[ceph.git] / ceph / qa / tasks / rados.py

"""
Rados modle-based integration tests
"""
import contextlib
import logging
import gevent
from teuthology import misc as teuthology


from teuthology.orchestra import run

log = logging.getLogger(__name__)

@contextlib.contextmanager
def task(ctx, config):
    """
    Run RadosModel-based integration tests.

    The config should be as follows::

        rados:
          clients: [client list]
          ops: <number of ops>
          objects: <number of objects to use>
          max_in_flight: <max number of operations in flight>
          object_size: <size of objects in bytes>
          min_stride_size: <minimum write stride size in bytes>
          max_stride_size: <maximum write stride size in bytes>
          op_weights: <dictionary mapping operation type to integer weight>
          runs: <number of times to run> - the pool is remade between runs
          ec_pool: use an ec pool
          erasure_code_profile: profile to use with the erasure coded pool
          fast_read: enable ec_pool's fast_read
          min_size: set the min_size of created pool
          pool_snaps: use pool snapshots instead of selfmanaged snapshots
	  write_fadvise_dontneed: write behavior like with LIBRADOS_OP_FLAG_FADVISE_DONTNEED.
	                          This mean data don't access in the near future.
				  Let osd backend don't keep data in cache.

    For example::

        tasks:
        - ceph:
        - rados:
            clients: [client.0]
            ops: 1000
            max_seconds: 0   # 0 for no limit
            objects: 25
            max_in_flight: 16
            object_size: 4000000
            min_stride_size: 1024
            max_stride_size: 4096
            op_weights:
              read: 20
              write: 10
              delete: 2
              snap_create: 3
              rollback: 2
              snap_remove: 0
            ec_pool: create an ec pool, defaults to False
            erasure_code_use_overwrites: test overwrites, default false
            erasure_code_profile:
              name: teuthologyprofile
              k: 2
              m: 1
              crush-failure-domain: osd
            pool_snaps: true
	    write_fadvise_dontneed: true
            runs: 10
        - interactive:

    Optionally, you can provide the pool name to run against:

        tasks:
        - ceph:
        - exec:
            client.0:
              - ceph osd pool create foo
        - rados:
            clients: [client.0]
            pools: [foo]
            ...

    Alternatively, you can provide a pool prefix:

        tasks:
        - ceph:
        - exec:
            client.0:
              - ceph osd pool create foo.client.0
        - rados:
            clients: [client.0]
            pool_prefix: foo
            ...

    The tests are run asynchronously, they are not complete when the task
    returns. For instance:

        - rados:
            clients: [client.0]
            pools: [ecbase]
            ops: 4000
            objects: 500
            op_weights:
              read: 100
              write: 100
              delete: 50
              copy_from: 50
        - print: "**** done rados ec-cache-agent (part 2)"

     will run the print task immediately after the rados tasks begins but
     not after it completes. To make the rados task a blocking / sequential
     task, use:

        - sequential:
          - rados:
              clients: [client.0]
              pools: [ecbase]
              ops: 4000
              objects: 500
              op_weights:
                read: 100
                write: 100
                delete: 50
                copy_from: 50
        - print: "**** done rados ec-cache-agent (part 2)"

    """
    log.info('Beginning rados...')
    assert isinstance(config, dict), \
        "please list clients to run on"

    object_size = int(config.get('object_size', 4000000))
    op_weights = config.get('op_weights', {})
    testdir = teuthology.get_testdir(ctx)
    args = [
        'adjust-ulimits',
        'ceph-coverage',
        '{tdir}/archive/coverage'.format(tdir=testdir),
        'ceph_test_rados']
    if config.get('ec_pool', False):
        args.extend(['--no-omap'])
        if not config.get('erasure_code_use_overwrites', False):
            args.extend(['--ec-pool'])
    if config.get('write_fadvise_dontneed', False):
        args.extend(['--write-fadvise-dontneed'])
    if config.get('set_redirect', False):
        args.extend(['--set_redirect'])
    if config.get('set_chunk', False):
        args.extend(['--set_chunk'])
    if config.get('enable_dedup', False):
        args.extend(['--enable_dedup'])
    if config.get('low_tier_pool', None):
        args.extend(['--low_tier_pool', config.get('low_tier_pool', None)])
    if config.get('pool_snaps', False):
        args.extend(['--pool-snaps'])
    if config.get('balance_reads', False):
        args.extend(['--balance-reads'])
    if config.get('localize_reads', False):
        args.extend(['--localize-reads'])
    args.extend([
        '--max-ops', str(config.get('ops', 10000)),
        '--objects', str(config.get('objects', 500)),
        '--max-in-flight', str(config.get('max_in_flight', 16)),
        '--size', str(object_size),
        '--min-stride-size', str(config.get('min_stride_size', object_size // 10)),
        '--max-stride-size', str(config.get('max_stride_size', object_size // 5)),
        '--max-seconds', str(config.get('max_seconds', 0))
        ])

    weights = {}
    weights['read'] = 100
    weights['write'] = 100
    weights['delete'] = 10
    # Parallel of the op_types in test/osd/TestRados.cc
    for field in [
        # read handled above
        # write handled above
        # delete handled above
        "snap_create",
        "snap_remove",
        "rollback",
        "setattr",
        "rmattr",
        "watch",
        "copy_from",
        "hit_set_list",
        "is_dirty",
        "undirty",
        "cache_flush",
        "cache_try_flush",
        "cache_evict",
        "append",
        "write",
        "read",
        "delete"
        ]:
        if field in op_weights:
            weights[field] = op_weights[field]

    if config.get('write_append_excl', True):
        if 'write' in weights:
            weights['write'] = weights['write'] // 2
            weights['write_excl'] = weights['write']

        if 'append' in weights:
            weights['append'] = weights['append'] // 2
            weights['append_excl'] = weights['append']

    for op, weight in weights.items():
        args.extend([
            '--op', op, str(weight)
        ])
                

    def thread():
        """Thread spawned by gevent"""
        clients = ['client.{id}'.format(id=id_) for id_ in teuthology.all_roles_of_type(ctx.cluster, 'client')]
        log.info('clients are %s' % clients)
        manager = ctx.managers['ceph']
        if config.get('ec_pool', False):
            profile = config.get('erasure_code_profile', {})
            profile_name = profile.get('name', 'teuthologyprofile')
            manager.create_erasure_code_profile(profile_name, profile)
        else:
            profile_name = None
        for i in range(int(config.get('runs', '1'))):
            log.info("starting run %s out of %s", str(i), config.get('runs', '1'))
            tests = {}
            existing_pools = config.get('pools', [])
            created_pools = []
            for role in config.get('clients', clients):
                assert isinstance(role, str)
                PREFIX = 'client.'
                assert role.startswith(PREFIX)
                id_ = role[len(PREFIX):]

                pool = config.get('pool', None)
                if not pool and existing_pools:
                    pool = existing_pools.pop()
                else:
                    pool = manager.create_pool_with_unique_name(
                        erasure_code_profile_name=profile_name,
                        erasure_code_use_overwrites=
                          config.get('erasure_code_use_overwrites', False)
                    )
                    created_pools.append(pool)
                    if config.get('fast_read', False):
                        manager.raw_cluster_cmd(
                            'osd', 'pool', 'set', pool, 'fast_read', 'true')
                    min_size = config.get('min_size', None);
                    if min_size is not None:
                        manager.raw_cluster_cmd(
                            'osd', 'pool', 'set', pool, 'min_size', str(min_size))

                (remote,) = ctx.cluster.only(role).remotes.keys()
                proc = remote.run(
                    args=["CEPH_CLIENT_ID={id_}".format(id_=id_)] + args +
                    ["--pool", pool],
                    logger=log.getChild("rados.{id}".format(id=id_)),
                    stdin=run.PIPE,
                    wait=False
                    )
                tests[id_] = proc
            run.wait(tests.values())

            for pool in created_pools:
                manager.wait_snap_trimming_complete(pool);
                manager.remove_pool(pool)

    running = gevent.spawn(thread)

    try:
        yield
    finally:
        log.info('joining rados')
        running.get()
Commit	Line	Data
7c673cae FG	1	"""
	2	Rados modle-based integration tests
	3	"""
	4	import contextlib
	5	import logging
	6	import gevent
	7	from teuthology import misc as teuthology
	8
9f95a23c	9
7c673cae FG	10	from teuthology.orchestra import run
	11
	12	log = logging.getLogger(__name__)
	13
	14	@contextlib.contextmanager
	15	def task(ctx, config):
	16	"""
	17	Run RadosModel-based integration tests.
	18
	19	The config should be as follows::
	20
	21	rados:
	22	clients: [client list]
	23	ops: <number of ops>
	24	objects: <number of objects to use>
	25	max_in_flight: <max number of operations in flight>
	26	object_size: <size of objects in bytes>
	27	min_stride_size: <minimum write stride size in bytes>
	28	max_stride_size: <maximum write stride size in bytes>
	29	op_weights: <dictionary mapping operation type to integer weight>
	30	runs: <number of times to run> - the pool is remade between runs
	31	ec_pool: use an ec pool
	32	erasure_code_profile: profile to use with the erasure coded pool
	33	fast_read: enable ec_pool's fast_read
	34	min_size: set the min_size of created pool
	35	pool_snaps: use pool snapshots instead of selfmanaged snapshots
	36	write_fadvise_dontneed: write behavior like with LIBRADOS_OP_FLAG_FADVISE_DONTNEED.
	37	This mean data don't access in the near future.
	38	Let osd backend don't keep data in cache.
	39
	40	For example::
	41
	42	tasks:
	43	- ceph:
	44	- rados:
	45	clients: [client.0]
	46	ops: 1000
	47	max_seconds: 0 # 0 for no limit
	48	objects: 25
	49	max_in_flight: 16
	50	object_size: 4000000
	51	min_stride_size: 1024
	52	max_stride_size: 4096
	53	op_weights:
	54	read: 20
	55	write: 10
	56	delete: 2
	57	snap_create: 3
	58	rollback: 2
	59	snap_remove: 0
	60	ec_pool: create an ec pool, defaults to False
	61	erasure_code_use_overwrites: test overwrites, default false
	62	erasure_code_profile:
	63	name: teuthologyprofile
	64	k: 2
	65	m: 1
224ce89b	66	crush-failure-domain: osd
7c673cae FG	67	pool_snaps: true
	68	write_fadvise_dontneed: true
	69	runs: 10
	70	- interactive:
	71
	72	Optionally, you can provide the pool name to run against:
	73
	74	tasks:
	75	- ceph:
	76	- exec:
	77	client.0:
	78	- ceph osd pool create foo
	79	- rados:
	80	clients: [client.0]
	81	pools: [foo]
	82	...
	83
	84	Alternatively, you can provide a pool prefix:
	85
	86	tasks:
	87	- ceph:
	88	- exec:
	89	client.0:
	90	- ceph osd pool create foo.client.0
	91	- rados:
	92	clients: [client.0]
	93	pool_prefix: foo
	94	...
	95
	96	The tests are run asynchronously, they are not complete when the task
	97	returns. For instance:
	98
	99	- rados:
	100	clients: [client.0]
	101	pools: [ecbase]
	102	ops: 4000
	103	objects: 500
	104	op_weights:
	105	read: 100
	106	write: 100
	107	delete: 50
	108	copy_from: 50
	109	- print: "**** done rados ec-cache-agent (part 2)"
	110
	111	will run the print task immediately after the rados tasks begins but
	112	not after it completes. To make the rados task a blocking / sequential
	113	task, use:
	114
	115	- sequential:
	116	- rados:
	117	clients: [client.0]
	118	pools: [ecbase]
	119	ops: 4000
	120	objects: 500
	121	op_weights:
	122	read: 100
	123	write: 100
	124	delete: 50
	125	copy_from: 50
	126	- print: "**** done rados ec-cache-agent (part 2)"
	127
	128	"""
	129	log.info('Beginning rados...')
	130	assert isinstance(config, dict), \
131	"please list clients to run on"
132
133	object_size = int(config.get('object_size', 4000000))
134	op_weights = config.get('op_weights', {})
135	testdir = teuthology.get_testdir(ctx)
136	args = [
137	'adjust-ulimits',
138	'ceph-coverage',
139	'{tdir}/archive/coverage'.format(tdir=testdir),
140	'ceph_test_rados']
141	if config.get('ec_pool', False):
142	args.extend(['--no-omap'])
143	if not config.get('erasure_code_use_overwrites', False):
144	args.extend(['--ec-pool'])
145	if config.get('write_fadvise_dontneed', False):
146	args.extend(['--write-fadvise-dontneed'])
31f18b77 FG	147	if config.get('set_redirect', False):
31f18b77 FG	148	args.extend(['--set_redirect'])
11fdf7f2 TL	149	if config.get('set_chunk', False):
11fdf7f2 TL	150	args.extend(['--set_chunk'])
9f95a23c TL	151	if config.get('enable_dedup', False):
9f95a23c TL	152	args.extend(['--enable_dedup'])
11fdf7f2 TL	153	if config.get('low_tier_pool', None):
11fdf7f2 TL	154	args.extend(['--low_tier_pool', config.get('low_tier_pool', None)])
7c673cae FG	155	if config.get('pool_snaps', False):
7c673cae FG	156	args.extend(['--pool-snaps'])
9f95a23c TL	157	if config.get('balance_reads', False):
	158	args.extend(['--balance-reads'])
	159	if config.get('localize_reads', False):
	160	args.extend(['--localize-reads'])
7c673cae FG	161	args.extend([
	162	'--max-ops', str(config.get('ops', 10000)),
	163	'--objects', str(config.get('objects', 500)),
	164	'--max-in-flight', str(config.get('max_in_flight', 16)),
	165	'--size', str(object_size),
e306af50 TL	166	'--min-stride-size', str(config.get('min_stride_size', object_size // 10)),
e306af50 TL	167	'--max-stride-size', str(config.get('max_stride_size', object_size // 5)),
7c673cae FG	168	'--max-seconds', str(config.get('max_seconds', 0))
	169	])
	170
	171	weights = {}
	172	weights['read'] = 100
	173	weights['write'] = 100
	174	weights['delete'] = 10
	175	# Parallel of the op_types in test/osd/TestRados.cc
	176	for field in [
	177	# read handled above
	178	# write handled above
	179	# delete handled above
	180	"snap_create",
	181	"snap_remove",
	182	"rollback",
	183	"setattr",
	184	"rmattr",
	185	"watch",
	186	"copy_from",
	187	"hit_set_list",
	188	"is_dirty",
	189	"undirty",
	190	"cache_flush",
	191	"cache_try_flush",
	192	"cache_evict",
	193	"append",
	194	"write",
	195	"read",
	196	"delete"
	197	]:
	198	if field in op_weights:
	199	weights[field] = op_weights[field]
	200
	201	if config.get('write_append_excl', True):
	202	if 'write' in weights:
e306af50	203	weights['write'] = weights['write'] // 2
7c673cae FG	204	weights['write_excl'] = weights['write']
	205
	206	if 'append' in weights:
e306af50	207	weights['append'] = weights['append'] // 2
7c673cae FG	208	weights['append_excl'] = weights['append']
7c673cae FG	209
9f95a23c	210	for op, weight in weights.items():
7c673cae FG	211	args.extend([
	212	'--op', op, str(weight)
	213	])
	214
	215
	216	def thread():
	217	"""Thread spawned by gevent"""
	218	clients = ['client.{id}'.format(id=id_) for id_ in teuthology.all_roles_of_type(ctx.cluster, 'client')]
	219	log.info('clients are %s' % clients)
	220	manager = ctx.managers['ceph']
	221	if config.get('ec_pool', False):
	222	profile = config.get('erasure_code_profile', {})
	223	profile_name = profile.get('name', 'teuthologyprofile')
	224	manager.create_erasure_code_profile(profile_name, profile)
	225	else:
	226	profile_name = None
	227	for i in range(int(config.get('runs', '1'))):
	228	log.info("starting run %s out of %s", str(i), config.get('runs', '1'))
	229	tests = {}
	230	existing_pools = config.get('pools', [])
	231	created_pools = []
	232	for role in config.get('clients', clients):
f67539c2	233	assert isinstance(role, str)
7c673cae FG	234	PREFIX = 'client.'
	235	assert role.startswith(PREFIX)
	236	id_ = role[len(PREFIX):]
	237
	238	pool = config.get('pool', None)
	239	if not pool and existing_pools:
	240	pool = existing_pools.pop()
	241	else:
	242	pool = manager.create_pool_with_unique_name(
	243	erasure_code_profile_name=profile_name,
	244	erasure_code_use_overwrites=
	245	config.get('erasure_code_use_overwrites', False)
	246	)
	247	created_pools.append(pool)
	248	if config.get('fast_read', False):
	249	manager.raw_cluster_cmd(
	250	'osd', 'pool', 'set', pool, 'fast_read', 'true')
	251	min_size = config.get('min_size', None);
	252	if min_size is not None:
	253	manager.raw_cluster_cmd(
	254	'osd', 'pool', 'set', pool, 'min_size', str(min_size))
	255
9f95a23c	256	(remote,) = ctx.cluster.only(role).remotes.keys()
7c673cae FG	257	proc = remote.run(
	258	args=["CEPH_CLIENT_ID={id_}".format(id_=id_)] + args +
	259	["--pool", pool],
	260	logger=log.getChild("rados.{id}".format(id=id_)),
	261	stdin=run.PIPE,
	262	wait=False
	263	)
	264	tests[id_] = proc
9f95a23c	265	run.wait(tests.values())
7c673cae FG	266
	267	for pool in created_pools:
	268	manager.wait_snap_trimming_complete(pool);
	269	manager.remove_pool(pool)
	270
	271	running = gevent.spawn(thread)
	272
	273	try:
	274	yield
	275	finally:
	276	log.info('joining rados')
	277	running.get()