/
server.py
1243 lines (1149 loc) · 57.7 KB
/
server.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# Copyright (c) 2010-2012 OpenStack Foundation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
# implied.
# See the License for the specific language governing permissions and
# limitations under the License.
""" Object Server for Swift """
import six.moves.cPickle as pickle
import json
import os
import multiprocessing
import time
import traceback
import socket
import math
from swift import gettext_ as _
from hashlib import md5
from eventlet import sleep, wsgi, Timeout, tpool
from eventlet.greenthread import spawn
from swift.common.utils import public, get_logger, \
config_true_value, timing_stats, replication, \
normalize_delete_at_timestamp, get_log_line, Timestamp, \
get_expirer_container, parse_mime_headers, \
iter_multipart_mime_documents, extract_swift_bytes, safe_json_loads, \
config_auto_int_value
from swift.common.bufferedhttp import http_connect
from swift.common.constraints import check_object_creation, \
valid_timestamp, check_utf8
from swift.common.exceptions import ConnectionTimeout, DiskFileQuarantined, \
DiskFileNotExist, DiskFileCollision, DiskFileNoSpace, DiskFileDeleted, \
DiskFileDeviceUnavailable, DiskFileExpired, ChunkReadTimeout, \
ChunkReadError, DiskFileXattrNotSupported
from swift.obj import ssync_receiver
from swift.common.http import is_success
from swift.common.base_storage_server import BaseStorageServer
from swift.common.header_key_dict import HeaderKeyDict
from swift.common.request_helpers import get_name_and_placement, \
is_user_meta, is_sys_or_user_meta, is_object_transient_sysmeta, \
resolve_etag_is_at_header, is_sys_meta
from swift.common.swob import HTTPAccepted, HTTPBadRequest, HTTPCreated, \
HTTPInternalServerError, HTTPNoContent, HTTPNotFound, \
HTTPPreconditionFailed, HTTPRequestTimeout, HTTPUnprocessableEntity, \
HTTPClientDisconnect, HTTPMethodNotAllowed, Request, Response, \
HTTPInsufficientStorage, HTTPForbidden, HTTPException, HTTPConflict, \
HTTPServerError
from swift.obj.diskfile import DATAFILE_SYSTEM_META, DiskFileRouter
def iter_mime_headers_and_bodies(wsgi_input, mime_boundary, read_chunk_size):
mime_documents_iter = iter_multipart_mime_documents(
wsgi_input, mime_boundary, read_chunk_size)
for file_like in mime_documents_iter:
hdrs = parse_mime_headers(file_like)
yield (hdrs, file_like)
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
def _make_backend_fragments_header(fragments):
if fragments:
result = {}
for ts, frag_list in fragments.items():
result[ts.internal] = frag_list
return json.dumps(result)
return None
class EventletPlungerString(str):
"""
Eventlet won't send headers until it's accumulated at least
eventlet.wsgi.MINIMUM_CHUNK_SIZE bytes or the app iter is exhausted. If we
want to send the response body behind Eventlet's back, perhaps with some
zero-copy wizardry, then we have to unclog the plumbing in eventlet.wsgi
to force the headers out, so we use an EventletPlungerString to empty out
all of Eventlet's buffers.
"""
def __len__(self):
return wsgi.MINIMUM_CHUNK_SIZE + 1
class ObjectController(BaseStorageServer):
"""Implements the WSGI application for the Swift Object Server."""
server_type = 'object-server'
def __init__(self, conf, logger=None):
"""
Creates a new WSGI application for the Swift Object Server. An
example configuration is given at
<source-dir>/etc/object-server.conf-sample or
/etc/swift/object-server.conf-sample.
"""
super(ObjectController, self).__init__(conf)
self.logger = logger or get_logger(conf, log_route='object-server')
self.node_timeout = float(conf.get('node_timeout', 3))
self.container_update_timeout = float(
conf.get('container_update_timeout', 1))
self.conn_timeout = float(conf.get('conn_timeout', 0.5))
self.client_timeout = int(conf.get('client_timeout', 60))
self.disk_chunk_size = int(conf.get('disk_chunk_size', 65536))
self.network_chunk_size = int(conf.get('network_chunk_size', 65536))
self.log_requests = config_true_value(conf.get('log_requests', 'true'))
self.max_upload_time = int(conf.get('max_upload_time', 86400))
self.slow = int(conf.get('slow', 0))
self.keep_cache_private = \
config_true_value(conf.get('keep_cache_private', 'false'))
default_allowed_headers = '''
content-disposition,
content-encoding,
x-delete-at,
x-object-manifest,
x-static-large-object,
'''
extra_allowed_headers = [
header.strip().lower() for header in conf.get(
'allowed_headers', default_allowed_headers).split(',')
if header.strip()
]
self.allowed_headers = set()
for header in extra_allowed_headers:
if header not in DATAFILE_SYSTEM_META:
self.allowed_headers.add(header)
self.auto_create_account_prefix = \
conf.get('auto_create_account_prefix') or '.'
self.expiring_objects_account = self.auto_create_account_prefix + \
(conf.get('expiring_objects_account_name') or 'expiring_objects')
self.expiring_objects_container_divisor = \
int(conf.get('expiring_objects_container_divisor') or 86400)
# Initialization was successful, so now apply the network chunk size
# parameter as the default read / write buffer size for the network
# sockets.
#
# NOTE WELL: This is a class setting, so until we get set this on a
# per-connection basis, this affects reading and writing on ALL
# sockets, those between the proxy servers and external clients, and
# those between the proxy servers and the other internal servers.
#
# ** Because the primary motivation for this is to optimize how data
# is written back to the proxy server, we could use the value from the
# disk_chunk_size parameter. However, it affects all created sockets
# using this class so we have chosen to tie it to the
# network_chunk_size parameter value instead.
socket._fileobject.default_bufsize = self.network_chunk_size
# Provide further setup specific to an object server implementation.
self.setup(conf)
def setup(self, conf):
"""
Implementation specific setup. This method is called at the very end
by the constructor to allow a specific implementation to modify
existing attributes or add its own attributes.
:param conf: WSGI configuration parameter
"""
# Common on-disk hierarchy shared across account, container and object
# servers.
self._diskfile_router = DiskFileRouter(conf, self.logger)
# This is populated by global_conf_callback way below as the semaphore
# is shared by all workers.
if 'replication_semaphore' in conf:
# The value was put in a list so it could get past paste
self.replication_semaphore = conf['replication_semaphore'][0]
else:
self.replication_semaphore = None
self.replication_failure_threshold = int(
conf.get('replication_failure_threshold') or 100)
self.replication_failure_ratio = float(
conf.get('replication_failure_ratio') or 1.0)
servers_per_port = int(conf.get('servers_per_port', '0') or 0)
if servers_per_port:
# The typical servers-per-port deployment also uses one port per
# disk, so you really get N servers per disk. In that case,
# having a pool of 20 threads per server per disk is far too
# much. For example, given a 60-disk chassis and 4 servers per
# disk, the default configuration will give us 21 threads per
# server (the main thread plus the twenty tpool threads), for a
# total of around 60 * 21 * 4 = 5040 threads. This is clearly
# too high.
#
# Instead, we use a tpool size of 1, giving us 2 threads per
# process. In the example above, that's 60 * 2 * 4 = 480
# threads, which is reasonable since there are 240 processes.
default_tpool_size = 1
else:
# If we're not using servers-per-port, then leave the tpool size
# alone. The default (20) is typically good enough for one
# object server handling requests for many disks.
default_tpool_size = None
tpool_size = config_auto_int_value(
conf.get('eventlet_tpool_num_threads'),
default_tpool_size)
if tpool_size:
tpool.set_num_threads(tpool_size)
def get_diskfile(self, device, partition, account, container, obj,
policy, **kwargs):
"""
Utility method for instantiating a DiskFile object supporting a given
REST API.
An implementation of the object server that wants to use a different
DiskFile class would simply over-ride this method to provide that
behavior.
"""
return self._diskfile_router[policy].get_diskfile(
device, partition, account, container, obj, policy, **kwargs)
def async_update(self, op, account, container, obj, host, partition,
contdevice, headers_out, objdevice, policy,
logger_thread_locals=None):
"""
Sends or saves an async update.
:param op: operation performed (ex: 'PUT', or 'DELETE')
:param account: account name for the object
:param container: container name for the object
:param obj: object name
:param host: host that the container is on
:param partition: partition that the container is on
:param contdevice: device name that the container is on
:param headers_out: dictionary of headers to send in the container
request
:param objdevice: device name that the object is in
:param policy: the associated BaseStoragePolicy instance
:param logger_thread_locals: The thread local values to be set on the
self.logger to retain transaction
logging information.
"""
if logger_thread_locals:
self.logger.thread_locals = logger_thread_locals
headers_out['user-agent'] = 'object-server %s' % os.getpid()
full_path = '/%s/%s/%s' % (account, container, obj)
if all([host, partition, contdevice]):
try:
with ConnectionTimeout(self.conn_timeout):
ip, port = host.rsplit(':', 1)
conn = http_connect(ip, port, contdevice, partition, op,
full_path, headers_out)
with Timeout(self.node_timeout):
response = conn.getresponse()
response.read()
if is_success(response.status):
return
else:
self.logger.error(_(
'ERROR Container update failed '
'(saving for async update later): %(status)d '
'response from %(ip)s:%(port)s/%(dev)s'),
{'status': response.status, 'ip': ip, 'port': port,
'dev': contdevice})
except (Exception, Timeout):
self.logger.exception(_(
'ERROR container update failed with '
'%(ip)s:%(port)s/%(dev)s (saving for async update later)'),
{'ip': ip, 'port': port, 'dev': contdevice})
data = {'op': op, 'account': account, 'container': container,
'obj': obj, 'headers': headers_out}
timestamp = headers_out.get('x-meta-timestamp',
headers_out.get('x-timestamp'))
self._diskfile_router[policy].pickle_async_update(
objdevice, account, container, obj, data, timestamp, policy)
def container_update(self, op, account, container, obj, request,
headers_out, objdevice, policy):
"""
Update the container when objects are updated.
:param op: operation performed (ex: 'PUT', or 'DELETE')
:param account: account name for the object
:param container: container name for the object
:param obj: object name
:param request: the original request object driving the update
:param headers_out: dictionary of headers to send in the container
request(s)
:param objdevice: device name that the object is in
:param policy: the BaseStoragePolicy instance
"""
headers_in = request.headers
conthosts = [h.strip() for h in
headers_in.get('X-Container-Host', '').split(',')]
contdevices = [d.strip() for d in
headers_in.get('X-Container-Device', '').split(',')]
contpartition = headers_in.get('X-Container-Partition', '')
if len(conthosts) != len(contdevices):
# This shouldn't happen unless there's a bug in the proxy,
# but if there is, we want to know about it.
self.logger.error(_(
'ERROR Container update failed: different '
'numbers of hosts and devices in request: '
'"%(hosts)s" vs "%(devices)s"') % {
'hosts': headers_in.get('X-Container-Host', ''),
'devices': headers_in.get('X-Container-Device', '')})
return
if contpartition:
updates = zip(conthosts, contdevices)
else:
updates = []
headers_out['x-trans-id'] = headers_in.get('x-trans-id', '-')
headers_out['referer'] = request.as_referer()
headers_out['X-Backend-Storage-Policy-Index'] = int(policy)
update_greenthreads = []
for conthost, contdevice in updates:
gt = spawn(self.async_update, op, account, container, obj,
conthost, contpartition, contdevice, headers_out,
objdevice, policy,
logger_thread_locals=self.logger.thread_locals)
update_greenthreads.append(gt)
# Wait a little bit to see if the container updates are successful.
# If we immediately return after firing off the greenthread above, then
# we're more likely to confuse the end-user who does a listing right
# after getting a successful response to the object create. The
# `container_update_timeout` bounds the length of time we wait so that
# one slow container server doesn't make the entire request lag.
try:
with Timeout(self.container_update_timeout):
for gt in update_greenthreads:
gt.wait()
except Timeout:
# updates didn't go through, log it and return
self.logger.debug(
'Container update timeout (%.4fs) waiting for %s',
self.container_update_timeout, updates)
def delete_at_update(self, op, delete_at, account, container, obj,
request, objdevice, policy):
"""
Update the expiring objects container when objects are updated.
:param op: operation performed (ex: 'PUT', or 'DELETE')
:param delete_at: scheduled delete in UNIX seconds, int
:param account: account name for the object
:param container: container name for the object
:param obj: object name
:param request: the original request driving the update
:param objdevice: device name that the object is in
:param policy: the BaseStoragePolicy instance (used for tmp dir)
"""
if config_true_value(
request.headers.get('x-backend-replication', 'f')):
return
delete_at = normalize_delete_at_timestamp(delete_at)
updates = [(None, None)]
partition = None
hosts = contdevices = [None]
headers_in = request.headers
headers_out = HeaderKeyDict({
# system accounts are always Policy-0
'X-Backend-Storage-Policy-Index': 0,
'x-timestamp': request.timestamp.internal,
'x-trans-id': headers_in.get('x-trans-id', '-'),
'referer': request.as_referer()})
if op != 'DELETE':
delete_at_container = headers_in.get('X-Delete-At-Container', None)
if not delete_at_container:
self.logger.warning(
'X-Delete-At-Container header must be specified for '
'expiring objects background %s to work properly. Making '
'best guess as to the container name for now.' % op)
# TODO(gholt): In a future release, change the above warning to
# a raised exception and remove the guess code below.
delete_at_container = get_expirer_container(
delete_at, self.expiring_objects_container_divisor,
account, container, obj)
partition = headers_in.get('X-Delete-At-Partition', None)
hosts = headers_in.get('X-Delete-At-Host', '')
contdevices = headers_in.get('X-Delete-At-Device', '')
updates = [upd for upd in
zip((h.strip() for h in hosts.split(',')),
(c.strip() for c in contdevices.split(',')))
if all(upd) and partition]
if not updates:
updates = [(None, None)]
headers_out['x-size'] = '0'
headers_out['x-content-type'] = 'text/plain'
headers_out['x-etag'] = 'd41d8cd98f00b204e9800998ecf8427e'
else:
# DELETEs of old expiration data have no way of knowing what the
# old X-Delete-At-Container was at the time of the initial setting
# of the data, so a best guess is made here.
# Worst case is a DELETE is issued now for something that doesn't
# exist there and the original data is left where it is, where
# it will be ignored when the expirer eventually tries to issue the
# object DELETE later since the X-Delete-At value won't match up.
delete_at_container = get_expirer_container(
delete_at, self.expiring_objects_container_divisor,
account, container, obj)
delete_at_container = normalize_delete_at_timestamp(
delete_at_container)
for host, contdevice in updates:
self.async_update(
op, self.expiring_objects_account, delete_at_container,
'%s-%s/%s/%s' % (delete_at, account, container, obj),
host, partition, contdevice, headers_out, objdevice,
policy)
def _make_timeout_reader(self, file_like):
def timeout_reader():
with ChunkReadTimeout(self.client_timeout):
try:
return file_like.read(self.network_chunk_size)
except (IOError, ValueError):
raise ChunkReadError
return timeout_reader
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 ChunkReadError:
raise HTTPClientDisconnect()
except ChunkReadTimeout:
raise HTTPRequestTimeout()
except StopIteration:
raise HTTPBadRequest(body="couldn't find PUT commit MIME doc")
return rcvd_commit
def _read_metadata_footer(self, mime_documents_iter):
try:
with ChunkReadTimeout(self.client_timeout):
footer_hdrs, footer_iter = next(mime_documents_iter)
except ChunkReadError:
raise HTTPClientDisconnect()
except ChunkReadTimeout:
raise HTTPRequestTimeout()
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 ChunkReadError:
raise HTTPClientDisconnect()
except ChunkReadTimeout:
raise HTTPRequestTimeout()
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")
def _check_container_override(self, update_headers, metadata,
footers=None):
"""
Applies any overrides to the container update headers.
Overrides may be in the x-object-sysmeta-container-update- namespace or
the x-backend-container-update-override- namespace. The former is
preferred and is used by proxy middlewares. The latter is historical
but is still used with EC policy PUT requests; for backwards
compatibility the header names used with EC policy requests have not
been changed to the sysmeta namespace - that way the EC PUT path of a
newer proxy will remain compatible with an object server that pre-dates
the introduction of the x-object-sysmeta-container-update- namespace
and vice-versa.
:param update_headers: a dict of headers used in the container update
:param metadata: a dict that may container override items
:param footers: another dict that may container override items, at a
higher priority than metadata
"""
footers = footers or {}
# the order of this list is significant:
# x-object-sysmeta-container-update-override-* headers take precedence
# over x-backend-container-update-override-* headers
override_prefixes = ['x-backend-container-update-override-',
'x-object-sysmeta-container-update-override-']
for override_prefix in override_prefixes:
for key, val in metadata.items():
if key.lower().startswith(override_prefix):
override = key.lower().replace(override_prefix, 'x-')
update_headers[override] = val
# apply x-backend-container-update-override* from footers *before*
# x-object-sysmeta-container-update-override-* from headers
for key, val in footers.items():
if key.lower().startswith(override_prefix):
override = key.lower().replace(override_prefix, 'x-')
update_headers[override] = val
def _preserve_slo_manifest(self, update_metadata, orig_metadata):
if 'X-Static-Large-Object' in orig_metadata:
update_metadata['X-Static-Large-Object'] = \
orig_metadata['X-Static-Large-Object']
@public
@timing_stats()
def POST(self, request):
"""Handle HTTP POST 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)
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')
next_part_power = request.headers.get('X-Backend-Next-Part-Power')
try:
disk_file = self.get_diskfile(
device, partition, account, container, obj,
policy=policy, open_expired=config_true_value(
request.headers.get('x-backend-replication', 'false')),
next_part_power=next_part_power)
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):
return HTTPNotFound(request=request)
orig_timestamp = Timestamp(orig_metadata.get('X-Timestamp', 0))
orig_ctype_timestamp = disk_file.content_type_timestamp
req_ctype_time = '0'
req_ctype = request.headers.get('Content-Type')
if req_ctype:
req_ctype_time = request.headers.get('Content-Type-Timestamp',
req_timestamp.internal)
req_ctype_timestamp = Timestamp(req_ctype_time)
if orig_timestamp >= req_timestamp \
and orig_ctype_timestamp >= req_ctype_timestamp:
return HTTPConflict(
request=request,
headers={'X-Backend-Timestamp': orig_timestamp.internal})
if req_timestamp > orig_timestamp:
metadata = {'X-Timestamp': req_timestamp.internal}
self._preserve_slo_manifest(metadata, orig_metadata)
metadata.update(val for val in request.headers.items()
if (is_user_meta('object', val[0]) or
is_object_transient_sysmeta(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]
orig_delete_at = int(orig_metadata.get('X-Delete-At') or 0)
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)
else:
# preserve existing metadata, only content-type may be updated
metadata = dict(disk_file.get_metafile_metadata())
if req_ctype_timestamp > orig_ctype_timestamp:
# we have a new content-type, add to metadata and container update
content_type_headers = {
'Content-Type': request.headers['Content-Type'],
'Content-Type-Timestamp': req_ctype_timestamp.internal
}
metadata.update(content_type_headers)
else:
# send existing content-type with container update
content_type_headers = {
'Content-Type': disk_file.content_type,
'Content-Type-Timestamp': orig_ctype_timestamp.internal
}
if orig_ctype_timestamp != disk_file.data_timestamp:
# only add to metadata if it's not the datafile content-type
metadata.update(content_type_headers)
try:
disk_file.write_metadata(metadata)
except (DiskFileXattrNotSupported, DiskFileNoSpace):
return HTTPInsufficientStorage(drive=device, request=request)
if (content_type_headers['Content-Type-Timestamp']
!= disk_file.data_timestamp):
# Current content-type is not from the datafile, but the datafile
# content-type may have a swift_bytes param that was appended by
# SLO and we must continue to send that with the container update.
# Do this (rather than use a separate header) for backwards
# compatibility because there may be 'legacy' container updates in
# async pending that have content-types with swift_bytes params, so
# we have to be able to handle those in container server anyway.
_, swift_bytes = extract_swift_bytes(
disk_file.get_datafile_metadata()['Content-Type'])
if swift_bytes:
content_type_headers['Content-Type'] += (';swift_bytes=%s'
% swift_bytes)
update_headers = HeaderKeyDict({
'x-size': orig_metadata['Content-Length'],
'x-content-type': content_type_headers['Content-Type'],
'x-timestamp': disk_file.data_timestamp.internal,
'x-content-type-timestamp':
content_type_headers['Content-Type-Timestamp'],
'x-meta-timestamp': metadata['X-Timestamp'],
'x-etag': orig_metadata['ETag']})
# Special cases for backwards compatibility.
# For EC policy, send X-Object-Sysmeta-Ec-Etag which is same as the
# X-Backend-Container-Update-Override-Etag value sent with the original
# PUT. Similarly send X-Object-Sysmeta-Ec-Content-Length which is the
# same as the X-Backend-Container-Update-Override-Size value. We have
# to send Etag and size with a POST container update because the
# original PUT container update may have failed or be in async_pending.
if 'X-Object-Sysmeta-Ec-Etag' in orig_metadata:
update_headers['X-Etag'] = orig_metadata[
'X-Object-Sysmeta-Ec-Etag']
if 'X-Object-Sysmeta-Ec-Content-Length' in orig_metadata:
update_headers['X-Size'] = orig_metadata[
'X-Object-Sysmeta-Ec-Content-Length']
self._check_container_override(update_headers, orig_metadata)
# object POST updates are PUT to the container server
self.container_update(
'PUT', account, container, obj, request, update_headers,
device, policy)
# Add sysmeta to response
resp_headers = {}
for key, value in orig_metadata.items():
if is_sys_meta('object', key):
resp_headers[key] = value
return HTTPAccepted(request=request, headers=resp_headers)
@public
@timing_stats()
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)
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')
next_part_power = request.headers.get('X-Backend-Next-Part-Power')
try:
disk_file = self.get_diskfile(
device, partition, account, container, obj,
policy=policy, frag_index=frag_index,
next_part_power=next_part_power)
except DiskFileDeviceUnavailable:
return HTTPInsufficientStorage(drive=device, request=request)
try:
orig_metadata = disk_file.read_metadata()
orig_timestamp = disk_file.data_timestamp
except DiskFileXattrNotSupported:
return HTTPInsufficientStorage(drive=device, request=request)
except DiskFileDeleted as e:
orig_metadata = {}
orig_timestamp = e.timestamp
except (DiskFileNotExist, DiskFileQuarantined):
orig_metadata = {}
orig_timestamp = Timestamp(0)
# 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)
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 ChunkReadError:
return HTTPClientDisconnect(request=request)
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 ChunkReadError:
return HTTPClientDisconnect(request=request)
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]) or
is_object_transient_sysmeta(val[0])))
metadata.update(val for val in footer_meta.items()
if (is_sys_or_user_meta('object', val[0]) or
is_object_transient_sysmeta(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 (when required), call commit 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 ChunkReadError:
raise HTTPClientDisconnect()
except ChunkReadTimeout:
raise HTTPRequestTimeout()
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,
footer_meta)
self.container_update(
'PUT', account, container, obj, request,
update_headers,
device, policy)
return HTTPCreated(request=request, etag=etag)
@public
@timing_stats()
def GET(self, request):
"""Handle HTTP GET requests for the Swift Object Server."""
device, partition, account, container, obj, policy = \
get_name_and_placement(request, 5, 5, True)
frag_prefs = safe_json_loads(
request.headers.get('X-Backend-Fragment-Preferences'))
try:
disk_file = self.get_diskfile(
device, partition, account, container, obj,
policy=policy, frag_prefs=frag_prefs,
open_expired=config_true_value(
request.headers.get('x-backend-replication', 'false')))
except DiskFileDeviceUnavailable:
return HTTPInsufficientStorage(drive=device, request=request)
try:
with disk_file.open():
metadata = disk_file.get_metadata()
obj_size = int(metadata['Content-Length'])
file_x_ts = Timestamp(metadata['X-Timestamp'])
keep_cache = (self.keep_cache_private or
('X-Auth-Token' not in request.headers and
'X-Storage-Token' not in request.headers))
conditional_etag = resolve_etag_is_at_header(request, metadata)
response = Response(
app_iter=disk_file.reader(keep_cache=keep_cache),
request=request, conditional_response=True,
conditional_etag=conditional_etag)
response.headers['Content-Type'] = metadata.get(
'Content-Type', 'application/octet-stream')
for key, value in metadata.items():
if (is_sys_or_user_meta('object', key) or
is_object_transient_sysmeta(key) or
key.lower() in self.allowed_headers):
response.headers[key] = value
response.etag = metadata['ETag']
response.last_modified = math.ceil(float(file_x_ts))
response.content_length = obj_size
try:
response.content_encoding = metadata[
'Content-Encoding']
except KeyError:
pass
response.headers['X-Timestamp'] = file_x_ts.normal
response.headers['X-Backend-Timestamp'] = file_x_ts.internal
response.headers['X-Backend-Data-Timestamp'] = \
disk_file.data_timestamp.internal
if disk_file.durable_timestamp:
response.headers['X-Backend-Durable-Timestamp'] = \
disk_file.durable_timestamp.internal
response.headers['X-Backend-Fragments'] = \
_make_backend_fragments_header(disk_file.fragments)
resp = request.get_response(response)
except DiskFileXattrNotSupported:
return HTTPInsufficientStorage(drive=device, request=request)
except (DiskFileNotExist, DiskFileQuarantined) as e:
headers = {}
if hasattr(e, 'timestamp'):
headers['X-Backend-Timestamp'] = e.timestamp.internal
resp = HTTPNotFound(request=request, headers=headers,
conditional_response=True)
return resp
@public
@timing_stats(sample_rate=0.8)
def HEAD(self, request):
"""Handle HTTP HEAD requests for the Swift Object Server."""
device, partition, account, container, obj, policy = \
get_name_and_placement(request, 5, 5, True)
frag_prefs = safe_json_loads(
request.headers.get('X-Backend-Fragment-Preferences'))
try:
disk_file = self.get_diskfile(
device, partition, account, container, obj,
policy=policy, frag_prefs=frag_prefs)
except DiskFileDeviceUnavailable:
return HTTPInsufficientStorage(drive=device, request=request)
try:
metadata = disk_file.read_metadata()
except DiskFileXattrNotSupported:
return HTTPInsufficientStorage(drive=device, request=request)
except (DiskFileNotExist, DiskFileQuarantined) as e:
headers = {}
if hasattr(e, 'timestamp'):
headers['X-Backend-Timestamp'] = e.timestamp.internal
return HTTPNotFound(request=request, headers=headers,
conditional_response=True)
conditional_etag = resolve_etag_is_at_header(request, metadata)
response = Response(request=request, conditional_response=True,
conditional_etag=conditional_etag)
response.headers['Content-Type'] = metadata.get(
'Content-Type', 'application/octet-stream')
for key, value in metadata.items():
if (is_sys_or_user_meta('object', key) or
is_object_transient_sysmeta(key) or
key.lower() in self.allowed_headers):
response.headers[key] = value
response.etag = metadata['ETag']
ts = Timestamp(metadata['X-Timestamp'])
response.last_modified = math.ceil(float(ts))