def cod_setup(log, ctx, remote, NUM_OBJECTS, DATADIR, REP_NAME, DATALINECOUNT, REP_POOL, db):
    ERRORS = 0
    log.info("Creating {objs} objects in replicated pool".format(objs=NUM_OBJECTS))
    nullfd = open(os.devnull, "w")

    objects = range(1, NUM_OBJECTS + 1)
    for i in objects:
        NAME = REP_NAME + "{num}".format(num=i)
        DDNAME = os.path.join(DATADIR, NAME)

        proc = rados(ctx, remote, ['-p', REP_POOL, 'put', NAME, DDNAME], wait=False)
        # proc = remote.run(args=['rados', '-p', REP_POOL, 'put', NAME, DDNAME])
        ret = proc.wait()
        if ret != 0:
            log.critical("Rados put failed with status {ret}".format(ret=r[0].exitstatus))
            sys.exit(1)

        db[NAME] = {}

        keys = range(i)
        db[NAME]["xattr"] = {}
        for k in keys:
            if k == 0:
                continue
            mykey = "key{i}-{k}".format(i=i, k=k)
            myval = "val{i}-{k}".format(i=i, k=k)
            proc = remote.run(args=['rados', '-p', REP_POOL, 'setxattr', NAME, mykey, myval])
            ret = proc.wait()
            if ret != 0:
                log.error("setxattr failed with {ret}".format(ret=ret))
                ERRORS += 1
            db[NAME]["xattr"][mykey] = myval

        # Create omap header in all objects but REPobject1
        if i != 1:
            myhdr = "hdr{i}".format(i=i)
            proc = remote.run(args=['rados', '-p', REP_POOL, 'setomapheader', NAME, myhdr])
            ret = proc.wait()
            if ret != 0:
                log.critical("setomapheader failed with {ret}".format(ret=ret))
                ERRORS += 1
            db[NAME]["omapheader"] = myhdr

        db[NAME]["omap"] = {}
        for k in keys:
            if k == 0:
                continue
            mykey = "okey{i}-{k}".format(i=i, k=k)
            myval = "oval{i}-{k}".format(i=i, k=k)
            proc = remote.run(args=['rados', '-p', REP_POOL, 'setomapval', NAME, mykey, myval])
            ret = proc.wait()
            if ret != 0:
                log.critical("setomapval failed with {ret}".format(ret=ret))
            db[NAME]["omap"][mykey] = myval

    nullfd.close()
    return ERRORS

def task(ctx, config):
    """
    Test peering.
    """
    if config is None:
        config = {}
    assert isinstance(config, dict), \
        'peer task only accepts a dict for configuration'
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()

    manager = ceph_manager.CephManager(
        mon,
        ctx=ctx,
        logger=log.getChild('ceph_manager'),
        )

    while len(manager.get_osd_status()['up']) < 3:
        time.sleep(10)
    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.1', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.wait_for_clean()

    for i in range(3):
        manager.set_config(
            i,
            osd_recovery_delay_start=120)

    # take on osd down
    manager.kill_osd(2)
    manager.mark_down_osd(2)

    # kludge to make sure they get a map
    rados(ctx, mon, ['-p', 'data', 'get', 'dummy', '-'])

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.1', 'flush_pg_stats')
    manager.wait_for_recovery()

    # kill another and revive 2, so that some pgs can't peer.
    manager.kill_osd(1)
    manager.mark_down_osd(1)
    manager.revive_osd(2)
    manager.wait_till_osd_is_up(2)

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')

    manager.wait_for_active_or_down()

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')

    # look for down pgs
    num_down_pgs = 0
    pgs = manager.get_pg_stats()
    for pg in pgs:
        out = manager.raw_cluster_cmd('pg', pg['pgid'], 'query')
	log.debug("out string %s",out)
        j = json.loads(out)
        log.info("pg is %s, query json is %s", pg, j)

        if pg['state'].count('down'):
            num_down_pgs += 1
            # verify that it is blocked on osd.1
            rs = j['recovery_state']
            assert len(rs) > 0
            assert rs[0]['name'] == 'Started/Primary/Peering/GetInfo'
            assert rs[1]['name'] == 'Started/Primary/Peering'
            assert rs[1]['blocked']
            assert rs[1]['down_osds_we_would_probe'] == [1]
            assert len(rs[1]['peering_blocked_by']) == 1
            assert rs[1]['peering_blocked_by'][0]['osd'] == 1

    assert num_down_pgs > 0

    # bring it all back
    manager.revive_osd(1)
    manager.wait_till_osd_is_up(1)
    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.1', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.wait_for_clean()

def task(ctx, config):
    """
    Test handling of osd_failsafe_nearfull_ratio and osd_failsafe_full_ratio
    configuration settings

    In order for test to pass must use log-whitelist as follows

        tasks:
            - chef:
            - install:
            - ceph:
                log-whitelist: ['OSD near full', 'OSD full dropping all updates']
            - osd_failsafe_enospc:

    """
    if config is None:
        config = {}
    assert isinstance(config, dict), \
        'osd_failsafe_enospc task only accepts a dict for configuration'

    # Give 2 seconds for injectargs + osd_op_complaint_time (30) + 2 * osd_heartbeat_interval (6) + 6 padding
    sleep_time = 50

    # something that is always there
    dummyfile = '/etc/fstab'
    dummyfile2 = '/etc/resolv.conf'

    # create 1 pg pool with 1 rep which can only be on osd.0
    osds = ctx.manager.get_osd_dump()
    for osd in osds:
        if osd['osd'] != 0:
            ctx.manager.mark_out_osd(osd['osd'])

    log.info('creating pool foo')
    ctx.manager.create_pool("foo")
    ctx.manager.raw_cluster_cmd('osd', 'pool', 'set', 'foo', 'size', '1')

    # State NONE -> NEAR
    log.info('1. Verify warning messages when exceeding nearfull_ratio')

    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()

    proc = mon.run(
             args=[
                'daemon-helper',
                'kill',
                'ceph', '-w'
             ],
             stdin=run.PIPE,
             stdout=StringIO(),
             wait=False,
        )

    ctx.manager.raw_cluster_cmd('tell', 'osd.0', 'injectargs', '--osd_failsafe_nearfull_ratio .00001')

    time.sleep(sleep_time)
    proc.stdin.close() # causes daemon-helper send SIGKILL to ceph -w
    proc.wait()

    lines = proc.stdout.getvalue().split('\n')

    count = len(filter(lambda line: '[WRN] OSD near full' in line, lines))
    assert count == 2, 'Incorrect number of warning messages expected 2 got %d' % count
    count = len(filter(lambda line: '[ERR] OSD full dropping all updates' in line, lines))
    assert count == 0, 'Incorrect number of error messages expected 0 got %d' % count

    # State NEAR -> FULL
    log.info('2. Verify error messages when exceeding full_ratio')

    proc = mon.run(
             args=[
                'daemon-helper',
                'kill',
                'ceph', '-w'
             ],
             stdin=run.PIPE,
             stdout=StringIO(),
             wait=False,
        )

    ctx.manager.raw_cluster_cmd('tell', 'osd.0', 'injectargs', '--osd_failsafe_full_ratio .00001')

    time.sleep(sleep_time)
    proc.stdin.close() # causes daemon-helper send SIGKILL to ceph -w
    proc.wait()

    lines = proc.stdout.getvalue().split('\n')

    count = len(filter(lambda line: '[ERR] OSD full dropping all updates' in line, lines))
    assert count == 2, 'Incorrect number of error messages expected 2 got %d' % count

    log.info('3. Verify write failure when exceeding full_ratio')

    # Write data should fail
    ret = rados(ctx, mon, ['-p', 'foo', 'put', 'newfile1', dummyfile])
    assert ret != 0, 'Expected write failure but it succeeded with exit status 0'

    # Put back default
    ctx.manager.raw_cluster_cmd('tell', 'osd.0', 'injectargs', '--osd_failsafe_full_ratio .97')
#.........这里部分代码省略.........

def task(ctx, config):
    """
    Test handling resolve stuck peering

    requires 3 osds on a single test node
    """
    if config is None:
        config = {}
        assert isinstance(config, dict), \
            'Resolve stuck peering only accepts a dict for config'

    manager = ctx.managers['ceph']

    while len(manager.get_osd_status()['up']) < 3:
        time.sleep(10)


    manager.wait_for_clean()

    dummyfile = '/etc/fstab'
    dummyfile1 = '/etc/resolv.conf'

    #create 1 PG pool
    pool='foo'
    log.info('creating pool foo')
    manager.raw_cluster_cmd('osd', 'pool', 'create', '%s' % pool, '1')

    #set min_size of the pool to 1
    #so that we can continue with I/O
    #when 2 osds are down
    manager.set_pool_property(pool, "min_size", 1)

    osds = [0, 1, 2]

    primary = manager.get_pg_primary('foo', 0)
    log.info("primary osd is %d", primary)

    others = list(osds)
    others.remove(primary)

    log.info('writing initial objects')
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()
    #create few objects
    for i in range(100):
        rados(ctx, mon, ['-p', 'foo', 'put', 'existing_%d' % i, dummyfile])

    manager.wait_for_clean()

    #kill other osds except primary
    log.info('killing other osds except primary')
    for i in others:
        manager.kill_osd(i)
    for i in others:
        manager.mark_down_osd(i)


    for i in range(100):
        rados(ctx, mon, ['-p', 'foo', 'put', 'new_%d' % i, dummyfile1])

    #kill primary osd
    manager.kill_osd(primary)
    manager.mark_down_osd(primary)

    #revive other 2 osds
    for i in others:
        manager.revive_osd(i)

    #make sure that pg is down
    #Assuming pg number for single pg pool will start from 0
    pgnum=0
    pgstr = manager.get_pgid(pool, pgnum)
    stats = manager.get_single_pg_stats(pgstr)
    print stats['state']

    timeout=60
    start=time.time()

    while 'down' not in stats['state']:
        assert time.time() - start < timeout, \
            'failed to reach down state before timeout expired'
        stats = manager.get_single_pg_stats(pgstr)

    #mark primary as lost
    manager.raw_cluster_cmd('osd', 'lost', '%d' % primary,\
                            '--yes-i-really-mean-it')


    #expect the pg status to be active+undersized+degraded
    #pg should recover and become active+clean within timeout
    stats = manager.get_single_pg_stats(pgstr)
    print stats['state']

    timeout=10
    start=time.time()

    while manager.get_num_down():
        assert time.time() - start < timeout, \
            'failed to recover before timeout expired'

#.........这里部分代码省略.........

#.........这里部分代码省略.........
                       format(id=osdid, pg=pg, file=fpath))
                proc = remote.run(args=cmd, check_status=False,
                                  stdout=StringIO())
                proc.wait()
                if proc.exitstatus != 0:
                    log.error("Exporting failed for pg {pg} "
                              "on osd.{id} with {ret}".
                              format(pg=pg, id=osdid, ret=proc.exitstatus))
                    EXP_ERRORS += 1

    ERRORS += EXP_ERRORS

    log.info("Test pg removal")
    RM_ERRORS = 0
    for remote in osds.remotes.iterkeys():
        for role in osds.remotes[remote]:
            if string.find(role, "osd.") != 0:
                continue
            osdid = int(role.split('.')[1])
            if osdid not in pgs:
                continue

            for pg in pgs[osdid]:
                cmd = ((prefix + "--force --op remove --pgid {pg}").
                       format(pg=pg, id=osdid))
                proc = remote.run(args=cmd, check_status=False,
                                  stdout=StringIO())
                proc.wait()
                if proc.exitstatus != 0:
                    log.error("Removing failed for pg {pg} "
                              "on osd.{id} with {ret}".
                              format(pg=pg, id=osdid, ret=proc.exitstatus))
                    RM_ERRORS += 1

    ERRORS += RM_ERRORS

    IMP_ERRORS = 0
    if EXP_ERRORS == 0 and RM_ERRORS == 0:
        log.info("Test pg import")

        for remote in osds.remotes.iterkeys():
            for role in osds.remotes[remote]:
                if string.find(role, "osd.") != 0:
                    continue
                osdid = int(role.split('.')[1])
                if osdid not in pgs:
                    continue

                for pg in pgs[osdid]:
                    fpath = os.path.join(DATADIR, "osd{id}.{pg}".
                                         format(id=osdid, pg=pg))

                    cmd = ((prefix + "--op import --file {file}").
                           format(id=osdid, file=fpath))
                    proc = remote.run(args=cmd, check_status=False,
                                      stdout=StringIO())
                    proc.wait()
                    if proc.exitstatus != 0:
                        log.error("Import failed from {file} with {ret}".
                                  format(file=fpath, ret=proc.exitstatus))
                        IMP_ERRORS += 1
    else:
        log.warning("SKIPPING IMPORT TESTS DUE TO PREVIOUS FAILURES")

    ERRORS += IMP_ERRORS

    if EXP_ERRORS == 0 and RM_ERRORS == 0 and IMP_ERRORS == 0:
        log.info("Restarting OSDs....")
        # They are still look to be up because of setting nodown
        for osd in manager.get_osd_status()['up']:
            manager.revive_osd(osd)
        # Wait for health?
        time.sleep(5)
        # Let scrub after test runs verify consistency of all copies
        log.info("Verify replicated import data")
        objects = range(1, NUM_OBJECTS + 1)
        for i in objects:
            NAME = REP_NAME + "{num}".format(num=i)
            TESTNAME = os.path.join(DATADIR, "gettest")
            REFNAME = os.path.join(DATADIR, NAME)

            proc = rados(ctx, cli_remote,
                         ['-p', REP_POOL, 'get', NAME, TESTNAME], wait=False)

            ret = proc.wait()
            if ret != 0:
                log.error("After import, rados get failed with {ret}".
                          format(ret=proc.exitstatus))
                ERRORS += 1
                continue

            cmd = "diff -q {gettest} {ref}".format(gettest=TESTNAME,
                                                   ref=REFNAME)
            proc = cli_remote.run(args=cmd, check_status=False)
            proc.wait()
            if proc.exitstatus != 0:
                log.error("Data comparison failed for {obj}".format(obj=NAME))
                ERRORS += 1

    return ERRORS

def task(ctx, config):
    """
    Test handling of lost objects.

    A pretty rigid cluseter is brought up andtested by this task
    """
    POOL = 'unfound_pool'
    if config is None:
        config = {}
    assert isinstance(config, dict), \
        'lost_unfound task only accepts a dict for configuration'
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()

    manager = ceph_manager.CephManager(
        mon,
        ctx=ctx,
        logger=log.getChild('ceph_manager'),
        )

    while len(manager.get_osd_status()['up']) < 3:
        time.sleep(10)

    manager.wait_for_clean()

    manager.create_pool(POOL)

    # something that is always there
    dummyfile = '/etc/fstab'

    # take an osd out until the very end
    manager.kill_osd(2)
    manager.mark_down_osd(2)
    manager.mark_out_osd(2)

    # kludge to make sure they get a map
    rados(ctx, mon, ['-p', POOL, 'put', 'dummy', dummyfile])

    manager.flush_pg_stats([0, 1])
    manager.wait_for_recovery()

    # create old objects
    for f in range(1, 10):
        rados(ctx, mon, ['-p', POOL, 'put', 'existing_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', POOL, 'put', 'existed_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', POOL, 'rm', 'existed_%d' % f])

    # delay recovery, and make the pg log very long (to prevent backfill)
    manager.raw_cluster_cmd(
            'tell', 'osd.1',
            'injectargs',
            '--osd-recovery-delay-start 1000 --osd-min-pg-log-entries 100000000'
            )

    manager.kill_osd(0)
    manager.mark_down_osd(0)
    
    for f in range(1, 10):
        rados(ctx, mon, ['-p', POOL, 'put', 'new_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', POOL, 'put', 'existed_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', POOL, 'put', 'existing_%d' % f, dummyfile])

    # bring osd.0 back up, let it peer, but don't replicate the new
    # objects...
    log.info('osd.0 command_args is %s' % 'foo')
    log.info(ctx.daemons.get_daemon('osd', 0).command_args)
    ctx.daemons.get_daemon('osd', 0).command_kwargs['args'].extend([
            '--osd-recovery-delay-start', '1000'
            ])
    manager.revive_osd(0)
    manager.mark_in_osd(0)
    manager.wait_till_osd_is_up(0)

    manager.flush_pg_stats([1, 0])
    manager.wait_till_active()

    # take out osd.1 and the only copy of those objects.
    manager.kill_osd(1)
    manager.mark_down_osd(1)
    manager.mark_out_osd(1)
    manager.raw_cluster_cmd('osd', 'lost', '1', '--yes-i-really-mean-it')

    # bring up osd.2 so that things would otherwise, in theory, recovery fully
    manager.revive_osd(2)
    manager.mark_in_osd(2)
    manager.wait_till_osd_is_up(2)

    manager.flush_pg_stats([0, 2])
    manager.wait_till_active()
    manager.flush_pg_stats([0, 2])

    # verify that there are unfound objects
    unfound = manager.get_num_unfound_objects()
    log.info("there are %d unfound objects" % unfound)
    assert unfound

    testdir = teuthology.get_testdir(ctx)
    procs = []
    if config.get('parallel_bench', True):
        procs.append(mon.run(
#.........这里部分代码省略.........

def task(ctx, config):
    """
    Test handling of lost objects on an ec pool.

    A pretty rigid cluster is brought up andtested by this task
    """
    if config is None:
        config = {}
    assert isinstance(config, dict), \
        'lost_unfound task only accepts a dict for configuration'
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()

    manager = ceph_manager.CephManager(
        mon,
        ctx=ctx,
        logger=log.getChild('ceph_manager'),
        )

    manager.wait_for_clean()

    profile = config.get('erasure_code_profile', {
        'k': '2',
        'm': '2',
        'ruleset-failure-domain': 'osd'
    })
    profile_name = profile.get('name', 'lost_unfound')
    manager.create_erasure_code_profile(profile_name, profile)
    pool = manager.create_pool_with_unique_name(erasure_code_profile_name=profile_name)

    # something that is always there, readable and never empty
    dummyfile = '/etc/group'

    # kludge to make sure they get a map
    rados(ctx, mon, ['-p', pool, 'put', 'dummy', dummyfile])

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.1', 'flush_pg_stats')
    manager.wait_for_recovery()

    # create old objects
    for f in range(1, 10):
        rados(ctx, mon, ['-p', pool, 'put', 'existing_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'put', 'existed_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'rm', 'existed_%d' % f])

    # delay recovery, and make the pg log very long (to prevent backfill)
    manager.raw_cluster_cmd(
            'tell', 'osd.1',
            'injectargs',
            '--osd-recovery-delay-start 1000 --osd-min-pg-log-entries 100000000'
            )

    manager.kill_osd(0)
    manager.mark_down_osd(0)
    manager.kill_osd(3)
    manager.mark_down_osd(3)
    
    for f in range(1, 10):
        rados(ctx, mon, ['-p', pool, 'put', 'new_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'put', 'existed_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'put', 'existing_%d' % f, dummyfile])

    # take out osd.1 and a necessary shard of those objects.
    manager.kill_osd(1)
    manager.mark_down_osd(1)
    manager.raw_cluster_cmd('osd', 'lost', '1', '--yes-i-really-mean-it')
    manager.revive_osd(0)
    manager.wait_till_osd_is_up(0)
    manager.revive_osd(3)
    manager.wait_till_osd_is_up(3)

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.3', 'flush_pg_stats')
    manager.wait_till_active()
    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.3', 'flush_pg_stats')

    # verify that there are unfound objects
    unfound = manager.get_num_unfound_objects()
    log.info("there are %d unfound objects" % unfound)
    assert unfound

    testdir = teuthology.get_testdir(ctx)
    procs = []
    if config.get('parallel_bench', True):
        procs.append(mon.run(
            args=[
                "/bin/sh", "-c",
                " ".join(['adjust-ulimits',
                          'ceph-coverage',
                          '{tdir}/archive/coverage',
                          'rados',
                          '--no-log-to-stderr',
                          '--name', 'client.admin',
                          '-b', str(4<<10),
                          '-p' , pool,
                          '-t', '20',
#.........这里部分代码省略.........

def task(ctx, config):
    """
    Test handling of divergent entries with prior_version
    prior to log_tail

    overrides:
      ceph:
        conf:
          osd:
            debug osd: 5

    Requires 3 osds on a single test node.
    """
    if config is None:
        config = {}
    assert isinstance(config, dict), "divergent_priors task only accepts a dict for configuration"

    manager = ctx.managers["ceph"]

    while len(manager.get_osd_status()["up"]) < 3:
        time.sleep(10)
    manager.raw_cluster_cmd("tell", "osd.0", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.1", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.2", "flush_pg_stats")
    manager.raw_cluster_cmd("osd", "set", "noout")
    manager.raw_cluster_cmd("osd", "set", "noin")
    manager.raw_cluster_cmd("osd", "set", "nodown")
    manager.wait_for_clean()

    # something that is always there
    dummyfile = "/etc/fstab"
    dummyfile2 = "/etc/resolv.conf"

    # create 1 pg pool
    log.info("creating foo")
    manager.raw_cluster_cmd("osd", "pool", "create", "foo", "1")

    osds = [0, 1, 2]
    for i in osds:
        manager.set_config(i, osd_min_pg_log_entries=10)
        manager.set_config(i, osd_max_pg_log_entries=10)
        manager.set_config(i, osd_pg_log_trim_min=5)

    # determine primary
    divergent = manager.get_pg_primary("foo", 0)
    log.info("primary and soon to be divergent is %d", divergent)
    non_divergent = list(osds)
    non_divergent.remove(divergent)

    log.info("writing initial objects")
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()
    # write 100 objects
    for i in range(100):
        rados(ctx, mon, ["-p", "foo", "put", "existing_%d" % i, dummyfile])

    manager.wait_for_clean()

    # blackhole non_divergent
    log.info("blackholing osds %s", str(non_divergent))
    for i in non_divergent:
        manager.set_config(i, filestore_blackhole=1)

    DIVERGENT_WRITE = 5
    DIVERGENT_REMOVE = 5
    # Write some soon to be divergent
    log.info("writing divergent objects")
    for i in range(DIVERGENT_WRITE):
        rados(ctx, mon, ["-p", "foo", "put", "existing_%d" % i, dummyfile2], wait=False)
    # Remove some soon to be divergent
    log.info("remove divergent objects")
    for i in range(DIVERGENT_REMOVE):
        rados(ctx, mon, ["-p", "foo", "rm", "existing_%d" % (i + DIVERGENT_WRITE)], wait=False)
    time.sleep(10)
    mon.run(args=["killall", "-9", "rados"], wait=True, check_status=False)

    # kill all the osds but leave divergent in
    log.info("killing all the osds")
    for i in osds:
        manager.kill_osd(i)
    for i in osds:
        manager.mark_down_osd(i)
    for i in non_divergent:
        manager.mark_out_osd(i)

    # bring up non-divergent
    log.info("bringing up non_divergent %s", str(non_divergent))
    for i in non_divergent:
        manager.revive_osd(i)
    for i in non_divergent:
        manager.mark_in_osd(i)

    # write 1 non-divergent object (ensure that old divergent one is divergent)
    objname = "existing_%d" % (DIVERGENT_WRITE + DIVERGENT_REMOVE)
    log.info("writing non-divergent object " + objname)
    rados(ctx, mon, ["-p", "foo", "put", objname, dummyfile2])

    manager.wait_for_recovery()

    # ensure no recovery of up osds first
#.........这里部分代码省略.........

def configure_regions_and_zones(ctx, config, regions, role_endpoints):
    """
    Configure regions and zones from rados and rgw.
    """
    if not regions:
        log.debug(
            'In rgw.configure_regions_and_zones() and regions is None. '
            'Bailing')
        yield
        return

    log.info('Configuring regions and zones...')

    log.debug('config is %r', config)
    log.debug('regions are %r', regions)
    log.debug('role_endpoints = %r', role_endpoints)
    # extract the zone info
    role_zones = dict([(client, extract_zone_info(ctx, client, c_config))
                       for client, c_config in config.iteritems()])
    log.debug('roles_zones = %r', role_zones)

    # extract the user info and append it to the payload tuple for the given
    # client
    for client, c_config in config.iteritems():
        if not c_config:
            user_info = None
        else:
            user_info = extract_user_info(c_config)

        (region, zone, zone_info) = role_zones[client]
        role_zones[client] = (region, zone, zone_info, user_info)

    region_info = dict([
        (region_name, extract_region_info(region_name, r_config))
        for region_name, r_config in regions.iteritems()])

    fill_in_endpoints(region_info, role_zones, role_endpoints)

    # clear out the old defaults
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()
    # removing these objects from .rgw.root and the per-zone root pools
    # may or may not matter
    rados(ctx, mon,
          cmd=['-p', '.rgw.root', 'rm', 'region_info.default'])
    rados(ctx, mon,
          cmd=['-p', '.rgw.root', 'rm', 'zone_info.default'])

    for client in config.iterkeys():
        for role, (_, zone, zone_info, user_info) in role_zones.iteritems():
            rados(ctx, mon,
                  cmd=['-p', zone_info['domain_root'],
                       'rm', 'region_info.default'])
            rados(ctx, mon,
                  cmd=['-p', zone_info['domain_root'],
                       'rm', 'zone_info.default'])

            (remote,) = ctx.cluster.only(role).remotes.keys()
            for pool_info in zone_info['placement_pools']:
                remote.run(args=['ceph', 'osd', 'pool', 'create',
                                 pool_info['val']['index_pool'], '64', '64'])
                if ctx.rgw.ec_data_pool:
                    create_ec_pool(remote, pool_info['val']['data_pool'],
                                   zone, 64, ctx.rgw.erasure_code_profile)
                else:
                    create_replicated_pool(
                        remote, pool_info['val']['data_pool'],
                        64)

            rgwadmin(ctx, client,
                     cmd=['-n', client, 'zone', 'set', '--rgw-zone', zone],
                     stdin=StringIO(json.dumps(dict(
                         zone_info.items() + user_info.items()))),
                     check_status=True)

        for region, info in region_info.iteritems():
            region_json = json.dumps(info)
            log.debug('region info is: %s', region_json)
            rgwadmin(ctx, client,
                     cmd=['-n', client, 'region', 'set'],
                     stdin=StringIO(region_json),
                     check_status=True)
            if info['is_master']:
                rgwadmin(ctx, client,
                         cmd=['-n', client,
                              'region', 'default',
                              '--rgw-region', region],
                         check_status=True)

        rgwadmin(ctx, client, cmd=['-n', client, 'regionmap', 'update'])
    yield

def task(ctx, config):
    """
    Test handling of divergent entries with prior_version
    prior to log_tail

    overrides:
      ceph:
        conf:
          osd:
            debug osd: 5

    Requires 3 osds on a single test node.
    """
    if config is None:
        config = {}
    assert isinstance(config, dict), \
        'divergent_priors task only accepts a dict for configuration'

    manager = ctx.managers['ceph']

    while len(manager.get_osd_status()['up']) < 3:
        time.sleep(10)
    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.1', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.raw_cluster_cmd('osd', 'set', 'noout')
    manager.raw_cluster_cmd('osd', 'set', 'noin')
    manager.raw_cluster_cmd('osd', 'set', 'nodown')
    manager.wait_for_clean()

    # something that is always there
    dummyfile = '/etc/fstab'
    dummyfile2 = '/etc/resolv.conf'

    # create 1 pg pool
    log.info('creating foo')
    manager.raw_cluster_cmd('osd', 'pool', 'create', 'foo', '1')

    osds = [0, 1, 2]
    for i in osds:
        manager.set_config(i, osd_min_pg_log_entries=10)
        manager.set_config(i, osd_max_pg_log_entries=10)
        manager.set_config(i, osd_pg_log_trim_min=5)

    # determine primary
    divergent = manager.get_pg_primary('foo', 0)
    log.info("primary and soon to be divergent is %d", divergent)
    non_divergent = list(osds)
    non_divergent.remove(divergent)

    log.info('writing initial objects')
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()
    # write 100 objects
    for i in range(100):
        rados(ctx, mon, ['-p', 'foo', 'put', 'existing_%d' % i, dummyfile])

    manager.wait_for_clean()

    # blackhole non_divergent
    log.info("blackholing osds %s", str(non_divergent))
    for i in non_divergent:
        manager.set_config(i, filestore_blackhole=1)

    DIVERGENT_WRITE = 5
    DIVERGENT_REMOVE = 5
    # Write some soon to be divergent
    log.info('writing divergent objects')
    for i in range(DIVERGENT_WRITE):
        rados(ctx, mon, ['-p', 'foo', 'put', 'existing_%d' % i,
                         dummyfile2], wait=False)
    # Remove some soon to be divergent
    log.info('remove divergent objects')
    for i in range(DIVERGENT_REMOVE):
        rados(ctx, mon, ['-p', 'foo', 'rm',
                         'existing_%d' % (i + DIVERGENT_WRITE)], wait=False)
    time.sleep(10)
    mon.run(
        args=['killall', '-9', 'rados'],
        wait=True,
        check_status=False)

    # kill all the osds but leave divergent in
    log.info('killing all the osds')
    for i in osds:
        manager.kill_osd(i)
    for i in osds:
        manager.mark_down_osd(i)
    for i in non_divergent:
        manager.mark_out_osd(i)

    # bring up non-divergent
    log.info("bringing up non_divergent %s", str(non_divergent))
    for i in non_divergent:
        manager.revive_osd(i)
    for i in non_divergent:
        manager.mark_in_osd(i)

    # write 1 non-divergent object (ensure that old divergent one is divergent)
    objname = "existing_%d" % (DIVERGENT_WRITE + DIVERGENT_REMOVE)
#.........这里部分代码省略.........

def task(ctx, config):
    """
    Test handling of lost objects on an ec pool.

    A pretty rigid cluster is brought up andtested by this task
    """
    if config is None:
        config = {}
    assert isinstance(config, dict), \
        'lost_unfound task only accepts a dict for configuration'
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()

    manager = ceph_manager.CephManager(
        mon,
        ctx=ctx,
        logger=log.getChild('ceph_manager'),
        )

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.1', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.3', 'flush_pg_stats')
    manager.wait_for_clean()

    profile = config.get('erasure_code_profile', {
        'k': '2',
        'm': '2',
        'ruleset-failure-domain': 'osd'
    })
    profile_name = profile.get('name', 'lost_unfound')
    manager.create_erasure_code_profile(profile_name, profile)
    pool = manager.create_pool_with_unique_name(erasure_code_profile_name=profile_name)

    # something that is always there
    dummyfile = '/etc/fstab'

    # kludge to make sure they get a map
    rados(ctx, mon, ['-p', pool, 'put', 'dummy', dummyfile])

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.1', 'flush_pg_stats')
    manager.wait_for_recovery()

    # create old objects
    for f in range(1, 10):
        rados(ctx, mon, ['-p', pool, 'put', 'existing_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'put', 'existed_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'rm', 'existed_%d' % f])

    # delay recovery, and make the pg log very long (to prevent backfill)
    manager.raw_cluster_cmd(
            'tell', 'osd.1',
            'injectargs',
            '--osd-recovery-delay-start 1000 --osd-min-pg-log-entries 100000000'
            )

    manager.kill_osd(0)
    manager.mark_down_osd(0)
    manager.kill_osd(3)
    manager.mark_down_osd(3)
    
    for f in range(1, 10):
        rados(ctx, mon, ['-p', pool, 'put', 'new_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'put', 'existed_%d' % f, dummyfile])
        rados(ctx, mon, ['-p', pool, 'put', 'existing_%d' % f, dummyfile])

    # take out osd.1 and a necessary shard of those objects.
    manager.kill_osd(1)
    manager.mark_down_osd(1)
    manager.raw_cluster_cmd('osd', 'lost', '1', '--yes-i-really-mean-it')
    manager.revive_osd(0)
    manager.wait_till_osd_is_up(0)
    manager.revive_osd(3)
    manager.wait_till_osd_is_up(3)

    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.3', 'flush_pg_stats')
    manager.wait_till_active()
    manager.raw_cluster_cmd('tell', 'osd.0', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.2', 'flush_pg_stats')
    manager.raw_cluster_cmd('tell', 'osd.3', 'flush_pg_stats')

    # verify that there are unfound objects
    unfound = manager.get_num_unfound_objects()
    log.info("there are %d unfound objects" % unfound)
    assert unfound

    # mark stuff lost
    pgs = manager.get_pg_stats()
    for pg in pgs:
        if pg['stat_sum']['num_objects_unfound'] > 0:
            # verify that i can list them direct from the osd
            log.info('listing missing/lost in %s state %s', pg['pgid'],
                     pg['state']);
            m = manager.list_pg_missing(pg['pgid'])
            log.info('%s' % m)
            assert m['num_unfound'] == pg['stat_sum']['num_objects_unfound']

#.........这里部分代码省略.........

#.........这里部分代码省略.........
    EXP_ERRORS = 0
    for remote in osds.remotes.iterkeys():
        for role in osds.remotes[remote]:
            if string.find(role, "osd.") != 0:
                continue
            osdid = int(role.split('.')[1])

            for pg in pgs[osdid]:
                fpath = os.path.join(DATADIR, "osd{id}.{pg}".format(id=osdid, pg=pg))

                cmd = (prefix + "--op export --pgid {pg} --file {file}").format(id=osdid, pg=pg, file=fpath)
                proc = remote.run(args=cmd, check_status=False, stdout=StringIO())
                proc.wait()
                if proc.exitstatus != 0:
                    log.error("Exporting failed for pg {pg} on osd.{id} with {ret}".format(pg=pg, id=osdid, ret=proc.exitstatus))
                    EXP_ERRORS += 1

    ERRORS += EXP_ERRORS

    log.info("Test pg removal")
    RM_ERRORS = 0
    for remote in osds.remotes.iterkeys():
        for role in osds.remotes[remote]:
            if string.find(role, "osd.") != 0:
                continue
            osdid = int(role.split('.')[1])

            for pg in pgs[osdid]:
                cmd = (prefix + "--op remove --pgid {pg}").format(pg=pg, id=osdid)
                proc = remote.run(args=cmd, check_status=False, stdout=StringIO())
                proc.wait()
                if proc.exitstatus != 0:
                    log.error("Removing failed for pg {pg} on osd.{id} with {ret}".format(pg=pg, id=osdid, ret=proc.exitstatus))
                    RM_ERRORS += 1

    ERRORS += RM_ERRORS

    IMP_ERRORS = 0
    if EXP_ERRORS == 0 and RM_ERRORS == 0:
        log.info("Test pg import")

        for remote in osds.remotes.iterkeys():
            for role in osds.remotes[remote]:
                if string.find(role, "osd.") != 0:
                    continue
                osdid = int(role.split('.')[1])

                for pg in pgs[osdid]:
                    fpath = os.path.join(DATADIR, "osd{id}.{pg}".format(id=osdid, pg=pg))

                    cmd = (prefix + "--op import --file {file}").format(id=osdid, file=fpath)
                    proc = remote.run(args=cmd, check_status=False, stdout=StringIO())
                    proc.wait()
                    if proc.exitstatus != 0:
                        log.error("Import failed from {file} with {ret}".format(file=fpath, ret=proc.exitstatus))
                        IMP_ERRORS += 1
    else:
        log.warning("SKIPPING IMPORT TESTS DUE TO PREVIOUS FAILURES")

    ERRORS += IMP_ERRORS

    if EXP_ERRORS == 0 and RM_ERRORS == 0 and IMP_ERRORS == 0:
        log.info("Restarting OSDs....")
        # They are still look to be up because of setting nodown
        for osd in manager.get_osd_status()['up']:
            manager.revive_osd(osd)
        # Wait for health?
        time.sleep(5)
        # Let scrub after test runs verify consistency of all copies
        log.info("Verify replicated import data")
        objects = range(1, NUM_OBJECTS + 1)
        for i in objects:
            NAME = REP_NAME + "{num}".format(num=i)
            TESTNAME = os.path.join(DATADIR, "gettest")
            REFNAME = os.path.join(DATADIR, NAME)

            proc = rados(ctx, cli_remote, ['-p', REP_POOL, 'get', NAME, TESTNAME], wait=False)

            ret = proc.wait()
            if ret != 0:
               log.errors("After import, rados get failed with {ret}".format(ret=r[0].exitstatus))
               ERRORS += 1
               continue

            cmd = "diff -q {gettest} {ref}".format(gettest=TESTNAME, ref=REFNAME)
            proc = cli_remote.run(args=cmd, check_status=False)
            proc.wait()
            if proc.exitstatus != 0:
                log.error("Data comparison failed for {obj}".format(obj=NAME))
                ERRORS += 1

    if ERRORS == 0:
        log.info("TEST PASSED")
    else:
        log.error("TEST FAILED WITH {errcount} ERRORS".format(errcount=ERRORS))

    try:
        yield
    finally:
        log.info('Ending ceph_objectstore_tool')

def task(ctx, config):
    """
    Test handling of lost objects.

    A pretty rigid cluseter is brought up andtested by this task
    """
    POOL = "unfound_pool"
    if config is None:
        config = {}
    assert isinstance(config, dict), "lost_unfound task only accepts a dict for configuration"
    first_mon = teuthology.get_first_mon(ctx, config)
    (mon,) = ctx.cluster.only(first_mon).remotes.iterkeys()

    manager = ceph_manager.CephManager(mon, ctx=ctx, logger=log.getChild("ceph_manager"))

    while len(manager.get_osd_status()["up"]) < 3:
        time.sleep(10)
    manager.raw_cluster_cmd("tell", "osd.0", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.1", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.2", "flush_pg_stats")
    manager.wait_for_clean()

    manager.create_pool(POOL)

    # something that is always there
    dummyfile = "/etc/fstab"

    # take an osd out until the very end
    manager.kill_osd(2)
    manager.mark_down_osd(2)
    manager.mark_out_osd(2)

    # kludge to make sure they get a map
    rados(ctx, mon, ["-p", POOL, "put", "dummy", dummyfile])

    manager.raw_cluster_cmd("tell", "osd.0", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.1", "flush_pg_stats")
    manager.wait_for_recovery()

    # create old objects
    for f in range(1, 10):
        rados(ctx, mon, ["-p", POOL, "put", "existing_%d" % f, dummyfile])
        rados(ctx, mon, ["-p", POOL, "put", "existed_%d" % f, dummyfile])
        rados(ctx, mon, ["-p", POOL, "rm", "existed_%d" % f])

    # delay recovery, and make the pg log very long (to prevent backfill)
    manager.raw_cluster_cmd(
        "tell", "osd.1", "injectargs", "--osd-recovery-delay-start 1000 --osd-min-pg-log-entries 100000000"
    )

    manager.kill_osd(0)
    manager.mark_down_osd(0)

    for f in range(1, 10):
        rados(ctx, mon, ["-p", POOL, "put", "new_%d" % f, dummyfile])
        rados(ctx, mon, ["-p", POOL, "put", "existed_%d" % f, dummyfile])
        rados(ctx, mon, ["-p", POOL, "put", "existing_%d" % f, dummyfile])

    # bring osd.0 back up, let it peer, but don't replicate the new
    # objects...
    log.info("osd.0 command_args is %s" % "foo")
    log.info(ctx.daemons.get_daemon("osd", 0).command_args)
    ctx.daemons.get_daemon("osd", 0).command_kwargs["args"].extend(["--osd-recovery-delay-start", "1000"])
    manager.revive_osd(0)
    manager.mark_in_osd(0)
    manager.wait_till_osd_is_up(0)

    manager.raw_cluster_cmd("tell", "osd.1", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.0", "flush_pg_stats")
    manager.wait_till_active()

    # take out osd.1 and the only copy of those objects.
    manager.kill_osd(1)
    manager.mark_down_osd(1)
    manager.mark_out_osd(1)
    manager.raw_cluster_cmd("osd", "lost", "1", "--yes-i-really-mean-it")

    # bring up osd.2 so that things would otherwise, in theory, recovery fully
    manager.revive_osd(2)
    manager.mark_in_osd(2)
    manager.wait_till_osd_is_up(2)

    manager.raw_cluster_cmd("tell", "osd.0", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.2", "flush_pg_stats")
    manager.wait_till_active()
    manager.raw_cluster_cmd("tell", "osd.0", "flush_pg_stats")
    manager.raw_cluster_cmd("tell", "osd.2", "flush_pg_stats")

    # verify that there are unfound objects
    unfound = manager.get_num_unfound_objects()
    log.info("there are %d unfound objects" % unfound)
    assert unfound

    testdir = teuthology.get_testdir(ctx)
    procs = []
    if config.get("parallel_bench", True):
        procs.app