[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('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'])
    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.iteritems():
        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, basestring)
                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.iterkeys()
                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.itervalues())

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