OSDManager.py 22.7 KB
Newer Older
1
2
3
import os
import pickle
import subprocess
Felix Seibert's avatar
Felix Seibert committed
4
5
from urllib import request
import urllib.error
6
7
import shutil

8
9
10
import xtreemfs_client.dataDistribution as dataDistribution
import xtreemfs_client.div_util as div_util
import xtreemfs_client.folder as folder
11
import xtreemfs_client.dir_status_page_parser as dir_status_page_parser
12
13
14
15
16
17
18

'''
xOSDManager - a python module to manage OSD selection in XtreemFS
currently only depth (level) 2 subdirectories can be managed
'''


Felix Seibert's avatar
Felix Seibert committed
19
20
# TODO add support for arbitrary subdirectory level
# (currently depth=2 is hardcoded, which is fine for GeoMultiSens purposes)
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43

class OSDManager(object):
    def __init__(self, path_to_managed_folder, config_file='.das_config', value_map=None, debug=False):

        self.managedFolder = path_to_managed_folder
        self.configFile = config_file
        self.debug = debug

        if value_map is None:

            if not div_util.check_for_executable('xtfsutil'):
                raise ExecutableNotFoundException("No xtfsutil found. Please make sure it is contained in your PATH.")

            output_1 = subprocess.run(["xtfsutil", self.managedFolder], stdout=subprocess.PIPE, stderr=subprocess.PIPE,
                                      universal_newlines=True)

            if output_1.stderr.startswith("xtfsutil failed: Path doesn't point to an entity on an XtreemFS volume!"):
                raise NotAXtreemFSVolume("The specified folder '" + path_to_managed_folder +
                                         "' is not part of an XtreemFS volume!")

            if len(output_1.stderr) > 0:
                raise Exception("xtfsutil produced some error: " + output_1.stderr)

Felix Seibert's avatar
Felix Seibert committed
44
            self.pathOnVolume = div_util.remove_leading_trailing_slashes(
45
46
47
48
49
50
                str(output_1.stdout).split("\n")[0].split()[-1])
            self.pathToMountPoint = self.managedFolder[0:(len(self.managedFolder) - len(self.pathOnVolume) - 1)]

            output_2 = subprocess.run(["xtfsutil", self.pathToMountPoint],
                                      stdout=subprocess.PIPE, universal_newlines=True)

Felix Seibert's avatar
Felix Seibert committed
51
52
53
54
55
            self.volume_information = div_util.extract_volume_information(output_2.stdout)
            self.volumeName = self.volume_information[0]
            osd_list = list(map(lambda x: x[0], self.volume_information[1]))
            self.osdSelectionPolicy = self.volume_information[2]
            self.volumeAddress = self.volume_information[3]
56
57
58
59
60

            self.dataDistribution = None
            if not self.__read_configuration():
                self.dataDistribution = dataDistribution.DataDistribution()

Felix Seibert's avatar
Felix Seibert committed
61
62
63
64
65
66
67
68
            self.dataDistribution.add_osd_list(osd_list)

            self.osdInformation = None

            try:
                answer = request.urlopen(div_util.get_http_address(self.volumeAddress))
                html_data = answer.read().decode('UTF-8')

69
                parser = dir_status_page_parser.dir_status_page_parser()
Felix Seibert's avatar
Felix Seibert committed
70
71
72
73
74
75
76
                parser.feed(html_data)

                # filter out data sets without last update time or wrong service type
                filtered_data_sets = list(filter(lambda x: int(x['last updated'].split()[0]) != 0, parser.dataSets))
                filtered_data_sets = list(filter(lambda x: x['type'] == 'SERVICE_TYPE_OSD', filtered_data_sets))

                self.osdInformation = {}
77

Felix Seibert's avatar
Felix Seibert committed
78
79
80
81
82
83
84
85
86
87
88
89
                for data_set in filtered_data_sets:
                    uuid = data_set['uuid']
                    current_osd = {}
                    current_osd['usable_space'] = int(data_set['usable'].split()[0])
                    current_osd['total_space'] = int(data_set['total'].split()[0])

                    self.osdInformation[uuid] = current_osd

            except urllib.error.URLError as error:
                print("osd information could not be fetched! Probably the http status page could not be found at:",
                      div_util.get_http_address(self.volumeAddress))
                print(error)
90
        else:
Felix Seibert's avatar
Felix Seibert committed
91
92
93
94
95
96
97
98
99
100
101
            try:
                self.pathOnVolume = value_map['path_on_volume']
                self.pathToMountPoint = value_map['path_to_mount']
                self.volumeName = value_map['volume_name']
                self.osdSelectionPolicy = value_map['osd_selection_policy']
                self.dataDistribution = value_map['data_distribution']
                self.volumeAddress = value_map['volume_address']
                self.osdInformation = value_map['osd_information']
            except KeyError as error:
                print('key not found:', error)
                print('leaving in OSDManager field empty!')
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122

    def __read_configuration(self):
        assert self.dataDistribution is None
        path_to_config = os.path.join(self.managedFolder, self.configFile)
        try:
            f = open(path_to_config, "rb")
            self.dataDistribution = pickle.load(f)
            return True
        except IOError:
            return False

    def __write_configuration(self):
        path_to_config = os.path.join(self.managedFolder, self.configFile)
        f = open(path_to_config, "wb")
        pickle.dump(self.dataDistribution, f)

    '''
    create empty folders and assign OSDs.
    '''

    def create_empty_folders(self, folders):
Felix Seibert's avatar
Felix Seibert committed
123
        average_size = int(self.dataDistribution.get_average_folder_size())
124
125
126
127
128
129
130
131
132
        if average_size <= 0:
            average_size = 1

        tiles = []

        for input_folder in folders:
            new_tile = folder.Folder(self.path_on_volume(input_folder), average_size, None)
            tiles.append(new_tile)

Felix Seibert's avatar
Felix Seibert committed
133
        new_tiles = self.dataDistribution.add_folders(tiles)
134
135
136
137
138
139
140
141
142
143
144
145
146

        self.apply_osd_assignments(new_tiles)

        for input_folder in folders:
            os.makedirs(input_folder, exist_ok=True)

        self.__write_configuration()

    '''
    copy a list of given folders into the managed folder, assigning OSDs to new folders and updating
    self.dataDistribution
    '''

Felix Seibert's avatar
Felix Seibert committed
147
148
    def copy_folders(self, folders, environment='LOCAL', remote_source=None, sshfs_mount_dir='/tmp/sshfs_tmp_mnt',
                     apply_layout=True):
149
150
151
152
153
154
155
156
157
158
159
160
        if not div_util.check_for_executable('du'):
            raise ExecutableNotFoundException("No du found. Please make sure it is contained in your PATH.")

        if remote_source is not None:
            if not div_util.check_for_executable('sshfs'):
                raise ExecutableNotFoundException("No sshfs found. Please make sure it is contained in your PATH.")
            if not div_util.check_for_executable('scp'):
                raise ExecutableNotFoundException("No scp found. Please make sure it is contained in your PATH.")
            if not div_util.check_for_executable('fusermount'):
                raise ExecutableNotFoundException("No fusermount found. Please make sure it is contained in your PATH.")

        if remote_source is not None:
Felix Seibert's avatar
Felix Seibert committed
161
            os.makedirs(sshfs_mount_dir, exist_ok=True)
162
163
164
165
166
167
168
169

        new_folders = []

        for input_folder in folders:
            last_2_path_elements = os.path.join(os.path.split(os.path.split(input_folder)[0])[1],
                                                os.path.split(input_folder)[1])
            if remote_source is not None:
                mount_point = os.path.join(sshfs_mount_dir, last_2_path_elements)
Felix Seibert's avatar
Felix Seibert committed
170
                os.makedirs(mount_point, exist_ok=True)
171
172
173
174
                subprocess.run(["sshfs", remote_source + ":" + input_folder, mount_point])
                du = subprocess.run(["du", "-s", mount_point], stdout=subprocess.PIPE,
                                    universal_newlines=True)
                subprocess.run(["fusermount", "-uz", mount_point])
Felix Seibert's avatar
Felix Seibert committed
175
                shutil.rmtree(mount_point)
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
                folder_size = int(du.stdout.split()[0])
            else:
                du = subprocess.run(["du", "-s", input_folder], stdout=subprocess.PIPE,
                                    universal_newlines=True)
                folder_size = int(du.stdout.split()[0])

            new_folder = folder.Folder(os.path.join(self.volumeName, self.pathOnVolume, last_2_path_elements),
                                       folder_size,
                                       input_folder)
            if self.debug:
                print("new folder: " + str(new_folder))

            new_folders.append(new_folder)

        if remote_source is not None:
            shutil.rmtree(sshfs_mount_dir)

Felix Seibert's avatar
Felix Seibert committed
193
194
195
196
197
        new_assignments = self.dataDistribution.add_folders(new_folders)
        if apply_layout:
            self.apply_osd_assignments(new_assignments)
        elif self.debug:
            print("NOT applying data layout!")
198
199
200
201
202
203
204
205
206

        self.__write_configuration()

        if self.debug:
            print("osd manager after new folders have been added to data distribution:")
            print(str(self))

        self.__copy_data(new_folders, environment, remote_source)

Felix Seibert's avatar
Felix Seibert committed
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
    '''
    moves a folder from one OSD to another OSD. you may specify a temporary folder.
    '''

    def move_folder_to_osd(self, folder_id: str, new_osd_id: str, tmp_dir='/tmp/move_folder_tmp_dir'):
        folder_path = os.path.join(self.get_target_dir(folder_id),
                                   os.path.split(folder_id)[1])

        os.makedirs(tmp_dir)

        if not div_util.check_for_executable('xtfsutil'):
            raise ExecutableNotFoundException("No xtfsutil found. Please make sure it is contained in your PATH.")
        # step 1: add folder to new OSD, update data distribution and xtreemfs configuration
        self.dataDistribution.assign_new_osd(folder_id, new_osd_id)
        if self.debug:
            subprocess.run(["xtfsutil",
                            "--set-pattr", "1004.filenamePrefix", "--value",
                            "add " + folder_id + " " + new_osd_id + "", self.pathToMountPoint])
        else:
            subprocess.run(["xtfsutil",
                            "--set-pattr", "1004.filenamePrefix", "--value",
                            "add " + folder_id + " " + new_osd_id + "", self.pathToMountPoint],
                           stdout=subprocess.PIPE, universal_newlines=True)

        # step 2: one by one, move files to tmp_location and then back to the folder, which means that they should now
        # be located onto the new OSD.

        for root, dirs, files in os.walk(folder_path):
            for file in files:
                current_file_path = os.path.join(root, file)
                copied_file_path = os.path.join(tmp_dir, file)
                shutil.move(current_file_path, copied_file_path)
                shutil.move(copied_file_path, current_file_path)

        shutil.rmtree(tmp_dir, ignore_errors=True)

    '''
    removes a folder from the distribution. this does NOT delete the folder from the file system.
    '''

    def remove_folder(self, folder_id):
        containing_osd = self.dataDistribution.get_containing_osd(folder_id)
        if containing_osd is not None:
            if not div_util.check_for_executable('xtfsutil'):
                raise ExecutableNotFoundException("No xtfsutil found. Please make sure it is contained in your PATH.")

            containing_osd.remove_folder(folder_id)
            if self.debug:
                subprocess.run(["xtfsutil",
                                "--set-pattr", "1004.filenamePrefix", "--value",
                                "remove " + folder_id + "", self.pathToMountPoint])
            else:
                subprocess.run(["xtfsutil",
                                "--set-pattr", "1004.filenamePrefix", "--value",
                                "remove " + folder_id + "", self.pathToMountPoint],
                               stdout=subprocess.PIPE, universal_newlines=True)

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
    '''
    update the given (by absolute path) folders, such that the values held by self.dataDistribution
    matches their size on disk.
    if no argument is given, all folders are updated.
    '''

    def update(self, arg_folders=None):

        if arg_folders is not None:
            for folder_for_update in arg_folders:
                if not folder_for_update.startswith(self.managedFolder):
                    raise PathNotManagedException(
                        "The path :" + folder_for_update + "is not managed by this instance of the XtreemFS OSD"
                                                           "manager!")

        folder_size_updates = {}

        folders = arg_folders
        if arg_folders is None:
            folders = self.get_assigned_folders()

        for folder_for_update in folders:
            folder_id = self.path_on_volume(folder_for_update)
            command = ["du", "-s", folder_for_update]
            if self.debug:
                print("executing: " + str(command))
            du = subprocess.run(command, stdout=subprocess.PIPE, universal_newlines=True)
            folder_disk_size = int(du.stdout.split()[0])
            folder_size_updates[folder_id] = folder_disk_size

        for folder_for_update, size in folder_size_updates.items():
            self.dataDistribution.update_folder(folder_for_update, size)

        self.__write_configuration()

        if self.debug:
            print(str(self))

    '''
    apply the given assignments to the XtreemFS volume, using xtfsutil.
    the assignments are given as a list containing tuples (tile_id, osd),
    where tile_id is given by applying path_on_volume() onto the absolute path of the folder.
    '''

    def apply_osd_assignments(self, assignments):
        if not div_util.check_for_executable('xtfsutil'):
            raise ExecutableNotFoundException("No xtfsutil found. Please make sure it is contained in your PATH.")

        if self.osdSelectionPolicy is not "1000,1004":
            if self.debug:
                subprocess.run(["xtfsutil", "--set-osp", "prefix", self.pathToMountPoint])
            else:
                subprocess.run(["xtfsutil", "--set-osp", "prefix", self.pathToMountPoint],
                               stdout=subprocess.PIPE, universal_newlines=True)

        for new_tile in assignments:
            if self.debug:
                subprocess.run(["xtfsutil",
                                "--set-pattr", "1004.filenamePrefix", "--value",
                                "add " + new_tile[0] + " " + new_tile[1] + "",
                                self.pathToMountPoint])
            else:
                subprocess.run(["xtfsutil",
                                "--set-pattr", "1004.filenamePrefix", "--value",
                                "add " + new_tile[0] + " " + new_tile[1] + "",
                                self.pathToMountPoint],
                               stdout=subprocess.PIPE, universal_newlines=True)

    '''
    copy data onto XtreemFS volume
    '''

    def __copy_data(self, input_folders, environment, remote_source):
Felix Seibert's avatar
Felix Seibert committed
337
338
        if self.debug:
            print('calling copy_data with: ', input_folders, environment, remote_source)
339
340
341
342
343
        osd_list = self.dataDistribution.get_osd_list()
        osd_to_folders_map = {}
        for osd in osd_list:
            osd_to_folders_map[osd] = []
        for input_folder in input_folders:
Felix Seibert's avatar
Felix Seibert committed
344
            osd_for_tile = self.dataDistribution.get_containing_osd(input_folder.id).uuid
345
346
347
348
349
350
351
352
353
354
355
            osd_to_folders_map[osd_for_tile].append(input_folder)

        if self.debug:
            print("osd to folders map:")
            for key, value in osd_to_folders_map.items():
                print("osd: " + key)
                print("assigned folders:")
                for input_folder in value:
                    print(str(input_folder))

        # trigger the copying!
Felix Seibert's avatar
Felix Seibert committed
356
        if environment == "SLURM":
357
358
            assert remote_source is not None
            copy_commands = self.__generate_copy_commands_slurm(osd_to_folders_map, remote_source)
Felix Seibert's avatar
Felix Seibert committed
359
        elif environment == "HU_CLUSTER":
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
            copy_commands = self.__generate_copy_commands_hu_cluster(osd_to_folders_map)
        else:
            copy_commands = self.__generate_copy_commands_local(osd_to_folders_map)

        for input_folder in input_folders:
            path_to_target_dir = self.get_target_dir(input_folder.id)
            os.makedirs(path_to_target_dir, exist_ok=True)

        self.__execute_copy_commands(copy_commands)

    '''
    generates a list of copy commands, one command for each OSD that receives new data.
    the copy commands are constructed such that they can be executed in a slurm environment (that is, within a slurm job
    allocation) at ZIB.
    each command is a copy command including all new folders for the corresponding OSD, preceded by a srun command
    to execute the copy command locally on the slurm node on which the target OSD resides.
    '''

    def __generate_copy_commands_slurm(self, osd_to_folders_map, remote_source):
        if self.debug:
            print("Using SLURM mode for copying...")

Felix Seibert's avatar
Felix Seibert committed
382
383
384
385
386
387
388
389
390
        slurm_hosts = div_util.get_slurm_hosts()

        if self.debug:
            print('slurm_hosts: ', slurm_hosts)

        osd_to_host_map = div_util.get_osd_to_hostname_map(self.volume_information[1], slurm_hosts)

        if self.debug:
            print('osd_to_host_map: ', osd_to_host_map)
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

        command_list = []

        for key in osd_to_folders_map.keys():
            host_name = osd_to_host_map[key]
            command = ""
            for copy_folder in osd_to_folders_map[key]:
                command += "srun -N1-1 --nodelist=" + host_name
                command += " scp -rq " + remote_source + ":" + copy_folder.origin
                command += " " + self.get_target_dir(copy_folder.id)
                command += " ;"
            if len(osd_to_folders_map[key]) > 0:
                command_list.append(command)

        return command_list

    '''
    generates a list of copy commands, one command for each OSD that receives new data.
    the copy commands are constructed such that they can be executed on the GeoMultiSens cluster at HU Berlin.
    each command is a copy command including all new folders for the corresponding OSD, preceded by a ssh command
    to execute the copy command locally on the node of the target OSD.
    '''

    def __generate_copy_commands_hu_cluster(self, osd_to_folders_map):
        if self.debug:
            print("Using HU_CLUSTER mode for copying...")

        if not div_util.check_for_executable('xtfsutil'):
            raise ExecutableNotFoundException("No xtfsutil found. Please make sure it is contained in your PATH.")

        xtfsutil = subprocess.run(["xtfsutil", self.pathToMountPoint],
                                  stdout=subprocess.PIPE, universal_newlines=True)
Felix Seibert's avatar
Felix Seibert committed
423
        volume_information = div_util.extract_volume_information(xtfsutil.stdout)
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
        osd_to_ip_address = {}
        for (osd, ip) in volume_information[1]:
            osd_to_ip_address[osd] = ip

        command_list = []

        for key in osd_to_folders_map.keys():
            ip_address = osd_to_ip_address[key]
            command = "ssh " + ip_address + " \'"
            for copy_folder in osd_to_folders_map[key]:
                command += " cp -r"
                command += " " + copy_folder.origin
                command += " " + self.get_target_dir(copy_folder.id)
                command += " ;"
            command += " \' "
            if len(osd_to_folders_map[key]) > 0:
                command_list.append(command)

        return command_list

    '''
    generates a list of copy commands, one command for each OSD that receives new data.
    plain old cp is used for the actual copying.
    '''

    def __generate_copy_commands_local(self, osd_to_folders_map):
        if self.debug:
            print("Using local cp for copying...")

        command_list = []

        for key in osd_to_folders_map.keys():
            command = ""
            for copy_folder in osd_to_folders_map[key]:
                command += "cp -r "
                command += copy_folder.origin
                command += " " + self.get_target_dir(copy_folder.id)
                command += " ; "

            if len(osd_to_folders_map[key]) > 0:
                command_list.append(command)

        return command_list

    '''
    execute, in parallel, a given set of commands. note that the degree of parallelism will match the length of
    command_list.
    '''

    def __execute_copy_commands(self, command_list):
        if self.debug:
            print("Executing copy commands: ")
            for command in command_list:
                print(str(command))
            print("in total " + str(len(command_list)) + " commands.")
        processes = []
        for command in command_list:
Felix Seibert's avatar
Felix Seibert committed
481
            process = subprocess.Popen(command, stderr=subprocess.STDOUT, stdout=subprocess.PIPE, shell=True)
482
483
484
485
486
487
            processes.append(process)

        for p in processes:
            p.wait()

        if self.debug:
Felix Seibert's avatar
Felix Seibert committed
488
489
            for terminated_process in processes:
                print(str(terminated_process.communicate()))
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
            print("Executing copy commands done.")

    '''
    creates a list of all folders, managed by this XtreemFS OSD manager, that have an assigned OSD.
    '''

    def get_assigned_folders(self):
        assigned_folders = []
        for depth_1_folder in os.listdir(self.managedFolder):
            depth_1_path = os.path.join(self.managedFolder, depth_1_folder)
            if os.path.isdir(depth_1_path):
                for depth_2_folder in os.listdir(depth_1_path):
                    depth_2_path = os.path.join(self.managedFolder, depth_1_folder, depth_2_folder)
                    if os.path.isdir(depth_2_path):
                        assigned_folders.append(depth_2_path)

        return assigned_folders

    '''
    gets the path to the target dir, given the folder_id
    '''
Felix Seibert's avatar
Felix Seibert committed
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
    def get_target_dir(self, folder_id):
        path_on_mount_point = folder_id[len(self.volumeName) + 1:]
        return os.path.split(os.path.join(self.pathToMountPoint, path_on_mount_point))[0]

    '''
    remove the leading part of the path, such that only the part onto the xtreemfs volume remains, including
    the volume itself.
    throws an exception when the path is not managed by this XtreemFS OSD manager.
    '''

    def path_on_volume(self, path):
        if not path.startswith(self.managedFolder):
            raise PathNotManagedException("Path " + path + " is not managed by this instance of the XtreemFS OSD"
                                                           "manager!")
        return os.path.join(self.volumeName, path[len(self.pathToMountPoint) + 1:])

    def __str__(self):
        representation = "pathToMountPoint: " + self.pathToMountPoint + " volumeName: " + self.volumeName + " pathOnVolume: " \
                         + self.pathOnVolume
        representation += ("\nconfigFile: " + self.configFile + "\n")
        representation += self.dataDistribution.description()
        return representation


class ExecutableNotFoundException(Exception):
    """raise this when an external executable can not be found"""


class NotAXtreemFSVolume(Exception):
    """raise this when a path does not point to a folder on a xtreemfs volume"""


class PathNotManagedException(Exception):
    """raise this when a path is handled, that is not managed by xOSDManager"""