"))
        #         temp_nodes.append(item)
        #         flag = 1
        #         break
        #     else:
        #         pickle.dump(d,open("/home/hduser/nodes.p","wb"))
        #         # d = pickle.load(open("/home/hduser/nodes.p","rb"))
        #         import ast
        #         d = ast.literal_eval(open("/home/hduser/nodes.txt","r").read())
        #         print("===Current dict===:",d)
        #         if(partition in d):
        #             print("In IF")
        #             d[partition].append(item)
        #             print("===Modified dict===:",d)
        #         else:
        #             print("In ELSE")
        #             d[partition] = [item]
        #             print("===Modified dict===:",d)
        #         pickle.dump(d,open("/home/hduser/nodes.p","wb"))
        #         fo = open("/home/hduser/nodes.txt","w")
        #         fo.write(str(d)+"\n")

        # Code to spin up a device if none are running already.
        if (len(upnodes) == 0):
            dev = nodes[0]['device']
            print("===ALL NODES DOWN===")
            print("===Mounting device===", dev)
            os.system("mount /dev/" + str(dev))

        print('===In controller PUT===:')
        print("===Partition===", partition)
        nodes = temp_nodes

        print('===In controller PUT===:')
        print("===Partition===", partition)
        nodes = temp_nodes
        print("===Nodes===:", nodes)

        check_ssd()
        ############################################  CHANGED_CODE  ########################################################

        # do a HEAD request for checking object versions
        if object_versions and not req.environ.get('swift_versioned_copy'):
            # make sure proxy-server uses the right policy index
            _headers = {
                'X-Backend-Storage-Policy-Index': policy_index,
                'X-Newest': 'True'
            }
            hreq = Request.blank(req.path_info,
                                 headers=_headers,
                                 environ={'REQUEST_METHOD': 'HEAD'})
            hresp = self.GETorHEAD_base(hreq, _('Object'), obj_ring, partition,
                                        hreq.swift_entity_path)

        # Used by container sync feature
        if 'x-timestamp' in req.headers:
            try:
                req_timestamp = Timestamp(req.headers['X-Timestamp'])
            except ValueError:
                return HTTPBadRequest(
                    request=req,
                    content_type='text/plain',
                    body='X-Timestamp should be a UNIX timestamp float value; '
                    'was %r' % req.headers['x-timestamp'])
            req.headers['X-Timestamp'] = req_timestamp.internal
        else:
            req.headers['X-Timestamp'] = Timestamp(time.time()).internal

        if object_versions and not req.environ.get('swift_versioned_copy'):
            is_manifest = 'X-Object-Manifest' in req.headers or \
                          'X-Object-Manifest' in hresp.headers
            if hresp.status_int != HTTP_NOT_FOUND and not is_manifest:
                # This is a version manifest and needs to be handled
                # differently. First copy the existing data to a new object,
                # then write the data from this request to the version manifest
                # object.
                lcontainer = object_versions.split('/')[0]
                prefix_len = '%03x' % len(self.object_name)
                lprefix = prefix_len + self.object_name + '/'
                ts_source = hresp.environ.get('swift_x_timestamp')
                if ts_source is None:
                    ts_source = time.mktime(
                        time.strptime(hresp.headers['last-modified'],
                                      '%a, %d %b %Y %H:%M:%S GMT'))
                new_ts = Timestamp(ts_source).internal
                vers_obj_name = lprefix + new_ts
                copy_headers = {
                    'Destination': '%s/%s' % (lcontainer, vers_obj_name)
                }
                copy_environ = {
                    'REQUEST_METHOD': 'COPY',
                    'swift_versioned_copy': True
                }
                copy_req = Request.blank(req.path_info,
                                         headers=copy_headers,
                                         environ=copy_environ)
                copy_resp = self.COPY(copy_req)
                if is_client_error(copy_resp.status_int):
                    # missing container or bad permissions
                    return HTTPPreconditionFailed(request=req)
                elif not is_success(copy_resp.status_int):
                    # could not copy the data, bail
                    return HTTPServiceUnavailable(request=req)

        reader = req.environ['wsgi.input'].read
        data_source = iter(lambda: reader(self.app.client_chunk_size), '')
        source_header = req.headers.get('X-Copy-From')
        source_resp = None
        if source_header:
            if req.environ.get('swift.orig_req_method', req.method) != 'POST':
                req.environ.setdefault('swift.log_info', []).append(
                    'x-copy-from:%s' % source_header)
            ver, acct, _rest = req.split_path(2, 3, True)
            src_account_name = req.headers.get('X-Copy-From-Account', None)
            if src_account_name:
                src_account_name = check_account_format(req, src_account_name)
            else:
                src_account_name = acct
            src_container_name, src_obj_name = check_copy_from_header(req)
            source_header = '/%s/%s/%s/%s' % (ver, src_account_name,
                                              src_container_name, src_obj_name)
            source_req = req.copy_get()

            # make sure the source request uses it's container_info
            source_req.headers.pop('X-Backend-Storage-Policy-Index', None)
            source_req.path_info = source_header
            source_req.headers['X-Newest'] = 'true'
            orig_obj_name = self.object_name
            orig_container_name = self.container_name
            orig_account_name = self.account_name
            self.object_name = src_obj_name
            self.container_name = src_container_name
            self.account_name = src_account_name
            sink_req = Request.blank(req.path_info,
                                     environ=req.environ,
                                     headers=req.headers)
            source_resp = self.GET(source_req)

            # This gives middlewares a way to change the source; for example,
            # this lets you COPY a SLO manifest and have the new object be the
            # concatenation of the segments (like what a GET request gives
            # the client), not a copy of the manifest file.
            hook = req.environ.get(
                'swift.copy_hook',
                (lambda source_req, source_resp, sink_req: source_resp))
            source_resp = hook(source_req, source_resp, sink_req)

            if source_resp.status_int >= HTTP_MULTIPLE_CHOICES:
                return source_resp
            self.object_name = orig_obj_name
            self.container_name = orig_container_name
            self.account_name = orig_account_name
            data_source = iter(source_resp.app_iter)
            sink_req.content_length = source_resp.content_length
            if sink_req.content_length is None:
                # This indicates a transfer-encoding: chunked source object,
                # which currently only happens because there are more than
                # CONTAINER_LISTING_LIMIT segments in a segmented object. In
                # this case, we're going to refuse to do the server-side copy.
                return HTTPRequestEntityTooLarge(request=req)
            if sink_req.content_length > constraints.MAX_FILE_SIZE:
                return HTTPRequestEntityTooLarge(request=req)
            sink_req.etag = source_resp.etag

            # we no longer need the X-Copy-From header
            del sink_req.headers['X-Copy-From']
            if 'X-Copy-From-Account' in sink_req.headers:
                del sink_req.headers['X-Copy-From-Account']
            if not content_type_manually_set:
                sink_req.headers['Content-Type'] = \
                    source_resp.headers['Content-Type']
            if config_true_value(
                    sink_req.headers.get('x-fresh-metadata', 'false')):
                # post-as-copy: ignore new sysmeta, copy existing sysmeta
                condition = lambda k: is_sys_meta('object', k)
                remove_items(sink_req.headers, condition)
                copy_header_subset(source_resp, sink_req, condition)
            else:
                # copy/update existing sysmeta and user meta
                copy_headers_into(source_resp, sink_req)
                copy_headers_into(req, sink_req)

            # copy over x-static-large-object for POSTs and manifest copies
            if 'X-Static-Large-Object' in source_resp.headers and \
                    req.params.get('multipart-manifest') == 'get':
                sink_req.headers['X-Static-Large-Object'] = \
                    source_resp.headers['X-Static-Large-Object']

            req = sink_req

        req, delete_at_container, delete_at_part, \
            delete_at_nodes = self._config_obj_expiration(req)

        node_iter = GreenthreadSafeIterator(
            self.iter_nodes_local_first(obj_ring, partition))
        pile = GreenPile(len(nodes))
        te = req.headers.get('transfer-encoding', '')
        chunked = ('chunked' in te)

        outgoing_headers = self._backend_requests(
            req, len(nodes), container_partition, containers,
            delete_at_container, delete_at_part, delete_at_nodes)

        for nheaders in outgoing_headers:
            # RFC2616:8.2.3 disallows 100-continue without a body
            if (req.content_length > 0) or chunked:
                nheaders['Expect'] = '100-continue'

#################################  CHANGED_CODE  ###################################################################
# Replaced node_iter by nodes in the following line to make sure that a new list with different order isnt used.
# Change from node_iter to nodes to make sure it writes to the same device.
# Without this, it gets a new list of nodes from the ring in a different order and connects to the first one.

            pile.spawn(self._connect_put_node, nodes, partition,
                       req.swift_entity_path, nheaders,
                       self.app.logger.thread_locals)

#################################  CHANGED_CODE ###################################################################

        conns = [conn for conn in pile if conn]
        min_conns = quorum_size(len(nodes))

        if req.if_none_match is not None and '*' in req.if_none_match:
            statuses = [conn.resp.status for conn in conns if conn.resp]
            if HTTP_PRECONDITION_FAILED in statuses:
                # If we find any copy of the file, it shouldn't be uploaded
                self.app.logger.debug(
                    _('Object PUT returning 412, %(statuses)r'),
                    {'statuses': statuses})
                return HTTPPreconditionFailed(request=req)

        if any(conn for conn in conns
               if conn.resp and conn.resp.status == HTTP_CONFLICT):
            timestamps = [
                HeaderKeyDict(
                    conn.resp.getheaders()).get('X-Backend-Timestamp')
                for conn in conns if conn.resp
            ]
            self.app.logger.debug(
                _('Object PUT returning 202 for 409: '
                  '%(req_timestamp)s <= %(timestamps)r'), {
                      'req_timestamp': req.timestamp.internal,
                      'timestamps': ', '.join(timestamps)
                  })
            return HTTPAccepted(request=req)

        if len(conns) < min_conns:
            self.app.logger.error(
                _('Object PUT returning 503, %(conns)s/%(nodes)s '
                  'required connections'), {
                      'conns': len(conns),
                      'nodes': min_conns
                  })
            return HTTPServiceUnavailable(request=req)
        bytes_transferred = 0
        try:
            with ContextPool(len(nodes)) as pool:
                for conn in conns:
                    conn.failed = False
                    conn.queue = Queue(self.app.put_queue_depth)
                    pool.spawn(self._send_file, conn, req.path)
                while True:
                    with ChunkReadTimeout(self.app.client_timeout):
                        try:
                            chunk = next(data_source)
                        except StopIteration:
                            if chunked:
                                for conn in conns:
                                    conn.queue.put('0\r\n\r\n')
                            break
                    bytes_transferred += len(chunk)
                    if bytes_transferred > constraints.MAX_FILE_SIZE:
                        return HTTPRequestEntityTooLarge(request=req)
                    for conn in list(conns):
                        if not conn.failed:
                            conn.queue.put('%x\r\n%s\r\n' %
                                           (len(chunk),
                                            chunk) if chunked else chunk)
                        else:
                            conns.remove(conn)
                    if len(conns) < min_conns:
                        self.app.logger.error(
                            _('Object PUT exceptions during'
                              ' send, %(conns)s/%(nodes)s required connections'
                              ), {
                                  'conns': len(conns),
                                  'nodes': min_conns
                              })
                        return HTTPServiceUnavailable(request=req)
                for conn in conns:
                    if conn.queue.unfinished_tasks:
                        conn.queue.join()
            conns = [conn for conn in conns if not conn.failed]
        except ChunkReadTimeout as err:
            self.app.logger.warn(_('ERROR Client read timeout (%ss)'),
                                 err.seconds)
            self.app.logger.increment('client_timeouts')
            return HTTPRequestTimeout(request=req)
        except (Exception, Timeout):
            self.app.logger.exception(
                _('ERROR Exception causing client disconnect'))
            return HTTPClientDisconnect(request=req)
        if req.content_length and bytes_transferred < req.content_length:
            req.client_disconnect = True
            self.app.logger.warn(
                _('Client disconnected without sending enough data'))
            self.app.logger.increment('client_disconnects')
            return HTTPClientDisconnect(request=req)

        statuses, reasons, bodies, etags = self._get_put_responses(
            req, conns, nodes)

        if len(etags) > 1:
            self.app.logger.error(
                _('Object servers returned %s mismatched etags'), len(etags))
            return HTTPServerError(request=req)
        etag = etags.pop() if len(etags) else None
        resp = self.best_response(req,
                                  statuses,
                                  reasons,
                                  bodies,
                                  _('Object PUT'),
                                  etag=etag)
        if source_header:
            acct, path = source_header.split('/', 3)[2:4]
            resp.headers['X-Copied-From-Account'] = quote(acct)
            resp.headers['X-Copied-From'] = quote(path)
            if 'last-modified' in source_resp.headers:
                resp.headers['X-Copied-From-Last-Modified'] = \
                    source_resp.headers['last-modified']
            copy_headers_into(req, resp)
        resp.last_modified = math.ceil(
            float(Timestamp(req.headers['X-Timestamp'])))
        return resp
Example #12
0
    def __call__(self):
        """
        :return httplib.HTTP(S)Connection in success, and webob.exc.HTTPException in failure
        """
        if self.headers.has_key('content-length'):
            if int(self.headers['content-length']) >= MAX_FILE_SIZE:
                return HTTPRequestEntityTooLarge(request=self.req)

        parsed = urlparse(self.url)
        if self.proxy:
            proxy_parsed = urlparse(self.proxy)

        if self._proxy_request_check(parsed.path):
            host, port = self.split_netloc(proxy_parsed)
            path = self.url
            ssl = True if proxy_parsed.scheme == 'https' else False
        else:
            host, port = self.split_netloc(parsed)
            path = parsed.path
            ssl = True if parsed.scheme == 'https' else False
        self.headers['host'] = '%s:%s' % (host, port)

        if self.method == 'PUT' and len(parsed.path.split('/')) >= 5:
            if self.headers.has_key('content-length') and int(
                    self.headers['content-length']) != 0:
                if not self.headers.has_key('expect'):
                    self.headers['expect'] = '100-continue'
            chunked = self.req.headers.get('transfer-encoding')
            if isinstance(self.req.environ['wsgi.input'], str):
                reader = self.req.environ['wsgi.input'].read
                data_source = iter(lambda: reader(self.chunk_size), '')
            else:
                data_source = self.req.environ['wsgi.input']
            bytes_transferred = 0
            try:
                conn = self._connect_put_node(host,
                                              port,
                                              self.method,
                                              path,
                                              headers=self.headers,
                                              query_string=parsed.query,
                                              ssl=ssl)
                if not conn:
                    return HTTPServiceUnavailable(request=self.req)
                with ContextPool(1) as pool:
                    conn.failed = False
                    conn.queue = Queue(10)
                    pool.spawn(self._send_file, conn, path)
                    while True:
                        with ChunkReadTimeout(self.client_timeout):
                            try:
                                chunk = next(data_source)
                            except StopIteration:
                                if chunked:
                                    conn.queue.put('0\r\n\r\n')
                                break
                            except TypeError, err:
                                self.logger.info('Chunk Read Error: %s' % err)
                                break
                            except Exception, err:
                                self.logger.info('Chunk Read Error: %s' % err)
                                return HTTPServerError(request=self.req)
                        bytes_transferred += len(chunk)
                        if bytes_transferred > MAX_FILE_SIZE:
                            return HTTPRequestEntityTooLarge(request=self.req)
                        if not conn.failed:
                            conn.queue.put('%x\r\n%s\r\n' %
                                           (len(chunk),
                                            chunk) if chunked else chunk)
Example #13
0
    def PUT(self, req):
        """HTTP PUT request handler."""
        container_info = self.container_info(
            self.account_name, self.container_name)
        container_partition = container_info['partition']
        containers = container_info['nodes']
        req.acl = container_info['write_acl']
        req.environ['swift_sync_key'] = container_info['sync_key']
        object_versions = container_info['versions']
        if 'swift.authorize' in req.environ:
            aresp = req.environ['swift.authorize'](req)
            if aresp:
                return aresp
        if not containers:
            return HTTPNotFound(request=req)
        if 'x-delete-after' in req.headers:
            try:
                x_delete_after = int(req.headers['x-delete-after'])
            except ValueError:
                    return HTTPBadRequest(request=req,
                                          content_type='text/plain',
                                          body='Non-integer X-Delete-After')
            req.headers['x-delete-at'] = '%d' % (time.time() + x_delete_after)
        partition, nodes = self.app.object_ring.get_nodes(
            self.account_name, self.container_name, self.object_name)
        # do a HEAD request for container sync and checking object versions
        if 'x-timestamp' in req.headers or \
                (object_versions and not
                 req.environ.get('swift_versioned_copy')):
            hreq = Request.blank(req.path_info, headers={'X-Newest': 'True'},
                                 environ={'REQUEST_METHOD': 'HEAD'})
            hresp = self.GETorHEAD_base(
                hreq, _('Object'), self.app.object_ring, partition,
                hreq.path_info)
        # Used by container sync feature
        if 'x-timestamp' in req.headers:
            try:
                req.headers['X-Timestamp'] = \
                    normalize_timestamp(float(req.headers['x-timestamp']))
                if hresp.environ and 'swift_x_timestamp' in hresp.environ and \
                    float(hresp.environ['swift_x_timestamp']) >= \
                        float(req.headers['x-timestamp']):
                    return HTTPAccepted(request=req)
            except ValueError:
                return HTTPBadRequest(
                    request=req, content_type='text/plain',
                    body='X-Timestamp should be a UNIX timestamp float value; '
                         'was %r' % req.headers['x-timestamp'])
        else:
            req.headers['X-Timestamp'] = normalize_timestamp(time.time())
        # Sometimes the 'content-type' header exists, but is set to None.
        content_type_manually_set = True
        if not req.headers.get('content-type'):
            guessed_type, _junk = mimetypes.guess_type(req.path_info)
            req.headers['Content-Type'] = guessed_type or \
                'application/octet-stream'
            content_type_manually_set = False
        error_response = check_object_creation(req, self.object_name) or \
            check_content_type(req)
        if error_response:
            return error_response
        if object_versions and not req.environ.get('swift_versioned_copy'):
            is_manifest = 'x-object-manifest' in req.headers or \
                          'x-object-manifest' in hresp.headers
            if hresp.status_int != HTTP_NOT_FOUND and not is_manifest:
                # This is a version manifest and needs to be handled
                # differently. First copy the existing data to a new object,
                # then write the data from this request to the version manifest
                # object.
                lcontainer = object_versions.split('/')[0]
                prefix_len = '%03x' % len(self.object_name)
                lprefix = prefix_len + self.object_name + '/'
                ts_source = hresp.environ.get('swift_x_timestamp')
                if ts_source is None:
                    ts_source = time.mktime(time.strptime(
                                            hresp.headers['last-modified'],
                                            '%a, %d %b %Y %H:%M:%S GMT'))
                new_ts = normalize_timestamp(ts_source)
                vers_obj_name = lprefix + new_ts
                copy_headers = {
                    'Destination': '%s/%s' % (lcontainer, vers_obj_name)}
                copy_environ = {'REQUEST_METHOD': 'COPY',
                                'swift_versioned_copy': True
                                }
                copy_req = Request.blank(req.path_info, headers=copy_headers,
                                         environ=copy_environ)
                copy_resp = self.COPY(copy_req)
                if is_client_error(copy_resp.status_int):
                    # missing container or bad permissions
                    return HTTPPreconditionFailed(request=req)
                elif not is_success(copy_resp.status_int):
                    # could not copy the data, bail
                    return HTTPServiceUnavailable(request=req)

        reader = req.environ['wsgi.input'].read
        data_source = iter(lambda: reader(self.app.client_chunk_size), '')
        source_header = req.headers.get('X-Copy-From')
        source_resp = None
        if source_header:
            source_header = unquote(source_header)
            acct = req.path_info.split('/', 2)[1]
            if isinstance(acct, unicode):
                acct = acct.encode('utf-8')
            if not source_header.startswith('/'):
                source_header = '/' + source_header
            source_header = '/' + acct + source_header
            try:
                src_container_name, src_obj_name = \
                    source_header.split('/', 3)[2:]
            except ValueError:
                return HTTPPreconditionFailed(
                    request=req,
                    body='X-Copy-From header must be of the form'
                         '<container name>/<object name>')
            source_req = req.copy_get()
            source_req.path_info = source_header
            source_req.headers['X-Newest'] = 'true'
            orig_obj_name = self.object_name
            orig_container_name = self.container_name
            self.object_name = src_obj_name
            self.container_name = src_container_name
            source_resp = self.GET(source_req)
            if source_resp.status_int >= HTTP_MULTIPLE_CHOICES:
                return source_resp
            self.object_name = orig_obj_name
            self.container_name = orig_container_name
            new_req = Request.blank(req.path_info,
                                    environ=req.environ, headers=req.headers)
            data_source = source_resp.app_iter
            new_req.content_length = source_resp.content_length
            if new_req.content_length is None:
                # This indicates a transfer-encoding: chunked source object,
                # which currently only happens because there are more than
                # CONTAINER_LISTING_LIMIT segments in a segmented object. In
                # this case, we're going to refuse to do the server-side copy.
                return HTTPRequestEntityTooLarge(request=req)
            if new_req.content_length > MAX_FILE_SIZE:
                return HTTPRequestEntityTooLarge(request=req)
            new_req.etag = source_resp.etag
            # we no longer need the X-Copy-From header
            del new_req.headers['X-Copy-From']
            if not content_type_manually_set:
                new_req.headers['Content-Type'] = \
                    source_resp.headers['Content-Type']
            if not config_true_value(
                    new_req.headers.get('x-fresh-metadata', 'false')):
                copy_headers_into(source_resp, new_req)
                copy_headers_into(req, new_req)
            # copy over x-static-large-object for POSTs and manifest copies
            if 'X-Static-Large-Object' in source_resp.headers and \
                    req.params.get('multipart-manifest') == 'get':
                new_req.headers['X-Static-Large-Object'] = \
                    source_resp.headers['X-Static-Large-Object']

            req = new_req

        if 'x-delete-at' in req.headers:
            try:
                x_delete_at = int(req.headers['x-delete-at'])
                if x_delete_at < time.time():
                    return HTTPBadRequest(
                        body='X-Delete-At in past', request=req,
                        content_type='text/plain')
            except ValueError:
                return HTTPBadRequest(request=req, content_type='text/plain',
                                      body='Non-integer X-Delete-At')
            delete_at_container = str(
                x_delete_at /
                self.app.expiring_objects_container_divisor *
                self.app.expiring_objects_container_divisor)
            delete_at_part, delete_at_nodes = \
                self.app.container_ring.get_nodes(
                    self.app.expiring_objects_account, delete_at_container)
        else:
            delete_at_part = delete_at_nodes = None

        node_iter = GreenthreadSafeIterator(
            self.iter_nodes(self.app.object_ring, partition))
        pile = GreenPile(len(nodes))
        chunked = req.headers.get('transfer-encoding')

        outgoing_headers = self._backend_requests(
            req, len(nodes), container_partition, containers,
            delete_at_part, delete_at_nodes)

        for nheaders in outgoing_headers:
            # RFC2616:8.2.3 disallows 100-continue without a body
            if (req.content_length > 0) or chunked:
                nheaders['Expect'] = '100-continue'
            pile.spawn(self._connect_put_node, node_iter, partition,
                       req.path_info, nheaders, self.app.logger.thread_locals)

        conns = [conn for conn in pile if conn]
        if len(conns) <= len(nodes) / 2:
            self.app.logger.error(
                _('Object PUT returning 503, %(conns)s/%(nodes)s '
                  'required connections'),
                {'conns': len(conns), 'nodes': len(nodes) // 2 + 1})
            return HTTPServiceUnavailable(request=req)
        bytes_transferred = 0
        try:
            with ContextPool(len(nodes)) as pool:
                for conn in conns:
                    conn.failed = False
                    conn.queue = Queue(self.app.put_queue_depth)
                    pool.spawn(self._send_file, conn, req.path)
                while True:
                    with ChunkReadTimeout(self.app.client_timeout):
                        try:
                            chunk = next(data_source)
                        except StopIteration:
                            if chunked:
                                [conn.queue.put('0\r\n\r\n') for conn in conns]
                            break
                    bytes_transferred += len(chunk)
                    if bytes_transferred > MAX_FILE_SIZE:
                        return HTTPRequestEntityTooLarge(request=req)
                    for conn in list(conns):
                        if not conn.failed:
                            conn.queue.put(
                                '%x\r\n%s\r\n' % (len(chunk), chunk)
                                if chunked else chunk)
                        else:
                            conns.remove(conn)
                    if len(conns) <= len(nodes) / 2:
                        self.app.logger.error(_(
                            'Object PUT exceptions during'
                            ' send, %(conns)s/%(nodes)s required connections'),
                            {'conns': len(conns), 'nodes': len(nodes) / 2 + 1})
                        return HTTPServiceUnavailable(request=req)
                for conn in conns:
                    if conn.queue.unfinished_tasks:
                        conn.queue.join()
            conns = [conn for conn in conns if not conn.failed]
        except ChunkReadTimeout, err:
            self.app.logger.warn(
                _('ERROR Client read timeout (%ss)'), err.seconds)
            self.app.logger.increment('client_timeouts')
            return HTTPRequestTimeout(request=req)
Example #14
0
    def PUT(self, req):
        """HTTP PUT request handler."""
        if req.if_none_match is not None and '*' not in req.if_none_match:
            # Sending an etag with if-none-match isn't currently supported
            return HTTPBadRequest(request=req, content_type='text/plain',
                                  body='If-None-Match only supports *')
        container_info = self.container_info(
            self.account_name, self.container_name, req)
        policy_index = req.headers.get('X-Backend-Storage-Policy-Index',
                                       container_info['storage_policy'])
        obj_ring = self.app.get_object_ring(policy_index)
        # pass the policy index to storage nodes via req header
        req.headers['X-Backend-Storage-Policy-Index'] = policy_index
        container_partition = container_info['partition']
        containers = container_info['nodes']
        req.acl = container_info['write_acl']
        req.environ['swift_sync_key'] = container_info['sync_key']
        object_versions = container_info['versions']
        if 'swift.authorize' in req.environ:
            aresp = req.environ['swift.authorize'](req)
            if aresp:
                return aresp
        if not containers:
            return HTTPNotFound(request=req)
        try:
            ml = req.message_length()
        except ValueError as e:
            return HTTPBadRequest(request=req, content_type='text/plain',
                                  body=str(e))
        except AttributeError as e:
            return HTTPNotImplemented(request=req, content_type='text/plain',
                                      body=str(e))
        if ml is not None and ml > constraints.MAX_FILE_SIZE:
            return HTTPRequestEntityTooLarge(request=req)
        if 'x-delete-after' in req.headers:
            try:
                x_delete_after = int(req.headers['x-delete-after'])
            except ValueError:
                return HTTPBadRequest(request=req,
                                      content_type='text/plain',
                                      body='Non-integer X-Delete-After')
            req.headers['x-delete-at'] = normalize_delete_at_timestamp(
                time.time() + x_delete_after)
        partition, nodes = obj_ring.get_nodes(
            self.account_name, self.container_name, self.object_name)
        # do a HEAD request for container sync and checking object versions
        if 'x-timestamp' in req.headers or \
                (object_versions and not
                 req.environ.get('swift_versioned_copy')):
            # make sure proxy-server uses the right policy index
            _headers = {'X-Backend-Storage-Policy-Index': policy_index,
                        'X-Newest': 'True'}
            hreq = Request.blank(req.path_info, headers=_headers,
                                 environ={'REQUEST_METHOD': 'HEAD'})
            hresp = self.GETorHEAD_base(
                hreq, _('Object'), obj_ring, partition,
                hreq.swift_entity_path)
        # Used by container sync feature
        if 'x-timestamp' in req.headers:
            try:
                req_timestamp = Timestamp(req.headers['X-Timestamp'])
                if hresp.environ and 'swift_x_timestamp' in hresp.environ and \
                        hresp.environ['swift_x_timestamp'] >= req_timestamp:
                    return HTTPAccepted(request=req)
            except ValueError:
                return HTTPBadRequest(
                    request=req, content_type='text/plain',
                    body='X-Timestamp should be a UNIX timestamp float value; '
                         'was %r' % req.headers['x-timestamp'])
            req.headers['X-Timestamp'] = req_timestamp.internal
        else:
            req.headers['X-Timestamp'] = Timestamp(time.time()).internal
        # Sometimes the 'content-type' header exists, but is set to None.
        content_type_manually_set = True
        detect_content_type = \
            config_true_value(req.headers.get('x-detect-content-type'))
        if detect_content_type or not req.headers.get('content-type'):
            guessed_type, _junk = mimetypes.guess_type(req.path_info)
            req.headers['Content-Type'] = guessed_type or \
                'application/octet-stream'
            if detect_content_type:
                req.headers.pop('x-detect-content-type')
            else:
                content_type_manually_set = False

        error_response = check_object_creation(req, self.object_name) or \
            check_content_type(req)
        if error_response:
            return error_response
        if object_versions and not req.environ.get('swift_versioned_copy'):
            if hresp.status_int != HTTP_NOT_FOUND:
                # This is a version manifest and needs to be handled
                # differently. First copy the existing data to a new object,
                # then write the data from this request to the version manifest
                # object.
                lcontainer = object_versions.split('/')[0]
                prefix_len = '%03x' % len(self.object_name)
                lprefix = prefix_len + self.object_name + '/'
                ts_source = hresp.environ.get('swift_x_timestamp')
                if ts_source is None:
                    ts_source = time.mktime(time.strptime(
                                            hresp.headers['last-modified'],
                                            '%a, %d %b %Y %H:%M:%S GMT'))
                new_ts = Timestamp(ts_source).internal
                vers_obj_name = lprefix + new_ts
                copy_headers = {
                    'Destination': '%s/%s' % (lcontainer, vers_obj_name)}
                copy_environ = {'REQUEST_METHOD': 'COPY',
                                'swift_versioned_copy': True
                                }
                copy_req = Request.blank(req.path_info, headers=copy_headers,
                                         environ=copy_environ)
                copy_resp = self.COPY(copy_req)
                if is_client_error(copy_resp.status_int):
                    # missing container or bad permissions
                    return HTTPPreconditionFailed(request=req)
                elif not is_success(copy_resp.status_int):
                    # could not copy the data, bail
                    return HTTPServiceUnavailable(request=req)

        reader = req.environ['wsgi.input'].read
        data_source = iter(lambda: reader(self.app.client_chunk_size), '')
        source_header = req.headers.get('X-Copy-From')
        source_resp = None
        if source_header:
            if req.environ.get('swift.orig_req_method', req.method) != 'POST':
                req.environ.setdefault('swift.log_info', []).append(
                    'x-copy-from:%s' % source_header)
            src_container_name, src_obj_name = check_copy_from_header(req)
            ver, acct, _rest = req.split_path(2, 3, True)
            if isinstance(acct, unicode):
                acct = acct.encode('utf-8')
            source_header = '/%s/%s/%s/%s' % (ver, acct,
                                              src_container_name, src_obj_name)
            source_req = req.copy_get()
            # make sure the source request uses it's container_info
            source_req.headers.pop('X-Backend-Storage-Policy-Index', None)
            source_req.path_info = source_header
            source_req.headers['X-Newest'] = 'true'
            orig_obj_name = self.object_name
            orig_container_name = self.container_name
            self.object_name = src_obj_name
            self.container_name = src_container_name
            sink_req = Request.blank(req.path_info,
                                     environ=req.environ, headers=req.headers)
            source_resp = self.GET(source_req)
            # This gives middlewares a way to change the source; for example,
            # this lets you COPY a SLO manifest and have the new object be the
            # concatenation of the segments (like what a GET request gives
            # the client), not a copy of the manifest file.
            hook = req.environ.get(
                'swift.copy_hook',
                (lambda source_req, source_resp, sink_req: source_resp))
            source_resp = hook(source_req, source_resp, sink_req)

            if source_resp.status_int >= HTTP_MULTIPLE_CHOICES:
                return source_resp
            self.object_name = orig_obj_name
            self.container_name = orig_container_name
            data_source = iter(source_resp.app_iter)
            sink_req.content_length = source_resp.content_length
            if sink_req.content_length is None:
                # This indicates a transfer-encoding: chunked source object,
                # which currently only happens because there are more than
                # CONTAINER_LISTING_LIMIT segments in a segmented object. In
                # this case, we're going to refuse to do the server-side copy.
                return HTTPRequestEntityTooLarge(request=req)
            if sink_req.content_length > constraints.MAX_FILE_SIZE:
                return HTTPRequestEntityTooLarge(request=req)
            sink_req.etag = source_resp.etag
            # we no longer need the X-Copy-From header
            del sink_req.headers['X-Copy-From']
            if not content_type_manually_set:
                sink_req.headers['Content-Type'] = \
                    source_resp.headers['Content-Type']
            if not config_true_value(
                    sink_req.headers.get('x-fresh-metadata', 'false')):
                copy_headers_into(source_resp, sink_req)
                copy_headers_into(req, sink_req)
            # copy over x-static-large-object for POSTs and manifest copies
            if 'X-Static-Large-Object' in source_resp.headers and \
                    req.params.get('multipart-manifest') == 'get':
                sink_req.headers['X-Static-Large-Object'] = \
                    source_resp.headers['X-Static-Large-Object']

            req = sink_req

        if 'x-delete-at' in req.headers:
            try:
                x_delete_at = normalize_delete_at_timestamp(
                    int(req.headers['x-delete-at']))
                if int(x_delete_at) < time.time():
                    return HTTPBadRequest(
                        body='X-Delete-At in past', request=req,
                        content_type='text/plain')
            except ValueError:
                return HTTPBadRequest(request=req, content_type='text/plain',
                                      body='Non-integer X-Delete-At')
            req.environ.setdefault('swift.log_info', []).append(
                'x-delete-at:%s' % x_delete_at)
            delete_at_container = normalize_delete_at_timestamp(
                int(x_delete_at) /
                self.app.expiring_objects_container_divisor *
                self.app.expiring_objects_container_divisor)
            delete_at_part, delete_at_nodes = \
                self.app.container_ring.get_nodes(
                    self.app.expiring_objects_account, delete_at_container)
        else:
            delete_at_container = delete_at_part = delete_at_nodes = None

        node_iter = GreenthreadSafeIterator(
            self.iter_nodes_local_first(obj_ring, partition))
        pile = GreenPile(len(nodes))
        te = req.headers.get('transfer-encoding', '')
        chunked = ('chunked' in te)

        outgoing_headers = self._backend_requests(
            req, len(nodes), container_partition, containers,
            delete_at_container, delete_at_part, delete_at_nodes)

        for nheaders in outgoing_headers:
            # RFC2616:8.2.3 disallows 100-continue without a body
            if (req.content_length > 0) or chunked:
                nheaders['Expect'] = '100-continue'
            pile.spawn(self._connect_put_node, node_iter, partition,
                       req.swift_entity_path, nheaders,
                       self.app.logger.thread_locals)

        conns = [conn for conn in pile if conn]
        min_conns = quorum_size(len(nodes))

        if req.if_none_match is not None and '*' in req.if_none_match:
            statuses = [conn.resp.status for conn in conns if conn.resp]
            if HTTP_PRECONDITION_FAILED in statuses:
                # If we find any copy of the file, it shouldn't be uploaded
                self.app.logger.debug(
                    _('Object PUT returning 412, %(statuses)r'),
                    {'statuses': statuses})
                return HTTPPreconditionFailed(request=req)

        if len(conns) < min_conns:
            self.app.logger.error(
                _('Object PUT returning 503, %(conns)s/%(nodes)s '
                  'required connections'),
                {'conns': len(conns), 'nodes': min_conns})
            return HTTPServiceUnavailable(request=req)
        bytes_transferred = 0
        try:
            with ContextPool(len(nodes)) as pool:
                for conn in conns:
                    conn.failed = False
                    conn.queue = Queue(self.app.put_queue_depth)
                    pool.spawn(self._send_file, conn, req.path)
                while True:
                    with ChunkReadTimeout(self.app.client_timeout):
                        try:
                            chunk = next(data_source)
                        except StopIteration:
                            if chunked:
                                for conn in conns:
                                    conn.queue.put('0\r\n\r\n')
                            break
                    bytes_transferred += len(chunk)
                    if bytes_transferred > constraints.MAX_FILE_SIZE:
                        return HTTPRequestEntityTooLarge(request=req)
                    for conn in list(conns):
                        if not conn.failed:
                            conn.queue.put(
                                '%x\r\n%s\r\n' % (len(chunk), chunk)
                                if chunked else chunk)
                        else:
                            conns.remove(conn)
                    if len(conns) < min_conns:
                        self.app.logger.error(_(
                            'Object PUT exceptions during'
                            ' send, %(conns)s/%(nodes)s required connections'),
                            {'conns': len(conns), 'nodes': min_conns})
                        return HTTPServiceUnavailable(request=req)
                for conn in conns:
                    if conn.queue.unfinished_tasks:
                        conn.queue.join()
            conns = [conn for conn in conns if not conn.failed]
        except ChunkReadTimeout as err:
            self.app.logger.warn(
                _('ERROR Client read timeout (%ss)'), err.seconds)
            self.app.logger.increment('client_timeouts')
            return HTTPRequestTimeout(request=req)
        except (Exception, Timeout):
            self.app.logger.exception(
                _('ERROR Exception causing client disconnect'))
            return HTTPClientDisconnect(request=req)
        if req.content_length and bytes_transferred < req.content_length:
            req.client_disconnect = True
            self.app.logger.warn(
                _('Client disconnected without sending enough data'))
            self.app.logger.increment('client_disconnects')
            return HTTPClientDisconnect(request=req)

        statuses, reasons, bodies, etags = self._get_put_responses(req, conns,
                                                                   nodes)

        if len(etags) > 1:
            self.app.logger.error(
                _('Object servers returned %s mismatched etags'), len(etags))
            return HTTPServerError(request=req)
        etag = etags.pop() if len(etags) else None
        resp = self.best_response(req, statuses, reasons, bodies,
                                  _('Object PUT'), etag=etag)
        if source_header:
            resp.headers['X-Copied-From'] = quote(
                source_header.split('/', 3)[3])
            if 'last-modified' in source_resp.headers:
                resp.headers['X-Copied-From-Last-Modified'] = \
                    source_resp.headers['last-modified']
            copy_headers_into(req, resp)
        resp.last_modified = math.ceil(
            float(Timestamp(req.headers['X-Timestamp'])))
        return resp
Example #15
0
 def timeout_reader():
     with ChunkReadTimeout(self.client_timeout):
         return file_like.read(self.network_chunk_size)
Example #16
0
 def timeout_reader():
     with ChunkReadTimeout(self.client_timeout):
         return request.environ['wsgi.input'].read(
             self.network_chunk_size)
Example #17
0
    def _transfer_data(self, req, data_source, conns, nodes):
        min_conns = quorum_size(len(nodes))

        bytes_transferred = 0
        try:
            with ContextPool(len(nodes)) as pool:
                for conn in conns:
                    conn.failed = False
                    conn.queue = Queue(self.app.put_queue_depth)
                    pool.spawn(self._send_file, conn, req.path)
                while True:
                    with ChunkReadTimeout(self.app.client_timeout):
                        try:
                            chunk = next(data_source)
                        except StopIteration:
                            if req.is_chunked:
                                for conn in conns:
                                    conn.queue.put('0\r\n\r\n')
                            break
                    bytes_transferred += len(chunk)
                    if bytes_transferred > constraints.MAX_FILE_SIZE:
                        raise HTTPRequestEntityTooLarge(request=req)
                    for conn in list(conns):
                        if not conn.failed:
                            conn.queue.put('%x\r\n%s\r\n' %
                                           (len(chunk), chunk) if req.
                                           is_chunked else chunk)
                        else:
                            conn.close()
                            conns.remove(conn)
                    self._check_min_conn(
                        req,
                        conns,
                        min_conns,
                        msg='Object PUT exceptions during'
                        ' send, %(conns)s/%(nodes)s required connections')
                for conn in conns:
                    if conn.queue.unfinished_tasks:
                        conn.queue.join()
            conns = [conn for conn in conns if not conn.failed]
            self._check_min_conn(req,
                                 conns,
                                 min_conns,
                                 msg='Object PUT exceptions after last send, '
                                 '%(conns)s/%(nodes)s required connections')
        except ChunkReadTimeout as err:
            self.app.logger.warn(_('ERROR Client read timeout (%ss)'),
                                 err.seconds)
            self.app.logger.increment('client_timeouts')
            raise HTTPRequestTimeout(request=req)
        except HTTPException:
            raise
        except (Exception, Timeout):
            self.app.logger.exception(
                _('ERROR Exception causing client disconnect'))
            raise HTTPClientDisconnect(request=req)
        if req.content_length and bytes_transferred < req.content_length:
            req.client_disconnect = True
            self.app.logger.warn(
                _('Client disconnected without sending enough data'))
            self.app.logger.increment('client_disconnects')
            raise HTTPClientDisconnect(request=req)