migration.py 39.2 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
# Copyright 2018 Markus Scheidgen
#
# 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.

"""
This module contains functions to read data from NOMAD coe, external sources,
other/older nomad@FAIRDI instances to mass upload it to a new nomad@FAIRDI instance.
18
19
20

.. autoclass:: NomadCOEMigration
.. autoclass:: SourceCalc
21
22
"""

Markus Scheidgen's avatar
Markus Scheidgen committed
23
from typing import Generator, Tuple, List, Iterable, IO, Any, Dict
Markus Scheidgen's avatar
Markus Scheidgen committed
24
import os
25
import os.path
26
import zipstream
27
import zipfile
28
import math
Markus Scheidgen's avatar
Markus Scheidgen committed
29
from mongoengine import Document, IntField, StringField, DictField, ListField
30
import time
31
import datetime
32
from bravado.exception import HTTPNotFound, HTTPBadRequest, HTTPGatewayTimeout
33
import glob
34
import os
35
import runstats
Markus Scheidgen's avatar
Markus Scheidgen committed
36
import io
Markus Scheidgen's avatar
Markus Scheidgen committed
37
import threading
38

39
from nomad import utils, infrastructure, config
40
from nomad.coe_repo import User, Calc, LoginException
41
42
from nomad.datamodel import CalcWithMetadata
from nomad.processing import FAILURE, SUCCESS
43
44


45
default_pid_prefix = 7000000
46
47
""" The default pid prefix for new non migrated calculations """

Markus Scheidgen's avatar
Markus Scheidgen committed
48
max_package_size = 16 * 1024 * 1024 * 1024  # 16 GB
49
""" The maximum size of a package that will be used as an upload on nomad@FAIRDI """
50
use_stats_for_filestats_threshold = 1024
51

52
53
54
default_comment = 'entry with unknown provernance'
default_uploader = dict(id=1)

55

Markus Scheidgen's avatar
Markus Scheidgen committed
56
57
58
59
60
61
62
63
64
65
66
67
def iterable_to_stream(iterable, buffer_size=io.DEFAULT_BUFFER_SIZE):
    """
    Lets you use an iterable (e.g. a generator) that yields bytestrings as a read-only
    input stream.

    The stream implements Python 3's newer I/O API (available in Python 2's io module).
    For efficiency, the stream is buffered.
    """
    class IterStream(io.RawIOBase):
        def __init__(self):
            self.leftover = None
            self.iterator = iter(iterable)
68

Markus Scheidgen's avatar
Markus Scheidgen committed
69
70
        def readable(self):
            return True
71

Markus Scheidgen's avatar
Markus Scheidgen committed
72
        def readinto(self, b):
73
            requested_len = len(b)  # We're supposed to return at most this much
Markus Scheidgen's avatar
Markus Scheidgen committed
74
75
76
            while True:
                try:
                    chunk = next(self.iterator)
77
                except StopIteration:
Markus Scheidgen's avatar
Markus Scheidgen committed
78
79
80
                    if len(self.leftover) == 0:
                        return 0  # indicate EOF
                    chunk = self.leftover
81
                output, self.leftover = chunk[:requested_len], chunk[requested_len:]
Markus Scheidgen's avatar
Markus Scheidgen committed
82
83
84
85
86
                len_output = len(output)
                if len_output == 0:
                    continue  # do not prematurely indicate EOF
                b[:len_output] = output
                return len_output
87

Markus Scheidgen's avatar
Markus Scheidgen committed
88
89
90
    return io.BufferedReader(IterStream(), buffer_size=buffer_size)


91
92
93
94
95
96
97
98
99
100
101
class Package(Document):
    """
    A Package represents split origin NOMAD CoE uploads. We use packages as uploads
    in nomad@FAIRDI. Some of the uploads in nomad are very big (alfow lib) and need
    to be split down to yield practical (i.e. for mirrors) upload sizes. Therefore,
    uploads are split over multiple packages if one upload gets to large. A package
    always contains full directories of files to preserve *mainfile* *aux* file relations.
    """

    package_id = StringField(primary_key=True)
    """ A random UUID for the package. Could serve later is target upload id."""
Markus Scheidgen's avatar
Markus Scheidgen committed
102
103
104
105
    filenames = ListField(StringField(), default=[])
    """ The files in the package relative to the upload path """
    upload_path = StringField()
    """ The absolute path of the source upload """
106
107
108
109
110
111
112
    upload_id = StringField()
    """ The source upload_id. There might be multiple packages per upload (this is the point). """
    restricted = IntField()
    """ The restricted in month, 0 for unrestricted """
    size = IntField()
    """ The sum of all file sizes """

113
    migration_version = IntField()
Markus Scheidgen's avatar
Markus Scheidgen committed
114
    report = DictField()
115
116

    meta = dict(indexes=['upload_id', 'migration_version'])
117

118
    def open_package_upload_file(self) -> IO:
119
        """ Creates a streaming zip file from the files of this package. """
120
        zip_file = zipstream.ZipFile(compression=zipstream.ZIP_STORED, allowZip64=True)
121
122
        for filename in self.filenames:
            filepath = os.path.join(self.upload_path, filename)
123
            zip_file.write(filepath, filename)
124

Markus Scheidgen's avatar
Markus Scheidgen committed
125
        return iterable_to_stream(zip_file)  # type: ignore
126

127
128
129
130
131
    def create_package_upload_file(self) -> Tuple[str, bool]:
        """
        Creates a zip file for the package in tmp and returns its path and whether it
        was created (``True``) or existed before (``False``).
        """
132
        upload_filepath = os.path.join(config.fs.nomad_tmp, '%s.zip' % self.package_id)
Markus Scheidgen's avatar
Markus Scheidgen committed
133
134
        if not os.path.exists(os.path.dirname(upload_filepath)):
            os.mkdir(os.path.dirname(upload_filepath))
135
136
137
138
139
140
141
        if not os.path.isfile(upload_filepath):
            with zipfile.ZipFile(
                    upload_filepath, 'w',
                    compression=zipfile.ZIP_STORED, allowZip64=True) as zip_file:
                for filename in self.filenames:
                    filepath = os.path.join(self.upload_path, filename)
                    zip_file.write(filepath, filename)
142
143
144
            created = True
        else:
            created = False
145

146
        return upload_filepath, created
147

148
149
150
151
152
153
154
155
156
157
    @classmethod
    def index(cls, *upload_paths):
        """
        Creates Package objects for the given uploads in nomad. The given uploads are
        supposed to be path to the extracted upload directories. If the upload is already
        in the db, the upload is skipped entirely.
        """
        logger = utils.get_logger(__name__)

        for upload_path in upload_paths:
Markus Scheidgen's avatar
Markus Scheidgen committed
158
159
160
161
162
163
            try:
                stats = runstats.Statistics()
                upload_path = os.path.abspath(upload_path)
                upload_id = os.path.basename(upload_path)
                if cls.objects(upload_id=upload_id).first() is not None:
                    logger.info('upload already exists, skip', upload_id=upload_id)
164
165
                    continue

Markus Scheidgen's avatar
Markus Scheidgen committed
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
                restrict_files = glob.glob(os.path.join(upload_path, 'RESTRICTED_*'))
                month = 0
                for restrict_file in restrict_files:
                    restrict_file = os.path.basename(restrict_file)
                    try:
                        new_month = int(restrict_file[len('RESTRICTED_'):])
                        if new_month > month:
                            month = new_month
                    except Exception:
                        month = 36
                if month > 36:
                    month = 36
                restricted = month

                def create_package():
                    cls.timer = time.time()
                    package = Package(
                        package_id=utils.create_uuid(),
                        upload_id=upload_id,
                        upload_path=upload_path,
                        restricted=restricted,
                        size=0)
                    return package

                def save_package(package):
                    if len(package.filenames) == 0:
                        return

                    if package.size > max_package_size:
                        # a single directory seems to big for a package
                        logger.error(
                            'directory exceeds max package size', directory=upload_path, size=package.size)

                    package.save()
                    logger.info(
                        'created package',
                        size=package.size,
                        files=len(package.filenames),
                        package_id=package.package_id,
                        exec_time=time.time() - cls.timer,
                        upload_id=package.upload_id)

                package = create_package()

                for root, _, files in os.walk(upload_path):
                    directory_filenames: List[str] = []
                    directory_size = 0

                    if len(files) == 0:
                        continue

                    for file in files:
                        filepath = os.path.join(root, file)
                        filename = filepath[len(upload_path) + 1:]
                        directory_filenames.append(filename)
                        # getting file stats is pretty expensive with gpfs
                        # if an upload has more then 1000 files, its pretty likely that
                        # size patterns repeat ... goood enough
                        if len(stats) < use_stats_for_filestats_threshold:
                            filesize = os.path.getsize(filepath)
                            stats.push(filesize)
                        else:
                            filesize = stats.mean()
                        directory_size += filesize

                    if (package.size + directory_size) > max_package_size and package.size > 0:
                        save_package(package)
                        package = create_package()

                    for filename in directory_filenames:
                        package.filenames.append(filename)
                    package.size += directory_size

                    logger.debug('packaged directory', directory=root, size=directory_size)

                save_package(package)

                logger.info('completed upload', directory=upload_path, upload_id=upload_id)
            except Exception as e:
                logger.error(
246
                    'could create package from upload',
Markus Scheidgen's avatar
Markus Scheidgen committed
247
248
                    upload_path=upload_path, upload_id=upload_id, exc_info=e)
                continue
Markus Scheidgen's avatar
Markus Scheidgen committed
249

250

251
class SourceCalc(Document):
252
    """
253
    Mongo document used as a calculation, upload, and metadata db and index
254
255
256
257
258
    build from a given source db. Each :class:`SourceCacl` entry relates
    a pid, mainfile, upload "id" with each other for a corressponding calculation.
    It might alos contain the user metadata. The uploads are "id"ed via the
    specific path segment that identifies an upload on the CoE repo FS(s) without
    any prefixes (e.g. $EXTRACTED, /data/upload, etc.)
259
260
261
262
263
264
    """
    pid = IntField(primary_key=True)
    mainfile = StringField()
    upload = StringField()
    metadata = DictField()

265
    migration_version = IntField()
266

267
268
    extracted_prefix = '$EXTRACTED/'
    sites = ['/data/nomad/extracted/', '/nomad/repository/extracted/']
269
270
    prefixes = [extracted_prefix] + sites

271
    meta = dict(indexes=['upload', 'mainfile', 'migration_version'])
272

273
    _dataset_cache: dict = {}
274
275

    @staticmethod
276
277
278
279
    def index(source, drop: bool = False, with_metadata: bool = True, per_query: int = 100) \
            -> Generator[Tuple['SourceCalc', int], None, None]:
        """
        Creates a collection of :class:`SourceCalc` documents that represent source repo
280
        db entries.
281
282
283

        Arguments:
            source: The source db sql alchemy session
284
285
286
287
288
289
290
            drop: True to drop and create a new collection, update the existing otherwise,
                default is False.
            with_metadata: True to also grab all metadata and store it, default is True.
            per_query: The implementation tries to grab almost all data with a heavely joined
                query on the CoE snoflake/star shaped schema.
                The query cannot ask for the whole db at once: choose how many calculations
                should be read at a time to optimize for your application.
291
292

        Returns:
293
            yields tuples (:class:`SourceCalc`, #calcs_total[incl. datasets])
294
        """
295
        logger = utils.get_logger(__name__)
296
297
298
        if drop:
            SourceCalc.drop_collection()

299
        last_source_calc = SourceCalc.objects().order_by('-pid').first()
300
301
        start_pid = last_source_calc.pid if last_source_calc is not None else 0
        source_query = source.query(Calc)
302
        total = source_query.count() - SourceCalc.objects.count()
303
304

        while True:
305
306
            query_timer = utils.timer(logger, 'query source db')
            query_timer.__enter__()  # pylint: disable=E1101
Markus Scheidgen's avatar
Markus Scheidgen committed
307
            calcs: Iterable[Calc] = source_query \
308
309
310
                .filter(Calc.coe_calc_id > start_pid) \
                .order_by(Calc.coe_calc_id) \
                .limit(per_query)
Markus Scheidgen's avatar
Markus Scheidgen committed
311

312
313
            source_calcs = []
            for calc in calcs:
314
                query_timer.__exit__(None, None, None)  # pylint: disable=E1101
315
                try:
Markus Scheidgen's avatar
Markus Scheidgen committed
316
317
                    filenames = calc.files
                    if filenames is None or len(filenames) == 0:
318
319
320
                        continue  # dataset case

                    filename = filenames[0]
321
322
323
                    if len(filenames) == 1 and (filename.endswith('.tgz') or filename.endswith('.zip')):
                        continue  # also a dataset, some datasets have a downloadable archive

324
325
326
327
328
329
330
331
                    for prefix in SourceCalc.prefixes:
                        filename = filename.replace(prefix, '')
                    segments = [file.strip('\\') for file in filename.split('/')]

                    source_calc = SourceCalc(pid=calc.pid)
                    source_calc.upload = segments[0]
                    source_calc.mainfile = os.path.join(*segments[1:])
                    if with_metadata:
Markus Scheidgen's avatar
Markus Scheidgen committed
332
                        source_calc.metadata = calc.to_calc_with_metadata().__dict__
333
334
335
336
337
338
                    source_calcs.append(source_calc)
                    start_pid = source_calc.pid

                    yield source_calc, total
                except Exception as e:
                    logger.error('could not index', pid=calc.pid, exc_info=e)
339

340
341
342
            if len(source_calcs) == 0:
                break
            else:
343
344
                with utils.timer(logger, 'write index'):
                    SourceCalc.objects.insert(source_calcs)
345
346
347
348


class NomadCOEMigration:
    """
349
    Drives a migration from the NOMAD coe repository db to nomad@FAIRDI.
350
351

    Arguments:
352
353
        migration_version: The migration version. Only packages/calculations with
            no migration version or a lower migration version are migrated.
Markus Scheidgen's avatar
Markus Scheidgen committed
354
355
        threads: Number of threads to run migration in parallel.
        quiet: Prints stats if not quiet
356
    """
357

358
359
360
361
362
363
364
365
366
367
368
    default_sites = [
        '/nomad/repository/data/uploads',
        '/nomad/repository/data/extracted',
        '/data/nomad/uploaded/',
        '/data/nomad/extracted/']

    default_pid_prefix = int(1e7)

    archive_filename = 'archive.tar.gz'
    """ The standard name for tarred uploads in the CoE repository. """

Markus Scheidgen's avatar
Markus Scheidgen committed
369
    def __init__(self, migration_version: int = 0, threads: int = 1, quiet: bool = False) -> None:
370
371
        self.logger = utils.get_logger(__name__, migration_version=migration_version)
        self.migration_version = migration_version
372
        self._client = None
Markus Scheidgen's avatar
Markus Scheidgen committed
373
374
        self._threads = threads
        self._quiet = quiet
375

376
377
378
379
380
381
382
383
384
        self.source = infrastructure.repository_db

    @property
    def client(self):
        if self._client is None:
            from nomad.client import create_client
            self._client = create_client()

        return self._client
385

386
387
    def copy_users(self):
        """ Copy all users. """
388
        for source_user in self.source.query(User).all():
389
390
391
392
393
394
395
396
397
398
            if source_user.user_id <= 2:
                # skip first two users to keep example users
                # they probably are either already the example users, or [root, Evgeny]
                continue

            create_user_payload = dict(
                user_id=source_user.user_id,
                email=source_user.email,
                first_name=source_user.first_name,
                last_name=source_user.last_name,
399
400
                password=source_user.password,
                created=source_user.created
401
402
403
404
405
406
407
408
409
410
411
412
413
414
            )

            try:
                create_user_payload.update(token=source_user.token)
            except LoginException:
                pass

            if source_user.affiliation is not None:
                create_user_payload.update(affiliation=dict(
                    name=source_user.affiliation.name,
                    address=source_user.affiliation.address))

            try:
                self.client.auth.create_user(payload=create_user_payload).response()
415
                self.logger.info('copied user', user_id=source_user.user_id)
416
417
            except HTTPBadRequest as e:
                self.logger.error('could not create user due to bad data', exc_info=e, user_id=source_user.user_id)
418

Markus Scheidgen's avatar
Markus Scheidgen committed
419
420
421
422
423
424
425
426
427
    def _to_comparable_list(self, list):
        for item in list:
            if isinstance(item, dict):
                for key in item.keys():
                    if key.endswith('id'):
                        yield item.get(key)
            else:
                yield item

Markus Scheidgen's avatar
Markus Scheidgen committed
428
429
430
431
432
433
434
    expected_differences = {
        '0d': 'molecule / cluster',
        '3d': 'bulk',
        '2d': '2d / surface',
        '+u': 'gga'
    }

435
    def _validate(self, repo_calc: dict, source_calc: CalcWithMetadata, logger) -> bool:
436
437
438
439
440
441
442
        """
        Validates the given processed calculation, assuming that the data in the given
        source_calc is correct.

        Returns:
            False, if the calculation differs from the source calc.
        """
443
444
445
        keys_to_validate = [
            'atoms', 'basis_set', 'xc_functional', 'system', 'crystal_system',
            'spacegroup', 'code_name', 'code_version']
446
447

        is_valid = True
Markus Scheidgen's avatar
Markus Scheidgen committed
448
        for key, target_value in repo_calc.items():
449
            if key not in keys_to_validate:
450
451
                continue

Markus Scheidgen's avatar
Markus Scheidgen committed
452
            source_value = getattr(source_calc, key, None)
453

454
455
            def check_mismatch() -> bool:
                # some exceptions
Markus Scheidgen's avatar
Markus Scheidgen committed
456
457
                if source_value in NomadCOEMigration.expected_differences and \
                        target_value == NomadCOEMigration.expected_differences.get(source_value):
458
459
                    return True

460
461
                logger.info(
                    'source target missmatch', quantity=key,
Markus Scheidgen's avatar
Markus Scheidgen committed
462
463
                    source_value=source_value, target_value=target_value,
                    value_diff='%s->%s' % (str(source_value), str(target_value)))
464
                return False
465

466
            if source_value is None and target_value is not None:
467
468
                continue

469
470
471
            if target_value is None and source_value is not None:
                is_valid &= check_mismatch()

472
            if isinstance(target_value, list):
Markus Scheidgen's avatar
Markus Scheidgen committed
473
474
475
                source_list = list(self._to_comparable_list(source_value))
                target_list = list(self._to_comparable_list(target_value))
                if len(set(source_list).intersection(target_list)) != len(target_list):
476
                    is_valid &= check_mismatch()
477
478
479
480
481
482
483
                continue

            if isinstance(source_value, str):
                source_value = source_value.lower()
                target_value = str(target_value).lower()

            if source_value != target_value:
484
                is_valid &= check_mismatch()
485
486
487

        return is_valid

488
    def _packages(
489
            self, source_upload_path: str,
Markus Scheidgen's avatar
Markus Scheidgen committed
490
            create: bool = False) -> Tuple[Any, str]:
491
492
493
494
495
496
497
498
499
        """
        Creates a iterator over packages for the given upload path. Packages are
        taken from the :class:`Package` index.

        Arguments:
            source_upload_path: The path to the extracted upload.
            create: If True, will index packages if they not exist.

        Returns:
Markus Scheidgen's avatar
Markus Scheidgen committed
500
            A tuple with the package query and the source_upload_id (last path segment)
501
        """
502

503
504
505
        source_upload_id = os.path.basename(source_upload_path)
        logger = self.logger.bind(
            source_upload_path=source_upload_path, source_upload_id=source_upload_id)
506
507
508

        if os.path.isfile(source_upload_path):
            # assume its a path to an archive files
Markus Scheidgen's avatar
Markus Scheidgen committed
509
            raise ValueError('currently no support for migrating archive files')
Markus Scheidgen's avatar
Markus Scheidgen committed
510
        if not os.path.exists(source_upload_path):
Markus Scheidgen's avatar
Markus Scheidgen committed
511
            raise ValueError('directory %s does not exist' % source_upload_path)
512
513
514
515
516
517
518
519
520

        package_query = Package.objects(upload_id=source_upload_id)

        if package_query.count() == 0:
            if create:
                Package.index(source_upload_path)
                package_query = Package.objects(upload_id=source_upload_id)
                if package_query.count() == 0:
                    logger.error('no package exists, even after indexing')
Markus Scheidgen's avatar
Markus Scheidgen committed
521
                    return package_query, source_upload_id
522
523
            else:
                logger.error('no package exists for upload')
Markus Scheidgen's avatar
Markus Scheidgen committed
524
                return package_query, source_upload_id
525
526

        logger.debug('identified packages for source upload', n_packages=package_query.count())
527
        return package_query, source_upload_id
528

529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
    def _surrogate_metadata(self, source: CalcWithMetadata):
        """
        Compute metadata from the given metadata that can be used for new calcs of the
        same upload.
        """
        return CalcWithMetadata(
            uploader=source.uploader,
            with_embargo=source.with_embargo,
            upload_time=source.upload_time,
            coauthors=source.coauthors,
            shared_with=source.shared_with,
            comment=source.comment,
            references=source.references,
            datasets=source.datasets)

544
545
546
547
548
549
550
551
    def set_pid_prefix(self, prefix: int = default_pid_prefix):
        """
        Sets the repo db pid counter to the given values. Allows to create new calcs
        without interfering with migration calcs with already existing PIDs.
        """
        self.logger.info('set pid prefix', pid_prefix=prefix)
        self.client.admin.exec_pidprefix_command(payload=dict(prefix=prefix)).response()

Markus Scheidgen's avatar
Markus Scheidgen committed
552
    def migrate(
553
554
            self, *args, create_packages: bool = True, local: bool = False,
            delete_local: bool = False) -> utils.POPO:
555
556
557
        """
        Migrate the given uploads.

558
559
560
561
562
        It takes paths to extracted uploads as arguments.

        Requires :class:`Package` instances for the given upload paths. Those will
        be created, if they do not already exists. The packages determine the uploads
        for the target infrastructure.
563
564
565
566
567

        Requires a build :func:`index` to look for existing data in the source db. This
        will be used to add user (and other, PID, ...) metadata and validate calculations.

        Uses PIDs of identified old calculations. Will create new PIDs for previously
568
        unknown uploads. See :func:`set_pid_prefix` on how to avoid conflicts.
569

570
        Arguments:
571
572
573
            upload_path: A filepath to the upload directory.
            create_packages: If True, will create non existing packages.
                Will skip with errors otherwise.
574
575
576
            local: Instead of streaming an upload, create a local file and use
                local_path on the upload.
            delete_local: Delete created local file upload files
577

Markus Scheidgen's avatar
Markus Scheidgen committed
578
        Returns: Dictionary with statistics on the migration.
579
        """
580

Markus Scheidgen's avatar
Markus Scheidgen committed
581
582
583
584
585
586
587
588
        cv = threading.Condition()
        overall_report = Report()
        upload_reports: Dict[str, Report] = {}
        threads = []

        def migrate_package(package: Package, of_packages: int):
            logger = self.logger.bind(package_id=package.package_id, source_upload_id=package.upload_id)
            try:
589
                package_report = self.migrate_package(package, local=local, delete_local=delete_local)
Markus Scheidgen's avatar
Markus Scheidgen committed
590
            except Exception as e:
591
                package_report = Report()
Markus Scheidgen's avatar
Markus Scheidgen committed
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
                logger.error(
                    'unexpected exception while migrating packages', exc_info=e)

            with cv:
                try:
                    overall_report.add(package_report)

                    upload_report = upload_reports[package.upload_id]
                    upload_report.add(package_report)

                    if upload_report.total_packages == of_packages:
                        missing_calcs = SourceCalc.objects(
                            upload=package.upload_id, migration_version__ne=self.migration_version).count()
                        total_source_calcs = SourceCalc.objects(upload=package.upload_id).count()

                        overall_report.missing_calcs -= upload_report.missing_calcs
                        overall_report.missing_calcs += missing_calcs
                        overall_report.total_source_calcs -= upload_report.total_source_calcs
                        overall_report.total_source_calcs += total_source_calcs

                        upload_report.missing_calcs = missing_calcs
                        upload_report.total_source_calcs = total_source_calcs

                        logger.info('migrated upload', **upload_report)

                    if not self._quiet:
                        print(
                            'packages: %d, source calcs: %d, migrated: %d, failed: %d, missing: %d' % (
                                overall_report.total_packages, overall_report.total_source_calcs,
                                overall_report.migrated_calcs, overall_report.failed_calcs,
                                overall_report.missing_calcs))
                except Exception as e:
                    logger.error('unexpected exception while migrating packages', exc_info=e)

                self._threads += 1
                cv.notify()

        for arg in args:
            package_query, source_upload_id = self._packages(arg, create=create_packages)
            number_of_packages = package_query.count()
            for package in package_query:
                with cv:
                    if source_upload_id not in upload_reports:
                        upload_reports[source_upload_id] = Report()

                    if package.migration_version is not None and package.migration_version >= self.migration_version:
                        self.logger.info(
                            'package already migrated, skip it',
                            package_id=package.package_id, source_upload_id=package.upload_id)
                        upload_report = upload_reports[package.upload_id]
                        upload_report.add(package.report)
                        upload_report.skipped_packages += 1
                        overall_report.add(package.report)
                        overall_report.skipped_packages += 1

                        continue
648

Markus Scheidgen's avatar
Markus Scheidgen committed
649
650
651
652
653
                    cv.wait_for(lambda: self._threads > 0)
                    self._threads -= 1
                    thread = threading.Thread(target=lambda: migrate_package(package, number_of_packages))
                    threads.append(thread)
                    thread.start()
654

Markus Scheidgen's avatar
Markus Scheidgen committed
655
656
        for thread in threads:
            thread.join()
657

Markus Scheidgen's avatar
Markus Scheidgen committed
658
        return overall_report
659

Markus Scheidgen's avatar
Markus Scheidgen committed
660
661
    _client_lock = threading.Lock()

662
663
664
    def nomad(self, operation: str, *args, **kwargs) -> Any:
        """
        Calls nomad via the bravado client. It deals with a very busy nomad and catches,
Markus Scheidgen's avatar
Markus Scheidgen committed
665
666
        backsoff, and retries on gateway timouts. It also circumvents bravados/jsonschemas
        thread safety issues using a global lock on client usage.
667
668
669
670
671
672
673
674
675
676
677
678
679

        Arguments:
            operation: Comma separated string of api, endpoint, operation,
                e.g. 'uploads.get_upload'.
        """
        op_path = operation.split('.')
        op = self.client
        for op_path_segment in op_path:
            op = getattr(op, op_path_segment)

        sleep = utils.SleepTimeBackoff()
        while True:
            try:
Markus Scheidgen's avatar
Markus Scheidgen committed
680
                NomadCOEMigration._client_lock.acquire(blocking=True)
681
682
683
684
685
                return op(*args, **kwargs).response().result
            except HTTPGatewayTimeout:
                sleep()
            except Exception as e:
                raise e
Markus Scheidgen's avatar
Markus Scheidgen committed
686
687
            finally:
                NomadCOEMigration._client_lock.release()
688

689
690
691
692
693
    def migrate_package(
            self, package: Package, local: bool = False,
            delete_local: bool = False) -> 'Report':
        """ Migrates the given package. For other params see :func:`migrate`. """

694
695
696
697
698
699
        source_upload_id = package.upload_id
        package_id = package.package_id

        logger = self.logger.bind(package_id=package_id, source_upload_id=source_upload_id)
        logger.debug('start to process package')

700
701
        report = Report()
        report.total_packages += 1
702
703
704

        # upload and process the upload file
        from nomad.client import stream_upload_with_client
705
        created_tmp_package_upload_file = None
706
707
708
        with utils.timer(logger, 'upload completed'):
            try:
                if local:
709
710
711
                    upload_filepath, created = package.create_package_upload_file()
                    if created:
                        created_tmp_package_upload_file = upload_filepath
712
                    self.logger.debug('created package upload file')
713
714
                    upload = self.nomad(
                        'uploads.upload', name=package_id, local_path=upload_filepath)
715
716
717
718
719
720
                else:
                    upload_f = package.open_package_upload_file()
                    self.logger.debug('opened package upload file')
                    upload = stream_upload_with_client(self.client, upload_f, name=package_id)
            except Exception as e:
                self.logger.error('could not upload package', exc_info=e)
721
722
                report.failed_packages += 1
                return report
723
724
725
726

        logger = logger.bind(
            source_upload_id=source_upload_id, upload_id=upload.upload_id)

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
        # grab source calcs, while waiting for upload
        source_calcs = dict()
        surrogate_source_calc_with_metadata = None
        with utils.timer(logger, 'loaded source metadata'):
            for filenames_chunk in utils.chunks(package.filenames, 1000):
                for source_calc in SourceCalc.objects(
                        upload=source_upload_id, mainfile__in=filenames_chunk):

                    source_calc_with_metadata = CalcWithMetadata(**source_calc.metadata)
                    source_calc_with_metadata.pid = source_calc.pid
                    source_calc_with_metadata.mainfile = source_calc.mainfile
                    source_calcs[source_calc.mainfile] = (source_calc, source_calc_with_metadata)

                    # establish a surrogate for new calcs
                    if surrogate_source_calc_with_metadata is None:
                        surrogate_source_calc_with_metadata = \
                            self._surrogate_metadata(source_calc_with_metadata)
            report.total_source_calcs = len(source_calcs)
        # try to find a surrogate outside the package, if necessary
        if surrogate_source_calc_with_metadata is None:
            source_calc = SourceCalc.objects(upload=source_upload_id).first()
            if source_calc is not None:
                source_calc_with_metadata = CalcWithMetadata(**source_calc.metadata)
                surrogate_source_calc_with_metadata = \
                    self._surrogate_metadata(source_calc_with_metadata)

753
754
755
756
        # wait for complete upload
        with utils.timer(logger, 'upload processing completed'):
            sleep = utils.SleepTimeBackoff()
            while upload.tasks_running:
757
                upload = self.nomad('uploads.get_upload', upload_id=upload.upload_id)
758
759
760
761
                sleep()

        if upload.tasks_status == FAILURE:
            logger.error('failed to process upload', process_errors=upload.errors)
762
763
764
            report.failed_packages += 1
            report.missing_calcs += report.total_source_calcs
            return report
765
766
767
768
769
770
771
772
        else:
            report.total_calcs += upload.calcs.pagination.total

        calc_mainfiles = []
        upload_total_calcs = upload.calcs.pagination.total

        # check for processing errors
        with utils.timer(logger, 'checked upload processing'):
773
            per_page = 500
774
            for page in range(1, math.ceil(upload_total_calcs / per_page) + 1):
775
776
777
                upload = self.nomad(
                    'uploads.get_upload', upload_id=upload.upload_id, per_page=per_page,
                    page=page, order_by='mainfile')
778
779
780
781
782
783
784
785

                for calc_proc in upload.calcs.results:
                    calc_logger = logger.bind(
                        calc_id=calc_proc.calc_id,
                        mainfile=calc_proc.mainfile)

                    if calc_proc.tasks_status == SUCCESS:
                        calc_mainfiles.append(calc_proc.mainfile)
786
                    else:
787
                        report.failed_calcs += 1
788
                        calc_logger.info(
789
790
                            'could not process a calc', process_errors=calc_proc.errors)
                        continue
791

792
793
794
795
        # verify upload against source
        calcs_in_search = 0
        with utils.timer(logger, 'varyfied upload against source calcs'):
            for page in range(1, math.ceil(upload_total_calcs / per_page) + 1):
796
797
798
799
                search = self.nomad(
                    'repo.search', page=page, per_page=per_page, upload_id=upload.upload_id,
                    order_by='mainfile')

800
801
802
803
804
805
806
807
808
                for calc in search.results:
                    calcs_in_search += 1
                    source_calc, source_calc_with_metadata = source_calcs.get(
                        calc['mainfile'], (None, None))

                    if source_calc is not None:
                        report.migrated_calcs += 1

                        calc_logger = logger.bind(calc_id=calc['calc_id'], mainfile=calc['mainfile'])
809
810
811
812
                        try:
                            if not self._validate(calc, source_calc_with_metadata, calc_logger):
                                report.calcs_with_diffs += 1
                        except Exception as e:
Markus Scheidgen's avatar
Markus Scheidgen committed
813
                            calc_logger.warning('unexpected exception during validation', exc_info=e)
814
815
816
817
                            report.calcs_with_diffs += 1
                    else:
                        calc_logger.info('processed a calc that has no source')
                        report.new_calcs += 1
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
                        # guessing the metadata from other calcs in upload/package
                        if surrogate_source_calc_with_metadata is not None:
                            new_calc_with_metadata = CalcWithMetadata(**surrogate_source_calc_with_metadata.to_dict())
                            new_calc_with_metadata.mainfile = calc['mainfile']
                        else:
                            calc_logger.warning('could not determine any metadata for new calc')
                            create_time_epoch = os.path.getctime(package.upload_path)
                            new_calc_with_metadata = CalcWithMetadata(
                                upload_time=datetime.datetime.fromtimestamp(create_time_epoch),
                                with_embargo=package.restricted > 0,
                                comment=default_comment,
                                uploader=default_uploader,
                                mainfile=calc['mainfile'])
                            surrogate_source_calc_with_metadata = new_calc_with_metadata

                        source_calcs[calc['mainfile']] = (None, new_calc_with_metadata)
834

835
836
            if len(calc_mainfiles) != calcs_in_search:
                logger.error('missmatch between processed calcs and calcs found with search')
Markus Scheidgen's avatar
Markus Scheidgen committed
837

838
839
840
841
842
843
844
        # publish upload
        if len(calc_mainfiles) > 0:
            with utils.timer(logger, 'upload published'):
                upload_metadata = dict(with_embargo=(package.restricted > 0))
                upload_metadata['calculations'] = [
                    self._to_api_metadata(source_calc_with_metadata)
                    for _, source_calc_with_metadata in source_calcs.values()]
845

846
847
848
                upload = self.nomad(
                    'uploads.exec_upload_operation', upload_id=upload.upload_id,
                    payload=dict(operation='publish', metadata=upload_metadata))
849

850
851
852
                sleep = utils.SleepTimeBackoff()
                while upload.process_running:
                    try:
853
                        upload = self.nomad('uploads.get_upload', upload_id=upload.upload_id)
854
855
856
857
858
859
860
861
862
863
864
                        sleep()
                    except HTTPNotFound:
                        # the proc upload will be deleted by the publish operation
                        break

                if upload.tasks_status == FAILURE:
                    logger.error('could not publish upload', process_errors=upload.errors)
                    report.failed_calcs = report.total_calcs
                    report.migrated_calcs = 0
                    report.calcs_with_diffs = 0
                    report.new_calcs = 0
865
                    report.failed_packages += 1
866
867
868
869
870
871
                else:
                    SourceCalc.objects(upload=source_upload_id, mainfile__in=calc_mainfiles) \
                        .update(migration_version=self.migration_version)
                    package.migration_version = self.migration_version
        else:
            logger.info('no successful calcs, skip publish')
872

873
        report.missing_calcs = report.total_source_calcs - report.migrated_calcs
Markus Scheidgen's avatar
Markus Scheidgen committed
874
875
876
        package.report = report
        package.save()

877
        logger.info('migrated package', **report)
878

879
880
881
882
883
884
        if created_tmp_package_upload_file is not None and delete_local:
            try:
                os.remove(created_tmp_package_upload_file)
            except Exception as e:
                logger.error('could not remove tmp package upload file', exc_info=e)

885
        return report
886
887

    def _to_api_metadata(self, calc_with_metadata: CalcWithMetadata) -> dict:
Markus Scheidgen's avatar
Markus Scheidgen committed
888
        """ Transforms to a dict that fullfils the API's uploade metadata model. """
889

Markus Scheidgen's avatar
Markus Scheidgen committed
890
        return dict(
891
892
893
894
            _upload_time=calc_with_metadata.upload_time,
            _uploader=calc_with_metadata.uploader['id'],
            _pid=calc_with_metadata.pid,
            references=[ref['value'] for ref in calc_with_metadata.references],
Markus Scheidgen's avatar
Markus Scheidgen committed
895
896
897
            datasets=[dict(
                id=ds['id'],
                _doi=ds.get('doi', {'value': None})['value'],
898
899
900
901
902
903
                _name=ds.get('name', None)) for ds in calc_with_metadata.datasets],
            mainfile=calc_with_metadata.mainfile,
            with_embargo=calc_with_metadata.with_embargo,
            comment=calc_with_metadata.comment,
            coauthors=list(int(user['id']) for user in calc_with_metadata.coauthors),
            shared_with=list(int(user['id']) for user in calc_with_metadata.shared_with)
Markus Scheidgen's avatar
Markus Scheidgen committed
904
905
        )

906
    def index(self, *args, **kwargs):
907
        """ see :func:`SourceCalc.index` """
908
        return SourceCalc.index(self.source, *args, **kwargs)
909
910
911
912

    def package(self, *args, **kwargs):
        """ see :func:`Package.add` """
        return Package.index(*args, **kwargs)
913
914
915


class Report(utils.POPO):
Markus Scheidgen's avatar
Markus Scheidgen committed
916
    def __init__(self, *args, **kwargs):
917
918
        self.total_packages = 0
        self.failed_packages = 0
Markus Scheidgen's avatar
Markus Scheidgen committed
919
        self.skipped_packages = 0
920
921
922
923
924
925
926
927
        self.total_calcs = 0  # the calcs that have been found by the target
        self.total_source_calcs = 0  # the calcs in the source index
        self.failed_calcs = 0  # the calcs found b the target that could not be processed/published
        self.migrated_calcs = 0   # the calcs from the source, successfully added to the target
        self.calcs_with_diffs = 0  # the calcs from the source, successfully added to the target with different metadata
        self.new_calcs = 0  # the calcs successfully added to the target that were not found in the source
        self.missing_calcs = 0  # the calcs in the source, that could not be added to the target due to failure or not founding the calc

Markus Scheidgen's avatar
Markus Scheidgen committed
928
929
        super().__init__(*args, **kwargs)

930
931
932
    def add(self, other: 'Report') -> None:
        for key, value in other.items():
            self[key] = self.get(key, 0) + value