summaryrefslogtreecommitdiffstats
path: root/qa/tasks/vstart_runner.py
blob: cfbaad7829aa697d29266345b5d5eb9ac3e8cd20 (plain)
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
"""
vstart_runner: override Filesystem and Mount interfaces to run a CephFSTestCase against a vstart
ceph instance instead of a packaged/installed cluster.  Use this to turn around test cases
quickly during development.

Simple usage (assuming teuthology and ceph checked out in ~/git):

    # Activate the teuthology virtualenv
    source ~/git/teuthology/virtualenv/bin/activate
    # Go into your ceph build directory
    cd ~/git/ceph/build
    # Invoke a test using this script
    python ~/git/ceph/qa/tasks/vstart_runner.py --create tasks.cephfs.test_data_scan

Alternative usage:

    # Alternatively, if you use different paths, specify them as follows:
    LD_LIBRARY_PATH=`pwd`/lib PYTHONPATH=~/git/teuthology:~/git/ceph/qa:`pwd`/../src/pybind:`pwd`/lib/cython_modules/lib.2 python ~/git/ceph/qa/tasks/vstart_runner.py

    # If you wish to drop to a python shell on failures, use --interactive:
    python ~/git/ceph/qa/tasks/vstart_runner.py --interactive

    # If you wish to run a named test case, pass it as an argument:
    python ~/git/ceph/qa/tasks/vstart_runner.py tasks.cephfs.test_data_scan

    # Also, you can create the cluster once and then run named test cases against it:
    python ~/git/ceph/qa/tasks/vstart_runner.py --create-cluster-only
    python ~/git/ceph/qa/tasks/vstart_runner.py tasks.mgr.dashboard.test_health
    python ~/git/ceph/qa/tasks/vstart_runner.py tasks.mgr.dashboard.test_rgw

"""

from io import BytesIO
from io import StringIO
from collections import defaultdict
import getpass
import signal
import tempfile
import threading
import datetime
import shutil
import re
import os
import time
import sys
import errno
from unittest import suite, loader
import unittest
import platform
from teuthology import misc
from teuthology.orchestra.run import Raw, quote
from teuthology.orchestra.daemon import DaemonGroup
from teuthology.config import config as teuth_config
import six
import logging
try:
    import urllib3
    urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
except:
    pass

log = logging.getLogger(__name__)

handler = logging.FileHandler("./vstart_runner.log")
formatter = logging.Formatter(
    fmt=u'%(asctime)s.%(msecs)03d %(levelname)s:%(name)s:%(message)s',
    datefmt='%Y-%m-%dT%H:%M:%S')
handler.setFormatter(formatter)
log.addHandler(handler)
log.setLevel(logging.INFO)


def respawn_in_path(lib_path, python_paths):
    execv_cmd = ['python']
    if platform.system() == "Darwin":
        lib_path_var = "DYLD_LIBRARY_PATH"
    else:
        lib_path_var = "LD_LIBRARY_PATH"

    py_binary = os.environ.get("PYTHON", "python")

    if lib_path_var in os.environ:
        if lib_path not in os.environ[lib_path_var]:
            os.environ[lib_path_var] += ':' + lib_path
            os.execvp(py_binary, execv_cmd + sys.argv)
    else:
        os.environ[lib_path_var] = lib_path
        os.execvp(py_binary, execv_cmd + sys.argv)

    for p in python_paths:
        sys.path.insert(0, p)


# Let's use some sensible defaults
if os.path.exists("./CMakeCache.txt") and os.path.exists("./bin"):

    # A list of candidate paths for each package we need
    guesses = [
        ["~/git/teuthology", "~/scm/teuthology", "~/teuthology"],
        ["lib/cython_modules/lib.2"],
        ["../src/pybind"],
    ]

    python_paths = []

    # Up one level so that "tasks.foo.bar" imports work
    python_paths.append(os.path.abspath(
        os.path.join(os.path.dirname(os.path.realpath(__file__)), "..")
    ))

    for package_guesses in guesses:
        for g in package_guesses:
            g_exp = os.path.abspath(os.path.expanduser(g))
            if os.path.exists(g_exp):
                python_paths.append(g_exp)

    ld_path = os.path.join(os.getcwd(), "lib/")
    print("Using guessed paths {0} {1}".format(ld_path, python_paths))
    respawn_in_path(ld_path, python_paths)


try:
    from teuthology.exceptions import CommandFailedError
    from tasks.ceph_manager import CephManager
    from tasks.cephfs.fuse_mount import FuseMount
    from tasks.cephfs.filesystem import Filesystem, MDSCluster, CephCluster
    from tasks.mgr.mgr_test_case import MgrCluster
    from teuthology.contextutil import MaxWhileTries
    from teuthology.task import interactive
except ImportError:
    sys.stderr.write("***\nError importing packages, have you activated your teuthology virtualenv "
                     "and set PYTHONPATH to point to teuthology and ceph-qa-suite?\n***\n\n")
    raise

# Must import after teuthology because of gevent monkey patching
import subprocess

if os.path.exists("./CMakeCache.txt"):
    # Running in build dir of a cmake build
    BIN_PREFIX = "./bin/"
    SRC_PREFIX = "../src"
else:
    # Running in src/ of an autotools build
    BIN_PREFIX = "./"
    SRC_PREFIX = "./"


class LocalRemoteProcess(object):
    def __init__(self, args, subproc, check_status, stdout, stderr):
        self.args = args
        self.subproc = subproc
        self.stdout = stdout
        self.stderr = stderr
        # this variable is meant for instance of this class named fuse_daemon.
        # child process of the command launched with sudo must be killed,
        # since killing parent process alone has no impact on the child
        # process.
        self.fuse_pid = -1

        self.check_status = check_status
        self.exitstatus = self.returncode = None

    def wait(self):
        if self.finished:
            # Avoid calling communicate() on a dead process because it'll
            # give you stick about std* already being closed
            if self.check_status and self.exitstatus != 0:
                raise CommandFailedError(self.args, self.exitstatus)
            else:
                return

        out, err = self.subproc.communicate()
        if isinstance(self.stdout, StringIO):
            self.stdout.write(out.decode(errors='ignore'))
        elif self.stdout is None:
            pass
        else:
            self.stdout.write(out)
        if isinstance(self.stderr, StringIO):
            self.stderr.write(err.decode(errors='ignore'))
        elif self.stderr is None:
            pass
        else:
            self.stderr.write(err)

        self.exitstatus = self.returncode = self.subproc.returncode

        if self.exitstatus != 0:
            sys.stderr.write(six.ensure_str(out))
            sys.stderr.write(six.ensure_str(err))

        if self.check_status and self.exitstatus != 0:
            raise CommandFailedError(self.args, self.exitstatus)

    @property
    def finished(self):
        if self.exitstatus is not None:
            return True

        if self.subproc.poll() is not None:
            out, err = self.subproc.communicate()
            if isinstance(self.stdout, StringIO):
                self.stdout.write(out.decode(errors='ignore'))
            elif self.stdout is None:
                pass
            else:
                self.stdout.write(out)
            if isinstance(self.stderr, StringIO):
                self.stderr.write(err.decode(errors='ignore'))
            elif self.stderr is None:
                pass
            else:
                self.stderr.write(err)
            self.exitstatus = self.returncode = self.subproc.returncode
            return True
        else:
            return False

    def kill(self):
        log.debug("kill ")
        if self.subproc.pid and not self.finished:
            log.debug("kill: killing pid {0} ({1})".format(
                self.subproc.pid, self.args))
            safe_kill(self.subproc.pid)
        else:
            log.debug("kill: already terminated ({0})".format(self.args))

    @property
    def stdin(self):
        class FakeStdIn(object):
            def __init__(self, mount_daemon):
                self.mount_daemon = mount_daemon

            def close(self):
                self.mount_daemon.kill()

        return FakeStdIn(self)


class LocalRemote(object):
    """
    Amusingly named class to present the teuthology RemoteProcess interface when we are really
    running things locally for vstart

    Run this inside your src/ dir!
    """

    def __init__(self):
        self.name = "local"
        self.hostname = "localhost"
        self.user = getpass.getuser()

    def get_file(self, path, sudo, dest_dir):
        tmpfile = tempfile.NamedTemporaryFile(delete=False).name
        shutil.copy(path, tmpfile)
        return tmpfile

    def put_file(self, src, dst, sudo=False):
        shutil.copy(src, dst)

    def run(self, args, check_status=True, wait=True,
            stdout=None, stderr=None, cwd=None, stdin=None,
            logger=None, label=None, env=None, timeout=None, omit_sudo=True):
        try:
            if args[args.index('sudo') + 1] in ['-u', 'passwd', 'chown']:
                omit_sudo = False
        except ValueError:
            pass

        # We don't need no stinkin' sudo
        if omit_sudo:
            args = [a for a in args if a != "sudo"]

        # We have to use shell=True if any run.Raw was present, e.g. &&
        shell = any([a for a in args if isinstance(a, Raw)])

        # Filter out helper tools that don't exist in a vstart environment
        args = [a for a in args if a not in (
            'adjust-ulimits', 'ceph-coverage', 'timeout')]

        # Adjust binary path prefix if given a bare program name
        if "/" not in args[0]:
            # If they asked for a bare binary name, and it exists
            # in our built tree, use the one there.
            local_bin = os.path.join(BIN_PREFIX, args[0])
            if os.path.exists(local_bin):
                args = [local_bin] + args[1:]
            else:
                log.debug("'{0}' is not a binary in the Ceph build dir".format(
                    args[0]
                ))

        log.debug("Running {0}".format(args))

        if shell:
            subproc = subprocess.Popen(quote(args),
                                       stdout=subprocess.PIPE,
                                       stderr=subprocess.PIPE,
                                       stdin=subprocess.PIPE,
                                       cwd=cwd,
                                       shell=True)
        else:
            # Sanity check that we've got a list of strings
            for arg in args:
                if not isinstance(arg, six.string_types):
                    raise RuntimeError("Oops, can't handle arg {0} type {1}".format(
                        arg, arg.__class__
                    ))

            subproc = subprocess.Popen(args,
                                       stdout=subprocess.PIPE,
                                       stderr=subprocess.PIPE,
                                       stdin=subprocess.PIPE,
                                       cwd=cwd,
                                       env=env)

        if stdin:
            # Hack: writing to stdin is not deadlock-safe, but it "always" works
            # as long as the input buffer is "small"
            if isinstance(stdin, str):
                subproc.stdin.write(stdin.encode())
            else:
                subproc.stdin.write(stdin)

        proc = LocalRemoteProcess(
            args, subproc, check_status,
            stdout, stderr
        )

        if wait:
            proc.wait()

        return proc

    # XXX: for compatibility keep this method same teuthology.orchestra.remote.sh
    def sh(self, script, **kwargs):
        """
        Shortcut for run method.

        Usage:
            my_name = remote.sh('whoami')
            remote_date = remote.sh('date')
        """
        if 'stdout' not in kwargs:
            kwargs['stdout'] = StringIO()
        if 'args' not in kwargs:
            kwargs['args'] = script
        proc = self.run(**kwargs)
        return proc.stdout.getvalue()


class LocalDaemon(object):
    def __init__(self, daemon_type, daemon_id):
        self.daemon_type = daemon_type
        self.daemon_id = daemon_id
        self.controller = LocalRemote()
        self.proc = None

    @property
    def remote(self):
        return LocalRemote()

    def running(self):
        return self._get_pid() is not None

    def check_status(self):
        if self.proc:
            return self.proc.poll()

    def _get_pid(self):
        """
        Return PID as an integer or None if not found
        """
        ps_txt = self.controller.run(args=["ps", "ww", "-u"+str(os.getuid())],
                                     stdout=StringIO()).\
            stdout.getvalue().strip()
        lines = ps_txt.split("\n")[1:]

        for line in lines:
            if line.find("ceph-{0} -i {1}".format(self.daemon_type, self.daemon_id)) != -1:
                log.debug("Found ps line for daemon: {0}".format(line))
                return int(line.split()[0])
        log.debug("No match for {0} {1}: {2}".format(
            self.daemon_type, self.daemon_id, ps_txt
            ))
        return None

    def wait(self, timeout):
        waited = 0
        while self._get_pid() is not None:
            if waited > timeout:
                raise MaxWhileTries("Timed out waiting for daemon {0}.{1}".format(self.daemon_type, self.daemon_id))
            time.sleep(1)
            waited += 1

    def stop(self, timeout=300):
        if not self.running():
            log.error('tried to stop a non-running daemon')
            return

        pid = self._get_pid()
        log.debug("Killing PID {0} for {1}.{2}".format(pid, self.daemon_type, self.daemon_id))
        os.kill(pid, signal.SIGTERM)

        waited = 0
        while pid is not None:
            new_pid = self._get_pid()
            if new_pid is not None and new_pid != pid:
                log.debug("Killing new PID {0}".format(new_pid))
                pid = new_pid
                os.kill(pid, signal.SIGTERM)

            if new_pid is None:
                break
            else:
                if waited > timeout:
                    raise MaxWhileTries(
                        "Timed out waiting for daemon {0}.{1}".format(
                            self.daemon_type, self.daemon_id))
                time.sleep(1)
                waited += 1

        self.wait(timeout=timeout)

    def restart(self):
        if self._get_pid() is not None:
            self.stop()

        self.proc = self.controller.run([os.path.join(BIN_PREFIX, "./ceph-{0}".format(self.daemon_type)), "-i", self.daemon_id])

    def signal(self, sig, silent=False):
        if not self.running():
            raise RuntimeError("Can't send signal to non-running daemon")

        os.kill(self._get_pid(), sig)
        if not silent:
            log.debug("Sent signal {0} to {1}.{2}".format(sig, self.daemon_type, self.daemon_id))


def safe_kill(pid):
    """
    os.kill annoyingly raises exception if process already dead.  Ignore it.
    """
    try:
        return os.kill(pid, signal.SIGKILL)
    except OSError as e:
        if e.errno == errno.ESRCH:
            # Raced with process termination
            pass
        else:
            raise


class LocalFuseMount(FuseMount):
    def __init__(self, ctx, test_dir, client_id):
        super(LocalFuseMount, self).__init__(ctx, None, test_dir, client_id, LocalRemote())

    @property
    def config_path(self):
        return "./ceph.conf"

    def get_keyring_path(self):
        # This is going to end up in a config file, so use an absolute path
        # to avoid assumptions about daemons' pwd
        return os.path.abspath("./client.{0}.keyring".format(self.client_id))

    def run_shell(self, args, wait=True, check_status=True, omit_sudo=True):
        # FIXME maybe should add a pwd arg to teuthology.orchestra so that
        # the "cd foo && bar" shenanigans isn't needed to begin with and
        # then we wouldn't have to special case this
        return self.client_remote.run(args, wait=wait, cwd=self.mountpoint,
                                      check_status=check_status,
                                      omit_sudo=omit_sudo)

    def setupfs(self, name=None):
        if name is None and self.fs is not None:
            # Previous mount existed, reuse the old name
            name = self.fs.name
        self.fs = LocalFilesystem(self.ctx, name=name)
        log.debug('Wait for MDS to reach steady state...')
        self.fs.wait_for_daemons()
        log.debug('Ready to start {}...'.format(type(self).__name__))

    @property
    def _prefix(self):
        return BIN_PREFIX

    def _asok_path(self):
        # In teuthology, the asok is named after the PID of the ceph-fuse process, because it's
        # run foreground.  When running it daemonized however, the asok is named after
        # the PID of the launching process, not the long running ceph-fuse process.  Therefore
        # we need to give an exact path here as the logic for checking /proc/ for which
        # asok is alive does not work.

        # Load the asok path from ceph.conf as vstart.sh now puts admin sockets
        # in a tmpdir. All of the paths are the same, so no need to select
        # based off of the service type.
        d = "./out"
        with open(self.config_path) as f:
            for line in f:
                asok_conf = re.search("^\s*admin\s+socket\s*=\s*(.*?)[^/]+$", line)
                if asok_conf:
                    d = asok_conf.groups(1)[0]
                    break
        path = "{0}/client.{1}.{2}.asok".format(d, self.client_id, self.fuse_daemon.subproc.pid)
        log.info("I think my launching pid was {0}".format(self.fuse_daemon.subproc.pid))
        return path

    def mount(self, mount_path=None, mount_fs_name=None, mountpoint=None):
        if mountpoint is not None:
            self.mountpoint = mountpoint
        self.setupfs(name=mount_fs_name)

        self.client_remote.run(args=['mkdir', '-p', self.mountpoint])

        def list_connections():
            self.client_remote.run(
                args=["mount", "-t", "fusectl", "/sys/fs/fuse/connections", "/sys/fs/fuse/connections"],
                check_status=False
            )

            p = self.client_remote.run(args=["ls", "/sys/fs/fuse/connections"],
                                       check_status=False, stdout=StringIO())
            if p.exitstatus != 0:
                log.warning("ls conns failed with {0}, assuming none".format(p.exitstatus))
                return []

            ls_str = p.stdout.getvalue().strip()
            if ls_str:
                return [int(n) for n in ls_str.split("\n")]
            else:
                return []

        # Before starting ceph-fuse process, note the contents of
        # /sys/fs/fuse/connections
        pre_mount_conns = list_connections()
        log.debug("Pre-mount connections: {0}".format(pre_mount_conns))

        prefix = [os.path.join(BIN_PREFIX, "ceph-fuse")]
        if os.getuid() != 0:
            prefix += ["--client_die_on_failed_dentry_invalidate=false"]

        if mount_path is not None:
            prefix += ["--client_mountpoint={0}".format(mount_path)]

        if mount_fs_name is not None:
            prefix += ["--client_mds_namespace={0}".format(mount_fs_name)]

        self.fuse_daemon = self.client_remote.run(args=
                                            prefix + [
                                                "-f",
                                                "--name",
                                                "client.{0}".format(self.client_id),
                                                self.mountpoint
                                            ], wait=False)

        log.debug("Mounting client.{0} with pid {1}".format(self.client_id, self.fuse_daemon.subproc.pid))

        # Wait for the connection reference to appear in /sys
        waited = 0
        post_mount_conns = list_connections()
        while len(post_mount_conns) <= len(pre_mount_conns):
            if self.fuse_daemon.finished:
                # Did mount fail?  Raise the CommandFailedError instead of
                # hitting the "failed to populate /sys/" timeout
                self.fuse_daemon.wait()
            time.sleep(1)
            waited += 1
            if waited > 30:
                raise RuntimeError("Fuse mount failed to populate /sys/ after {0} seconds".format(
                    waited
                ))
            post_mount_conns = list_connections()

        log.debug("Post-mount connections: {0}".format(post_mount_conns))

        # Record our fuse connection number so that we can use it when
        # forcing an unmount
        new_conns = list(set(post_mount_conns) - set(pre_mount_conns))
        if len(new_conns) == 0:
            raise RuntimeError("New fuse connection directory not found ({0})".format(new_conns))
        elif len(new_conns) > 1:
            raise RuntimeError("Unexpectedly numerous fuse connections {0}".format(new_conns))
        else:
            self._fuse_conn = new_conns[0]

        self.gather_mount_info()

        self.mounted = True

    def _run_python(self, pyscript, py_version='python'):
        """
        Override this to remove the daemon-helper prefix that is used otherwise
        to make the process killable.
        """
        return self.client_remote.run(args=[py_version, '-c', pyscript],
                                      wait=False, stdout=StringIO())

class LocalCephManager(CephManager):
    def __init__(self):
        # Deliberately skip parent init, only inheriting from it to get
        # util methods like osd_dump that sit on top of raw_cluster_cmd
        self.controller = LocalRemote()

        # A minority of CephManager fns actually bother locking for when
        # certain teuthology tests want to run tasks in parallel
        self.lock = threading.RLock()

        self.log = lambda x: log.debug(x)

        # Don't bother constructing a map of pools: it should be empty
        # at test cluster start, and in any case it would be out of date
        # in no time.  The attribute needs to exist for some of the CephManager
        # methods to work though.
        self.pools = {}

    def find_remote(self, daemon_type, daemon_id):
        """
        daemon_type like 'mds', 'osd'
        daemon_id like 'a', '0'
        """
        return LocalRemote()

    def run_ceph_w(self, watch_channel=None):
        """
        :param watch_channel: Specifies the channel to be watched.
                              This can be 'cluster', 'audit', ...
        :type watch_channel: str
        """
        args = [os.path.join(BIN_PREFIX, "ceph"), "-w"]
        if watch_channel is not None:
            args.append("--watch-channel")
            args.append(watch_channel)
        proc = self.controller.run(args=args, wait=False, stdout=StringIO())
        return proc

    def raw_cluster_cmd(self, *args, **kwargs):
        """
        args like ["osd", "dump"}
        return stdout string
        """
        proc = self.controller.run(args=[os.path.join(BIN_PREFIX, "ceph")] +\
                                   list(args), **kwargs, stdout=StringIO())
        return proc.stdout.getvalue()

    def raw_cluster_cmd_result(self, *args, **kwargs):
        """
        like raw_cluster_cmd but don't check status, just return rc
        """
        kwargs['check_status'] = False
        proc = self.controller.run([os.path.join(BIN_PREFIX, "ceph")] + list(args), **kwargs)
        return proc.exitstatus

    def admin_socket(self, daemon_type, daemon_id, command, check_status=True,
                     timeout=None, stdout=None):
        if stdout is None:
            stdout = StringIO()

        return self.controller.run(
            args=[os.path.join(BIN_PREFIX, "ceph"), "daemon",
                  "{0}.{1}".format(daemon_type, daemon_id)] + command,
            check_status=check_status, timeout=timeout, stdout=stdout)


class LocalCephCluster(CephCluster):
    def __init__(self, ctx):
        # Deliberately skip calling parent constructor
        self._ctx = ctx
        self.mon_manager = LocalCephManager()
        self._conf = defaultdict(dict)

    @property
    def admin_remote(self):
        return LocalRemote()

    def get_config(self, key, service_type=None):
        if service_type is None:
            service_type = 'mon'

        # FIXME hardcoded vstart service IDs
        service_id = {
            'mon': 'a',
            'mds': 'a',
            'osd': '0'
        }[service_type]

        return self.json_asok(['config', 'get', key], service_type, service_id)[key]

    def _write_conf(self):
        # In teuthology, we have the honour of writing the entire ceph.conf, but
        # in vstart land it has mostly already been written and we need to carefully
        # append to it.
        conf_path = "./ceph.conf"
        banner = "\n#LOCAL_TEST\n"
        existing_str = open(conf_path).read()

        if banner in existing_str:
            existing_str = existing_str[0:existing_str.find(banner)]

        existing_str += banner

        for subsys, kvs in self._conf.items():
            existing_str += "\n[{0}]\n".format(subsys)
            for key, val in kvs.items():
                # Comment out existing instance if it exists
                log.debug("Searching for existing instance {0}/{1}".format(
                    key, subsys
                ))
                existing_section = re.search("^\[{0}\]$([\n]|[^\[])+".format(
                    subsys
                ), existing_str, re.MULTILINE)

                if existing_section:
                    section_str = existing_str[existing_section.start():existing_section.end()]
                    existing_val = re.search("^\s*[^#]({0}) =".format(key), section_str, re.MULTILINE)
                    if existing_val:
                        start = existing_section.start() + existing_val.start(1)
                        log.debug("Found string to replace at {0}".format(
                            start
                        ))
                        existing_str = existing_str[0:start] + "#" + existing_str[start:]

                existing_str += "{0} = {1}\n".format(key, val)

        open(conf_path, "w").write(existing_str)

    def set_ceph_conf(self, subsys, key, value):
        self._conf[subsys][key] = value
        self._write_conf()

    def clear_ceph_conf(self, subsys, key):
        del self._conf[subsys][key]
        self._write_conf()


class LocalMDSCluster(LocalCephCluster, MDSCluster):
    def __init__(self, ctx):
        super(LocalMDSCluster, self).__init__(ctx)

        self.mds_ids = ctx.daemons.daemons['ceph.mds'].keys()
        self.mds_daemons = dict([(id_, LocalDaemon("mds", id_)) for id_ in self.mds_ids])

    def clear_firewall(self):
        # FIXME: unimplemented
        pass

    def newfs(self, name='cephfs', create=True):
        return LocalFilesystem(self._ctx, name=name, create=create)


class LocalMgrCluster(LocalCephCluster, MgrCluster):
    def __init__(self, ctx):
        super(LocalMgrCluster, self).__init__(ctx)

        self.mgr_ids = ctx.daemons.daemons['ceph.mgr'].keys()
        self.mgr_daemons = dict([(id_, LocalDaemon("mgr", id_)) for id_ in self.mgr_ids])


class LocalFilesystem(Filesystem, LocalMDSCluster):
    def __init__(self, ctx, fscid=None, name='cephfs', create=False):
        # Deliberately skip calling parent constructor
        self._ctx = ctx

        self.id = None
        self.name = None
        self.ec_profile = None
        self.metadata_pool_name = None
        self.metadata_overlay = False
        self.data_pool_name = None
        self.data_pools = None
        self.fs_config = None

        # Hack: cheeky inspection of ceph.conf to see what MDSs exist
        self.mds_ids = set()
        for line in open("ceph.conf").readlines():
            match = re.match("^\[mds\.(.+)\]$", line)
            if match:
                self.mds_ids.add(match.group(1))

        if not self.mds_ids:
            raise RuntimeError("No MDSs found in ceph.conf!")

        self.mds_ids = list(self.mds_ids)

        log.debug("Discovered MDS IDs: {0}".format(self.mds_ids))

        self.mon_manager = LocalCephManager()

        self.mds_daemons = dict([(id_, LocalDaemon("mds", id_)) for id_ in self.mds_ids])

        self.client_remote = LocalRemote()

        self._conf = defaultdict(dict)

        if name is not None:
            if fscid is not None:
                raise RuntimeError("cannot specify fscid when creating fs")
            if create and not self.legacy_configured():
                self.create()
        else:
            if fscid is not None:
                self.id = fscid
                self.getinfo(refresh=True)

        # Stash a reference to the first created filesystem on ctx, so
        # that if someone drops to the interactive shell they can easily
        # poke our methods.
        if not hasattr(self._ctx, "filesystem"):
            self._ctx.filesystem = self

    @property
    def _prefix(self):
        return BIN_PREFIX

    def set_clients_block(self, blocked, mds_id=None):
        raise NotImplementedError()


class InteractiveFailureResult(unittest.TextTestResult):
    """
    Specialization that implements interactive-on-error style
    behavior.
    """
    def addFailure(self, test, err):
        super(InteractiveFailureResult, self).addFailure(test, err)
        log.error(self._exc_info_to_string(err, test))
        log.error("Failure in test '{0}', going interactive".format(
            self.getDescription(test)
        ))
        interactive.task(ctx=None, config=None)

    def addError(self, test, err):
        super(InteractiveFailureResult, self).addError(test, err)
        log.error(self._exc_info_to_string(err, test))
        log.error("Error in test '{0}', going interactive".format(
            self.getDescription(test)
        ))
        interactive.task(ctx=None, config=None)


def enumerate_methods(s):
    log.debug("e: {0}".format(s))
    for t in s._tests:
        if isinstance(t, suite.BaseTestSuite):
            for sub in enumerate_methods(t):
                yield sub
        else:
            yield s, t


def load_tests(modules, loader):
    if modules:
        log.debug("Executing modules: {0}".format(modules))
        module_suites = []
        for mod_name in modules:
            # Test names like cephfs.test_auto_repair
            module_suites.append(loader.loadTestsFromName(mod_name))
        log.debug("Loaded: {0}".format(list(module_suites)))
        return suite.TestSuite(module_suites)
    else:
        log.debug("Executing all cephfs tests")
        return loader.discover(
            os.path.join(os.path.dirname(os.path.abspath(__file__)), "cephfs")
        )


def scan_tests(modules):
    overall_suite = load_tests(modules, loader.TestLoader())

    max_required_mds = 0
    max_required_clients = 0
    max_required_mgr = 0
    require_memstore = False

    for suite_, case in enumerate_methods(overall_suite):
        max_required_mds = max(max_required_mds,
                               getattr(case, "MDSS_REQUIRED", 0))
        max_required_clients = max(max_required_clients,
                               getattr(case, "CLIENTS_REQUIRED", 0))
        max_required_mgr = max(max_required_mgr,
                               getattr(case, "MGRS_REQUIRED", 0))
        require_memstore = getattr(case, "REQUIRE_MEMSTORE", False) \
                               or require_memstore

    return max_required_mds, max_required_clients, \
            max_required_mgr, require_memstore


class LocalCluster(object):
    def __init__(self, rolename="placeholder"):
        self.remotes = {
            LocalRemote(): [rolename]
        }

    def only(self, requested):
        return self.__class__(rolename=requested)


class LocalContext(object):
    def __init__(self):
        self.config = {}
        self.teuthology_config = teuth_config
        self.cluster = LocalCluster()
        self.daemons = DaemonGroup()

        # Shove some LocalDaemons into the ctx.daemons DaemonGroup instance so that any
        # tests that want to look these up via ctx can do so.
        # Inspect ceph.conf to see what roles exist
        for conf_line in open("ceph.conf").readlines():
            for svc_type in ["mon", "osd", "mds", "mgr"]:
                prefixed_type = "ceph." + svc_type
                if prefixed_type not in self.daemons.daemons:
                    self.daemons.daemons[prefixed_type] = {}
                match = re.match("^\[{0}\.(.+)\]$".format(svc_type), conf_line)
                if match:
                    svc_id = match.group(1)
                    self.daemons.daemons[prefixed_type][svc_id] = LocalDaemon(svc_type, svc_id)

    def __del__(self):
        shutil.rmtree(self.teuthology_config['test_path'])

def exec_test():
    # Parse arguments
    interactive_on_error = False
    create_cluster = False
    create_cluster_only = False
    ignore_missing_binaries = False
    opt_verbose = True

    args = sys.argv[1:]
    flags = [a for a in args if a.startswith("-")]
    modules = [a for a in args if not a.startswith("-")]
    for f in flags:
        if f == "--interactive":
            interactive_on_error = True
        elif f == "--create":
            create_cluster = True
        elif f == "--create-cluster-only":
            create_cluster_only = True
        elif f == "--ignore-missing-binaries":
            ignore_missing_binaries = True
        elif '--no-verbose' == f:
            opt_verbose = False
        else:
            log.error("Unknown option '{0}'".format(f))
            sys.exit(-1)

    # Help developers by stopping up-front if their tree isn't built enough for all the
    # tools that the tests might want to use (add more here if needed)
    require_binaries = ["ceph-dencoder", "cephfs-journal-tool", "cephfs-data-scan",
                        "cephfs-table-tool", "ceph-fuse", "rados"]
    missing_binaries = [b for b in require_binaries if not os.path.exists(os.path.join(BIN_PREFIX, b))]
    if missing_binaries and not ignore_missing_binaries:
        log.error("Some ceph binaries missing, please build them: {0}".format(" ".join(missing_binaries)))
        sys.exit(-1)

    max_required_mds, max_required_clients, \
            max_required_mgr, require_memstore = scan_tests(modules)

    remote = LocalRemote()

    # Tolerate no MDSs or clients running at start
    ps_txt = remote.run(args=["ps", "-u"+str(os.getuid())],
                        stdout=StringIO()).stdout.getvalue().strip()
    lines = ps_txt.split("\n")[1:]
    for line in lines:
        if 'ceph-fuse' in line or 'ceph-mds' in line:
            pid = int(line.split()[0])
            log.warning("Killing stray process {0}".format(line))
            os.kill(pid, signal.SIGKILL)

    # Fire up the Ceph cluster if the user requested it
    if create_cluster or create_cluster_only:
        log.info("Creating cluster with {0} MDS daemons".format(
            max_required_mds))
        remote.run([os.path.join(SRC_PREFIX, "stop.sh")], check_status=False)
        remote.run(["rm", "-rf", "./out"])
        remote.run(["rm", "-rf", "./dev"])
        vstart_env = os.environ.copy()
        vstart_env["FS"] = "0"
        vstart_env["MDS"] = max_required_mds.__str__()
        vstart_env["OSD"] = "4"
        vstart_env["MGR"] = max(max_required_mgr, 1).__str__()

        args = [
            os.path.join(SRC_PREFIX, "vstart.sh"),
            "-n",
            "--nolockdep",
        ]
        if require_memstore:
            args.append("--memstore")

        if opt_verbose:
            args.append("-d")

        remote.run(args, env=vstart_env)

        # Wait for OSD to come up so that subsequent injectargs etc will
        # definitely succeed
        LocalCephCluster(LocalContext()).mon_manager.wait_for_all_osds_up(timeout=30)

    if create_cluster_only:
        return

    # List of client mounts, sufficient to run the selected tests
    clients = [i.__str__() for i in range(0, max_required_clients)]

    test_dir = tempfile.mkdtemp()
    teuth_config['test_path'] = test_dir

    ctx = LocalContext()
    ceph_cluster = LocalCephCluster(ctx)
    mds_cluster = LocalMDSCluster(ctx)
    mgr_cluster = LocalMgrCluster(ctx)

    # Construct Mount classes
    mounts = []
    for client_id in clients:
        # Populate client keyring (it sucks to use client.admin for test clients
        # because it's awkward to find the logs later)
        client_name = "client.{0}".format(client_id)

        if client_name not in open("./keyring").read():
            p = remote.run(args=[os.path.join(BIN_PREFIX, "ceph"), "auth", "get-or-create", client_name,
                                 "osd", "allow rw",
                                 "mds", "allow",
                                 "mon", "allow r"], stdout=StringIO())

            open("./keyring", "at").write(p.stdout.getvalue())

        mount = LocalFuseMount(ctx, test_dir, client_id)
        mounts.append(mount)
        if mount.is_mounted():
            log.warning("unmounting {0}".format(mount.mountpoint))
            mount.umount_wait()
        else:
            if os.path.exists(mount.mountpoint):
                os.rmdir(mount.mountpoint)

    from tasks.cephfs_test_runner import DecoratingLoader

    class LogStream(object):
        def __init__(self):
            self.buffer = ""

        def write(self, data):
            self.buffer += data
            if "\n" in self.buffer:
                lines = self.buffer.split("\n")
                for line in lines[:-1]:
                    pass
                    # sys.stderr.write(line + "\n")
                    log.info(line)
                self.buffer = lines[-1]

        def flush(self):
            pass

    decorating_loader = DecoratingLoader({
        "ctx": ctx,
        "mounts": mounts,
        "ceph_cluster": ceph_cluster,
        "mds_cluster": mds_cluster,
        "mgr_cluster": mgr_cluster,
    })

    # For the benefit of polling tests like test_full -- in teuthology land we set this
    # in a .yaml, here it's just a hardcoded thing for the developer's pleasure.
    remote.run(args=[os.path.join(BIN_PREFIX, "ceph"), "tell", "osd.*", "injectargs", "--osd-mon-report-interval", "5"])
    ceph_cluster.set_ceph_conf("osd", "osd_mon_report_interval", "5")

    # Vstart defaults to two segments, which very easily gets a "behind on trimming" health warning
    # from normal IO latency.  Increase it for running teests.
    ceph_cluster.set_ceph_conf("mds", "mds log max segments", "10")

    # Make sure the filesystem created in tests has uid/gid that will let us talk to
    # it after mounting it (without having to  go root).  Set in 'global' not just 'mds'
    # so that cephfs-data-scan will pick it up too.
    ceph_cluster.set_ceph_conf("global", "mds root ino uid", "%s" % os.getuid())
    ceph_cluster.set_ceph_conf("global", "mds root ino gid", "%s" % os.getgid())

    # Monkeypatch get_package_version to avoid having to work out what kind of distro we're on
    def _get_package_version(remote, pkg_name):
        # Used in cephfs tests to find fuse version.  Your development workstation *does* have >=2.9, right?
        return "2.9"

    import teuthology.packaging
    teuthology.packaging.get_package_version = _get_package_version

    overall_suite = load_tests(modules, decorating_loader)

    # Filter out tests that don't lend themselves to interactive running,
    victims = []
    for case, method in enumerate_methods(overall_suite):
        fn = getattr(method, method._testMethodName)

        drop_test = False

        if hasattr(fn, 'is_for_teuthology') and getattr(fn, 'is_for_teuthology') is True:
            drop_test = True
            log.warning("Dropping test because long running: ".format(method.id()))

        if getattr(fn, "needs_trimming", False) is True:
            drop_test = (os.getuid() != 0)
            log.warning("Dropping test because client trim unavailable: ".format(method.id()))

        if drop_test:
            # Don't drop the test if it was explicitly requested in arguments
            is_named = False
            for named in modules:
                if named.endswith(method.id()):
                    is_named = True
                    break

            if not is_named:
                victims.append((case, method))

    log.debug("Disabling {0} tests because of is_for_teuthology or needs_trimming".format(len(victims)))
    for s, method in victims:
        s._tests.remove(method)

    if interactive_on_error:
        result_class = InteractiveFailureResult
    else:
        result_class = unittest.TextTestResult
    fail_on_skip = False

    class LoggingResult(result_class):
        def startTest(self, test):
            log.info("Starting test: {0}".format(self.getDescription(test)))
            test.started_at = datetime.datetime.utcnow()
            return super(LoggingResult, self).startTest(test)

        def stopTest(self, test):
            log.info("Stopped test: {0} in {1}s".format(
                self.getDescription(test),
                (datetime.datetime.utcnow() - test.started_at).total_seconds()
            ))

        def addSkip(self, test, reason):
            if fail_on_skip:
                # Don't just call addFailure because that requires a traceback
                self.failures.append((test, reason))
            else:
                super(LoggingResult, self).addSkip(test, reason)

    # Execute!
    result = unittest.TextTestRunner(
        stream=LogStream(),
        resultclass=LoggingResult,
        verbosity=2,
        failfast=True).run(overall_suite)

    if not result.wasSuccessful():
        result.printErrors()  # duplicate output at end for convenience

        bad_tests = []
        for test, error in result.errors:
            bad_tests.append(str(test))
        for test, failure in result.failures:
            bad_tests.append(str(test))

        sys.exit(-1)
    else:
        sys.exit(0)


if __name__ == "__main__":
    exec_test()