def __init__(self, settings, workload, prefix=None):
super(Latency, self).__init__(settings)
self.clients = []
for bucket in self.get_buckets():
client = CBGen(bucket=bucket,
host=settings.master_node,
username=bucket,
password=settings.bucket_password)
self.clients.append((bucket, client))
self.existing_keys = ExistingKey(workload.working_set,
workload.working_set_access,
prefix=prefix)
self.new_keys = NewKey(prefix=prefix, expiration=workload.expiration)
self.keys_for_removal = KeyForRemoval(prefix=prefix)
if not hasattr(workload, 'doc_gen') or workload.doc_gen == 'old':
self.new_docs = NewDocument(workload.size)
elif workload.doc_gen == 'new':
self.new_docs = NewNestedDocument(workload.size)
elif workload.doc_gen == 'reverse_lookup':
self.new_docs = ReverseLookupDocument(workload.size,
workload.doc_partitions)
elif workload.doc_gen == 'reverse_lookup_array_indexing':
self.new_docs = ReverseLookupDocumentArrayIndexing(
workload.size, workload.doc_partitions, workload.items)
self.items = workload.items
self.n1ql_op = workload.n1ql_op
def __init__(self, settings, workload, prefix=None):
super(Latency, self).__init__(settings)
self.clients = []
for bucket in self.get_buckets():
client = CBGen(bucket=bucket, host=settings.master_node,
username=bucket, password=settings.bucket_password)
self.clients.append((bucket, client))
self.existing_keys = ExistingKey(workload.working_set,
workload.working_set_access,
prefix=prefix)
self.new_keys = NewKey(prefix=prefix, expiration=workload.expiration)
self.keys_for_removal = KeyForRemoval(prefix=prefix)
if not hasattr(workload, 'doc_gen') or workload.doc_gen == 'old':
self.new_docs = NewDocument(workload.size)
elif workload.doc_gen == 'new':
self.new_docs = NewNestedDocument(workload.size)
elif workload.doc_gen == 'reverse_lookup':
self.new_docs = ReverseLookupDocument(workload.size,
workload.doc_partitions)
elif workload.doc_gen == 'reverse_lookup_array_indexing':
self.new_docs = ReverseLookupDocumentArrayIndexing(
workload.size, workload.doc_partitions, workload.items)
self.items = workload.items
self.n1ql_op = workload.n1ql_op
def __init__(self, workload_settings, target_settings, shutdown_event):
super(N1QLWorker, self).__init__(workload_settings, target_settings,
shutdown_event)
self.new_queries = N1QLQueryGen(workload_settings.n1ql_queries)
self.total_workers = self.ws.n1ql_workers
self.throughput = self.ws.n1ql_throughput
self.name = 'n1ql-worker'
host, port = self.ts.node.split(':')
params = {'bucket': self.ts.bucket, 'host': host, 'port': port,
'username': self.ts.bucket, 'password': self.ts.password}
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access,
'n1ql')
self.new_keys = NewKey('n1ql', self.ws.expiration)
self.keys_for_removal = KeyForRemoval('n1ql')
self.keys_for_casupdate = KeyForCASUpdate(self.total_workers, self.ws.working_set,
self.ws.working_set_access,
'n1ql')
if self.ws.doc_gen == 'merge':
self.docs = MergeDocument(self.ws.size,
self.ws.doc_partitions,
False)
elif self.ws.doc_gen == 'reverse_lookup':
self.docs = ReverseLookupDocument(self.ws.size,
self.ws.doc_partitions,
False)
self.cb = N1QLGen(**params)
def init_keys(self):
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access,
prefix='n1ql')
self.new_keys = NewKey(prefix='n1ql', expiration=self.ws.expiration)
self.keys_for_casupdate = KeyForCASUpdate(self.total_workers,
self.ws.working_set,
self.ws.working_set_access,
prefix='n1ql')
class SpringLatency(Latency):
COLLECTOR = "spring_latency"
METRICS = ("latency_set", "latency_get")
def __init__(self, settings, workload, prefix=None):
super(Latency, self).__init__(settings)
self.clients = []
for bucket in self.get_buckets():
client = CBGen(bucket=bucket,
host=settings.master_node,
username=bucket,
password=settings.bucket_password)
self.clients.append((bucket, client))
self.existing_keys = ExistingKey(workload.working_set,
workload.working_set_access,
prefix=prefix)
self.new_keys = NewKey(prefix=prefix, expiration=workload.expiration)
self.keys_for_removal = KeyForRemoval(prefix=prefix)
if not hasattr(workload, 'doc_gen') or workload.doc_gen == 'old':
self.new_docs = NewDocument(workload.size)
elif workload.doc_gen == 'new':
self.new_docs = NewNestedDocument(workload.size)
elif workload.doc_gen == 'reverse_lookup':
self.new_docs = ReverseLookupDocument(workload.size,
workload.doc_partitions)
elif workload.doc_gen == 'reverse_lookup_array_indexing':
self.new_docs = ReverseLookupDocumentArrayIndexing(
workload.size, workload.doc_partitions, workload.items)
self.items = workload.items
self.n1ql_op = workload.n1ql_op
def measure(self, client, metric, bucket):
key = self.existing_keys.next(curr_items=self.items, curr_deletes=0)
doc = self.new_docs.next(key)
t0 = time()
if metric == "latency_set":
client.create(key, doc)
elif metric == "latency_get":
client.read(key)
elif metric == "latency_cas":
client.cas(key, doc)
return 1000 * (time() - t0) # Latency in ms
def sample(self):
for bucket, client in self.clients:
samples = {}
for metric in self.METRICS:
samples[metric] = self.measure(client, metric, bucket)
self.store.append(samples,
cluster=self.cluster,
bucket=bucket,
collector=self.COLLECTOR)
def __init__(self, workload_settings, target_settings, shutdown_event):
super(N1QLWorker, self).__init__(workload_settings, target_settings,
shutdown_event)
self.new_queries = N1QLQueryGen(workload_settings.n1ql_queries)
self.total_workers = self.ws.n1ql_workers
self.throughput = self.ws.n1ql_throughput
self.name = 'n1ql-worker'
host, port = self.ts.node.split(':')
bucket = self.ts.bucket
if workload_settings.n1ql_op == 'ryow':
bucket += '?fetch_mutation_tokens=true'
params = {'bucket': bucket, 'host': host, 'port': port,
'username': self.ts.bucket, 'password': self.ts.password}
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access,
'n1ql')
self.new_keys = NewKey('n1ql', self.ws.expiration)
self.keys_for_removal = KeyForRemoval('n1ql')
self.keys_for_casupdate = KeyForCASUpdate(self.total_workers, self.ws.working_set,
self.ws.working_set_access,
'n1ql')
if self.ws.doc_gen == 'merge':
self.docs = MergeDocument(self.ws.size,
self.ws.doc_partitions,
False)
elif self.ws.doc_gen == 'reverse_lookup':
self.docs = ReverseLookupDocument(self.ws.size,
self.ws.doc_partitions,
False)
elif self.ws.doc_gen == 'reverse_lookup_array_indexing':
if self.ws.updates:
self.docs = ReverseLookupDocumentArrayIndexing(
self.ws.size, self.ws.doc_partitions, self.ws.items,
delta=random.randint(0, 10))
else:
self.docs = ReverseLookupDocumentArrayIndexing(
self.ws.size, self.ws.doc_partitions, self.ws.items)
self.cb = N1QLGen(**params)
class SpringLatency(Latency):
COLLECTOR = "spring_latency"
METRICS = ("latency_set", "latency_get")
def __init__(self, settings, workload, prefix=None):
super(Latency, self).__init__(settings)
self.clients = []
for bucket in self.get_buckets():
client = CBGen(bucket=bucket, host=settings.master_node,
username=bucket, password=settings.bucket_password)
self.clients.append((bucket, client))
self.existing_keys = ExistingKey(workload.working_set,
workload.working_set_access,
prefix=prefix)
self.new_keys = NewKey(prefix=prefix, expiration=workload.expiration)
self.keys_for_removal = KeyForRemoval(prefix=prefix)
if not hasattr(workload, 'doc_gen') or workload.doc_gen == 'old':
self.new_docs = NewDocument(workload.size)
elif workload.doc_gen == 'new':
self.new_docs = NewNestedDocument(workload.size)
elif workload.doc_gen == 'reverse_lookup':
self.new_docs = ReverseLookupDocument(workload.size,
workload.doc_partitions)
elif workload.doc_gen == 'reverse_lookup_array_indexing':
self.new_docs = ReverseLookupDocumentArrayIndexing(
workload.size, workload.doc_partitions, workload.items)
self.items = workload.items
self.n1ql_op = workload.n1ql_op
def measure(self, client, metric, bucket):
key = self.existing_keys.next(curr_items=self.items, curr_deletes=0)
doc = self.new_docs.next(key)
t0 = time()
if metric == "latency_set":
client.create(key, doc)
elif metric == "latency_get":
client.read(key)
elif metric == "latency_cas":
client.cas(key, doc)
return 1000 * (time() - t0) # Latency in ms
def sample(self):
for bucket, client in self.clients:
samples = {}
for metric in self.METRICS:
samples[metric] = self.measure(client, metric, bucket)
self.store.append(samples, cluster=self.cluster,
bucket=bucket, collector=self.COLLECTOR)
def __init__(self, settings, workload, prefix=None):
super(Latency, self).__init__(settings)
self.clients = []
for bucket in self.get_buckets():
client = CBGen(bucket=bucket, host=settings.master_node,
username=bucket, password=settings.bucket_password)
self.clients.append((bucket, client))
self.existing_keys = ExistingKey(workload.working_set,
workload.working_set_access,
prefix=prefix)
if not hasattr(workload, 'doc_gen') or workload.doc_gen == 'old':
self.new_docs = NewDocument(workload.size)
else:
self.new_docs = NewNestedDocument(workload.size)
self.items = workload.items
class SpringLatency(Latency):
COLLECTOR = "spring_latency"
METRICS = ("latency_set", "latency_get")
def __init__(self, settings, workload, prefix=None):
super(Latency, self).__init__(settings)
self.clients = []
for bucket in self.get_buckets():
client = CBGen(bucket=bucket, host=settings.master_node,
username=bucket, password=settings.bucket_password)
self.clients.append((bucket, client))
self.existing_keys = ExistingKey(workload.working_set,
workload.working_set_access,
prefix=prefix)
if not hasattr(workload, 'doc_gen') or workload.doc_gen == 'old':
self.new_docs = NewDocument(workload.size)
else:
self.new_docs = NewNestedDocument(workload.size)
self.items = workload.items
def measure(self, client, metric):
key = self.existing_keys.next(curr_items=self.items, curr_deletes=0)
doc = self.new_docs.next(key)
t0 = time()
if metric == "latency_set":
client.create(key, doc)
elif metric == "latency_get":
client.read(key)
elif metric == "latency_cas":
client.cas(key, doc)
return 1000 * (time() - t0) # Latency in ms
def sample(self):
for bucket, client in self.clients:
samples = {}
for metric in self.METRICS:
samples[metric] = self.measure(client, metric)
self.store.append(samples, cluster=self.cluster,
bucket=bucket, collector=self.COLLECTOR)
def __init__(self, settings, workload, prefix=None):
super(Latency, self).__init__(settings)
self.clients = []
for bucket in self.get_buckets():
client = CBGen(bucket=bucket, host=settings.master_node,
username=bucket, password=settings.bucket_password)
self.clients.append((bucket, client))
self.existing_keys = ExistingKey(workload.working_set,
workload.working_set_access,
prefix=prefix)
self.new_keys = NewKey(prefix=prefix, expiration=workload.expiration)
self.keys_for_removal = KeyForRemoval(prefix=prefix)
if not hasattr(workload, 'doc_gen') or workload.doc_gen == 'basic':
self.new_docs = Document(workload.size)
elif workload.doc_gen == 'nested':
self.new_docs = NestedDocument(workload.size)
self.items = workload.items
class N1QLWorker(Worker):
def __init__(self, workload_settings, target_settings, shutdown_event):
super(N1QLWorker, self).__init__(workload_settings, target_settings,
shutdown_event)
self.new_queries = N1QLQueryGen(workload_settings.n1ql_queries)
self.total_workers = self.ws.n1ql_workers
self.throughput = self.ws.n1ql_throughput
self.name = 'n1ql-worker'
host, port = self.ts.node.split(':')
bucket = self.ts.bucket
if workload_settings.n1ql_op == 'ryow':
bucket += '?fetch_mutation_tokens=true'
params = {'bucket': bucket, 'host': host, 'port': port,
'username': self.ts.bucket, 'password': self.ts.password}
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access,
'n1ql')
self.new_keys = NewKey('n1ql', self.ws.expiration)
self.keys_for_removal = KeyForRemoval('n1ql')
self.keys_for_casupdate = KeyForCASUpdate(self.total_workers, self.ws.working_set,
self.ws.working_set_access,
'n1ql')
if self.ws.doc_gen == 'merge':
self.docs = MergeDocument(self.ws.size,
self.ws.doc_partitions,
False)
elif self.ws.doc_gen == 'reverse_lookup':
self.docs = ReverseLookupDocument(self.ws.size,
self.ws.doc_partitions,
False)
elif self.ws.doc_gen == 'reverse_lookup_array_indexing':
if self.ws.updates:
self.docs = ReverseLookupDocumentArrayIndexing(
self.ws.size, self.ws.doc_partitions, self.ws.items,
delta=random.randint(0, 10))
else:
self.docs = ReverseLookupDocumentArrayIndexing(
self.ws.size, self.ws.doc_partitions, self.ws.items)
self.cb = N1QLGen(**params)
@with_sleep
def do_batch(self):
if self.ws.n1ql_op == 'read':
curr_items_spot = \
self.curr_items.value - self.ws.creates * self.ws.workers
deleted_spot = \
self.deleted_items.value + self.ws.deletes * self.ws.workers
for _ in xrange(self.BATCH_SIZE):
key = self.existing_keys.next(curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
return
curr_items_tmp = curr_items_spot = self.curr_items.value
if self.ws.n1ql_op == 'create':
with self.lock:
self.curr_items.value += self.BATCH_SIZE
curr_items_tmp = self.curr_items.value - self.BATCH_SIZE
curr_items_spot = (curr_items_tmp -
self.BATCH_SIZE * self.total_workers)
deleted_items_tmp = deleted_spot = 0
if self.ws.n1ql_op == 'delete':
with self.lock:
self.deleted_items.value += self.BATCH_SIZE
deleted_items_tmp = self.deleted_items.value - self.BATCH_SIZE
deleted_spot = (deleted_items_tmp +
self.BATCH_SIZE * self.total_workers)
deleted_capped_items_tmp = deleted_capped_spot = 0
if self.ws.n1ql_op == 'rangedelete':
with self.lock:
self.deleted_capped_items.value += self.BATCH_SIZE
deleted_capped_items_tmp = self.deleted_capped_items.value - self.BATCH_SIZE
deleted_capped_spot = (deleted_capped_items_tmp +
self.BATCH_SIZE * self.total_workers)
casupdated_items_tmp = casupdated_spot = 0
if self.ws.n1ql_op == 'update':
with self.lock:
self.casupdated_items.value += self.BATCH_SIZE
casupdated_items_tmp = self.casupdated_items.value - self.BATCH_SIZE
casupdated_spot = (casupdated_items_tmp +
self.BATCH_SIZE * self.total_workers)
if self.ws.n1ql_op == 'create':
for _ in xrange(self.BATCH_SIZE):
curr_items_tmp += 1
key, ttl = self.new_keys.next(curr_items_tmp)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'delete':
for _ in xrange(self.BATCH_SIZE):
deleted_items_tmp += 1
key = self.keys_for_removal.next(deleted_items_tmp)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'update' or self.ws.n1ql_op == 'lookupupdate':
for _ in xrange(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid, curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'ryow':
for _ in xrange(self.BATCH_SIZE):
query = self.ws.n1ql_queries[0]['statement'][1:-1]
if self.ws.n1ql_queries[0]['prepared'] == "singleton_unique_lookup":
by_key = 'email'
elif self.ws.n1ql_queries[0]['prepared'] == "range_scan":
by_key = 'capped_small'
else:
logger.error('n1ql_queries {} not defined'.format(self.ws.n1ql_queries))
key1 = self.keys_for_casupdate.next(self.sid, curr_items_spot, deleted_spot)
doc1 = self.docs.next(key1)
key2 = self.keys_for_casupdate.next(self.sid, curr_items_spot, deleted_spot)
doc2 = self.docs.next(key2)
rvs = self.cb.client.upsert_multi({key1: doc2, key2: doc1})
# This is a part of requirements:
# Each n1ql worker sleeps for 1 seconds.
time.sleep(float(self.ws.n1ql_queries[0]['time_sleep']))
ms = MutationState()
ms.add_results(*rvs.values())
nq = N1QLQuery(query.format(doc2[by_key]))
nq.consistent_with(ms)
len(list(self.cb.client.n1ql_query(nq)))
elif self.ws.n1ql_op == 'rangeupdate':
for _ in xrange(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid, curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'rangedelete':
for _ in xrange(self.BATCH_SIZE):
doc = {}
doc['capped_small'] = "n1ql-_100_" + str(deleted_capped_items_tmp)
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
deleted_capped_items_tmp += 1
elif self.ws.n1ql_op == 'merge': #run select * workload for merge
for _ in xrange(self.BATCH_SIZE):
key = self.existing_keys.next(curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
query['statement'] = "SELECT * FROM `bucket-1` USE KEYS[$1];"
query['args'] = "[\"{key}\"]".format(**doc)
del query['prepared']
self.cb.query(ddoc_name, view_name, query=query)
def run(self, sid, lock, curr_queries, curr_items, deleted_items,
casupdated_items, deleted_capped_items):
self.cb.start_updater()
if self.throughput < float('inf'):
self.target_time = float(self.BATCH_SIZE) * self.total_workers / \
self.throughput
else:
self.target_time = None
self.lock = lock
self.sid = sid
self.curr_items = curr_items
self.deleted_items = deleted_items
self.deleted_capped_items = deleted_capped_items
self.casupdated_items = casupdated_items
self.curr_queries = curr_queries
try:
logger.info('Started: {}-{}'.format(self.name, self.sid))
while curr_queries.value < self.ws.ops and not self.time_to_stop():
with self.lock:
curr_queries.value += self.BATCH_SIZE
self.do_batch()
self.report_progress(curr_queries.value)
except (KeyboardInterrupt, ValueFormatError, AttributeError) as e:
logger.info('Interrupted: {}-{}-{}'.format(self.name, self.sid, e))
else:
if self.fallingBehindCount > 0:
logger.info('Worker {0} fell behind {1} times.'.
format(self.name, self.fallingBehindCount))
logger.info('Finished: {}-{}'.format(self.name, self.sid))
def __init__(self, workload_settings, target_settings,
shutdown_event=None):
self.ws = workload_settings
self.ts = target_settings
self.shutdown_event = shutdown_event
logger.setLevel(logging.INFO)
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access,
self.ts.prefix)
self.new_keys = NewKey(self.ts.prefix, self.ws.expiration)
self.keys_for_removal = KeyForRemoval(self.ts.prefix)
if not hasattr(self.ws, 'doc_gen') or self.ws.doc_gen == 'old':
extra_fields = False
if (hasattr(self.ws, 'extra_doc_fields') and
self.ws['extra_doc_fields'] == 'yes'):
extra_fields = True
self.docs = NewDocument(self.ws.size, extra_fields)
elif self.ws.doc_gen == 'new':
self.docs = NewNestedDocument(self.ws.size)
elif self.ws.doc_gen == 'merge':
isRandom = True
if self.ts.prefix == 'n1ql':
isRandom = False
self.docs = MergeDocument(self.ws.size,
self.ws.doc_partitions,
isRandom)
elif self.ws.doc_gen == 'reverse_lookup':
isRandom = True
if self.ts.prefix == 'n1ql':
isRandom = False
self.docs = ReverseLookupDocument(self.ws.size,
self.ws.doc_partitions,
isRandom)
elif self.ws.doc_gen == 'reverse_lookup_array_indexing':
isRandom = True
if self.ts.prefix == 'n1ql':
isRandom = False
if self.ws.updates:
# plus 10 to all values in array when updating doc
self.docs = ReverseLookupDocumentArrayIndexing(
self.ws.size, self.ws.doc_partitions, self.ws.items,
delta=random.randint(0, 10))
else:
self.docs = ReverseLookupDocumentArrayIndexing(
self.ws.size, self.ws.doc_partitions, self.ws.items)
elif self.ws.doc_gen == 'spatial':
self.docs = NewDocumentFromSpatialFile(
self.ws.spatial.data,
self.ws.spatial.dimensionality)
elif self.ws.doc_gen == 'large_subdoc':
self.docs = NewLargeDocument(self.ws.size)
self.next_report = 0.05 # report after every 5% of completion
host, port = self.ts.node.split(':')
# Only FTS uses proxyPort and authless bucket right now.
# Instead of jumping hoops to specify proxyPort in target
# iterator/settings, which only passes down very specific attributes,
# just detect fts instead. The following does not work with
# authless bucket. FTS's worker does its own Couchbase.connect
if not (hasattr(self.ws, "fts") and hasattr(
self.ws.fts, "doc_database_url")):
# default sasl bucket
self.init_db({'bucket': self.ts.bucket, 'host': host, 'port': port,
'username': self.ts.bucket,
'password': self.ts.password})
self.fallingBehindCount = 0
class N1QLWorker(Worker):
NAME = 'n1ql-worker'
def __init__(self, workload_settings, target_settings, shutdown_event):
self.new_queries = N1QLQueryGen(workload_settings.n1ql_queries)
self.total_workers = workload_settings.n1ql_workers
self.throughput = workload_settings.n1ql_throughput
self.reservoir = Reservoir(num_workers=workload_settings.n1ql_workers)
super(N1QLWorker, self).__init__(workload_settings, target_settings,
shutdown_event)
def init_keys(self):
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access,
prefix='n1ql')
self.new_keys = NewKey(prefix='n1ql', expiration=self.ws.expiration)
self.keys_for_casupdate = KeyForCASUpdate(self.total_workers,
self.ws.working_set,
self.ws.working_set_access,
prefix='n1ql')
def init_docs(self):
if self.ws.doc_gen == 'reverse_lookup':
self.docs = ReverseLookupDocument(self.ws.size,
prefix='n1ql')
elif self.ws.doc_gen == 'reverse_range_lookup':
self.docs = ReverseRangeLookupDocument(self.ws.size,
prefix='n1ql',
range_distance=self.ws.range_distance)
elif self.ws.doc_gen == 'ext_reverse_lookup':
self.docs = ExtReverseLookupDocument(self.ws.size,
prefix='n1ql',
num_docs=self.ws.items)
elif self.ws.doc_gen == 'join':
self.docs = JoinedDocument(self.ws.size,
prefix='n1ql',
num_docs=self.ws.items,
num_categories=self.ws.num_categories,
num_replies=self.ws.num_replies)
elif self.ws.doc_gen == 'ref':
self.docs = RefDocument(self.ws.size,
prefix='n1ql')
elif self.ws.doc_gen == 'array_indexing':
self.docs = ArrayIndexingDocument(self.ws.size,
prefix='n1ql',
array_size=self.ws.array_size,
num_docs=self.ws.items)
def init_db(self):
host, port = self.ts.node.split(':')
self.cb = N1QLGen(bucket=self.ts.bucket, password=self.ts.password,
host=host, port=port)
def read(self):
curr_items_tmp = self.curr_items.value
if self.ws.doc_gen == 'ext_reverse_lookup':
curr_items_tmp /= 4
for _ in range(self.BATCH_SIZE):
key = self.existing_keys.next(curr_items=curr_items_tmp,
curr_deletes=0)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
query = self.new_queries.next(doc)
_, latency = self.cb.query(query)
self.reservoir.update(latency)
def create(self):
with self.lock:
self.curr_items.value += self.BATCH_SIZE
curr_items_tmp = self.curr_items.value - self.BATCH_SIZE
for _ in range(self.BATCH_SIZE):
curr_items_tmp += 1
key, ttl = self.new_keys.next(curr_items=curr_items_tmp)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
query = self.new_queries.next(doc)
_, latency = self.cb.query(query)
self.reservoir.update(latency)
def update(self):
with self.lock:
self.cas_updated_items.value += self.BATCH_SIZE
curr_items_tmp = self.curr_items.value - self.BATCH_SIZE
for _ in range(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid,
curr_items=curr_items_tmp,
curr_deletes=0)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
query = self.new_queries.next(doc)
_, latency = self.cb.query(query)
self.reservoir.update(latency)
def range_update(self):
with self.lock:
self.cas_updated_items.value += self.BATCH_SIZE
curr_items_tmp = self.curr_items.value - self.BATCH_SIZE
for _ in range(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid,
curr_items=curr_items_tmp,
curr_deletes=0)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
query = self.new_queries.next(doc)
_, latency = self.cb.query(query)
self.reservoir.update(latency)
@with_sleep
def do_batch(self):
if self.ws.n1ql_op == 'read':
self.read()
elif self.ws.n1ql_op == 'create':
self.create()
elif self.ws.n1ql_op == 'update':
self.update()
elif self.ws.n1ql_op == 'rangeupdate':
self.range_update()
def run(self, sid, lock, curr_queries, curr_items, deleted_items,
cas_updated_items):
if self.throughput < float('inf'):
self.target_time = float(self.BATCH_SIZE) * self.total_workers / \
self.throughput
else:
self.target_time = None
self.lock = lock
self.sid = sid
self.curr_items = curr_items
self.cas_updated_items = cas_updated_items
self.curr_queries = curr_queries
try:
logger.info('Started: {}-{}'.format(self.NAME, self.sid))
while curr_queries.value < self.ws.ops and not self.time_to_stop():
with self.lock:
curr_queries.value += self.BATCH_SIZE
self.do_batch()
except (KeyboardInterrupt, ValueFormatError, AttributeError) as e:
logger.info('Interrupted: {}-{}-{}'.format(self.NAME, self.sid, e))
else:
logger.info('Finished: {}-{}'.format(self.NAME, self.sid))
self.reservoir.dump(filename='{}-{}'.format(self.NAME, self.sid))
class N1QLWorker(Worker):
def __init__(self, workload_settings, target_settings, shutdown_event):
super(N1QLWorker, self).__init__(workload_settings, target_settings,
shutdown_event)
self.new_queries = N1QLQueryGen(workload_settings.n1ql_queries)
self.total_workers = self.ws.n1ql_workers
self.throughput = self.ws.n1ql_throughput
self.name = 'n1ql-worker'
host, port = self.ts.node.split(':')
params = {'bucket': self.ts.bucket, 'host': host, 'port': port,
'username': self.ts.bucket, 'password': self.ts.password}
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access,
'n1ql')
self.new_keys = NewKey('n1ql', self.ws.expiration)
self.keys_for_removal = KeyForRemoval('n1ql')
self.keys_for_casupdate = KeyForCASUpdate(self.total_workers, self.ws.working_set,
self.ws.working_set_access,
'n1ql')
if self.ws.doc_gen == 'merge':
self.docs = MergeDocument(self.ws.size,
self.ws.doc_partitions,
False)
elif self.ws.doc_gen == 'reverse_lookup':
self.docs = ReverseLookupDocument(self.ws.size,
self.ws.doc_partitions,
False)
self.cb = N1QLGen(**params)
@with_sleep
def do_batch(self):
if self.ws.n1ql_op == 'read':
curr_items_spot = \
self.curr_items.value - self.ws.creates * self.ws.workers
deleted_spot = \
self.deleted_items.value + self.ws.deletes * self.ws.workers
for _ in xrange(self.BATCH_SIZE):
key = self.existing_keys.next(curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
return
curr_items_tmp = curr_items_spot = self.curr_items.value
if self.ws.n1ql_op == 'create':
with self.lock:
self.curr_items.value += self.BATCH_SIZE
curr_items_tmp = self.curr_items.value - self.BATCH_SIZE
curr_items_spot = (curr_items_tmp -
self.BATCH_SIZE * self.total_workers)
deleted_items_tmp = deleted_spot = 0
if self.ws.n1ql_op == 'delete':
with self.lock:
self.deleted_items.value += self.BATCH_SIZE
deleted_items_tmp = self.deleted_items.value - self.BATCH_SIZE
deleted_spot = (deleted_items_tmp +
self.BATCH_SIZE * self.total_workers)
deleted_capped_items_tmp = deleted_capped_spot = 0
if self.ws.n1ql_op == 'rangedelete':
with self.lock:
self.deleted_capped_items.value += self.BATCH_SIZE
deleted_capped_items_tmp = self.deleted_capped_items.value - self.BATCH_SIZE
deleted_capped_spot = (deleted_capped_items_tmp +
self.BATCH_SIZE * self.total_workers)
casupdated_items_tmp = casupdated_spot = 0
if self.ws.n1ql_op == 'update':
with self.lock:
self.casupdated_items.value += self.BATCH_SIZE
casupdated_items_tmp = self.casupdated_items.value - self.BATCH_SIZE
casupdated_spot = (casupdated_items_tmp +
self.BATCH_SIZE * self.total_workers)
if self.ws.n1ql_op == 'create':
for _ in xrange(self.BATCH_SIZE):
curr_items_tmp += 1
key, ttl = self.new_keys.next(curr_items_tmp)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'delete':
for _ in xrange(self.BATCH_SIZE):
deleted_items_tmp += 1
key = self.keys_for_removal.next(deleted_items_tmp)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'update' or self.ws.n1ql_op == 'lookupupdate':
for _ in xrange(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid, curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'rangeupdate':
for _ in xrange(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid, curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'rangedelete':
for _ in xrange(self.BATCH_SIZE):
doc = {}
doc['capped_small'] = "n1ql-_100_" + str(deleted_capped_items_tmp)
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
deleted_capped_items_tmp += 1
elif self.ws.n1ql_op == 'merge': #run select * workload for merge
for _ in xrange(self.BATCH_SIZE):
key = self.existing_keys.next(curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
query['statement'] = "SELECT * FROM `bucket-1` USE KEYS[$1];"
query['args'] = "[\"{key}\"]".format(**doc)
del query['prepared']
self.cb.query(ddoc_name, view_name, query=query)
def run(self, sid, lock, curr_queries, curr_items, deleted_items,
casupdated_items, deleted_capped_items):
self.cb.start_updater()
if self.throughput < float('inf'):
self.target_time = float(self.BATCH_SIZE) * self.total_workers / \
self.throughput
else:
self.target_time = None
self.lock = lock
self.sid = sid
self.curr_items = curr_items
self.deleted_items = deleted_items
self.deleted_capped_items = deleted_capped_items
self.casupdated_items = casupdated_items
self.curr_queries = curr_queries
try:
logger.info('Started: {}-{}'.format(self.name, self.sid))
while curr_queries.value < self.ws.ops and not self.time_to_stop():
with self.lock:
curr_queries.value += self.BATCH_SIZE
self.do_batch()
self.report_progress(curr_queries.value)
except (KeyboardInterrupt, ValueFormatError, AttributeError) as e:
logger.info('Interrupted: {}-{}-{}'.format(self.name, self.sid, e))
else:
logger.info('Finished: {}-{}'.format(self.name, self.sid))
class N1QLWorker(Worker):
def __init__(self, workload_settings, target_settings, shutdown_event):
super(N1QLWorker, self).__init__(workload_settings, target_settings,
shutdown_event)
self.new_queries = N1QLQueryGen(workload_settings.n1ql_queries)
self.total_workers = self.ws.n1ql_workers
self.throughput = self.ws.n1ql_throughput
self.name = 'n1ql-worker'
host, port = self.ts.node.split(':')
params = {
'bucket': self.ts.bucket,
'host': host,
'port': port,
'username': self.ts.bucket,
'password': self.ts.password
}
self.existing_keys = ExistingKey(self.ws.working_set,
self.ws.working_set_access, 'n1ql')
self.new_keys = NewKey('n1ql', self.ws.expiration)
self.keys_for_removal = KeyForRemoval('n1ql')
self.keys_for_casupdate = KeyForCASUpdate(self.total_workers,
self.ws.working_set,
self.ws.working_set_access,
'n1ql')
if self.ws.doc_gen == 'merge':
self.docs = MergeDocument(self.ws.size, self.ws.doc_partitions,
False)
elif self.ws.doc_gen == 'reverse_lookup':
self.docs = ReverseLookupDocument(self.ws.size,
self.ws.doc_partitions, False)
elif self.ws.doc_gen == 'reverse_lookup_array_indexing':
self.docs = ReverseLookupDocumentArrayIndexing(
self.ws.size, self.ws.doc_partitions, self.ws.items)
self.cb = N1QLGen(**params)
@with_sleep
def do_batch(self):
if self.ws.n1ql_op == 'read':
curr_items_spot = \
self.curr_items.value - self.ws.creates * self.ws.workers
deleted_spot = \
self.deleted_items.value + self.ws.deletes * self.ws.workers
for _ in xrange(self.BATCH_SIZE):
key = self.existing_keys.next(curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
return
curr_items_tmp = curr_items_spot = self.curr_items.value
if self.ws.n1ql_op == 'create':
with self.lock:
self.curr_items.value += self.BATCH_SIZE
curr_items_tmp = self.curr_items.value - self.BATCH_SIZE
curr_items_spot = (curr_items_tmp -
self.BATCH_SIZE * self.total_workers)
deleted_items_tmp = deleted_spot = 0
if self.ws.n1ql_op == 'delete':
with self.lock:
self.deleted_items.value += self.BATCH_SIZE
deleted_items_tmp = self.deleted_items.value - self.BATCH_SIZE
deleted_spot = (deleted_items_tmp +
self.BATCH_SIZE * self.total_workers)
deleted_capped_items_tmp = deleted_capped_spot = 0
if self.ws.n1ql_op == 'rangedelete':
with self.lock:
self.deleted_capped_items.value += self.BATCH_SIZE
deleted_capped_items_tmp = self.deleted_capped_items.value - self.BATCH_SIZE
deleted_capped_spot = (deleted_capped_items_tmp +
self.BATCH_SIZE * self.total_workers)
casupdated_items_tmp = casupdated_spot = 0
if self.ws.n1ql_op == 'update':
with self.lock:
self.casupdated_items.value += self.BATCH_SIZE
casupdated_items_tmp = self.casupdated_items.value - self.BATCH_SIZE
casupdated_spot = (casupdated_items_tmp +
self.BATCH_SIZE * self.total_workers)
if self.ws.n1ql_op == 'create':
for _ in xrange(self.BATCH_SIZE):
curr_items_tmp += 1
key, ttl = self.new_keys.next(curr_items_tmp)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'delete':
for _ in xrange(self.BATCH_SIZE):
deleted_items_tmp += 1
key = self.keys_for_removal.next(deleted_items_tmp)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'update' or self.ws.n1ql_op == 'lookupupdate':
for _ in xrange(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid, curr_items_spot,
deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'rangeupdate':
for _ in xrange(self.BATCH_SIZE):
key = self.keys_for_casupdate.next(self.sid, curr_items_spot,
deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
elif self.ws.n1ql_op == 'rangedelete':
for _ in xrange(self.BATCH_SIZE):
doc = {}
doc['capped_small'] = "n1ql-_100_" + str(
deleted_capped_items_tmp)
ddoc_name, view_name, query = self.new_queries.next(doc)
self.cb.query(ddoc_name, view_name, query=query)
deleted_capped_items_tmp += 1
elif self.ws.n1ql_op == 'merge': #run select * workload for merge
for _ in xrange(self.BATCH_SIZE):
key = self.existing_keys.next(curr_items_spot, deleted_spot)
doc = self.docs.next(key)
doc['key'] = key
doc['bucket'] = self.ts.bucket
ddoc_name, view_name, query = self.new_queries.next(doc)
query['statement'] = "SELECT * FROM `bucket-1` USE KEYS[$1];"
query['args'] = "[\"{key}\"]".format(**doc)
del query['prepared']
self.cb.query(ddoc_name, view_name, query=query)
def run(self, sid, lock, curr_queries, curr_items, deleted_items,
casupdated_items, deleted_capped_items):
self.cb.start_updater()
if self.throughput < float('inf'):
self.target_time = float(self.BATCH_SIZE) * self.total_workers / \
self.throughput
else:
self.target_time = None
self.lock = lock
self.sid = sid
self.curr_items = curr_items
self.deleted_items = deleted_items
self.deleted_capped_items = deleted_capped_items
self.casupdated_items = casupdated_items
self.curr_queries = curr_queries
try:
logger.info('Started: {}-{}'.format(self.name, self.sid))
while curr_queries.value < self.ws.ops and not self.time_to_stop():
with self.lock:
curr_queries.value += self.BATCH_SIZE
self.do_batch()
self.report_progress(curr_queries.value)
except (KeyboardInterrupt, ValueFormatError, AttributeError) as e:
logger.info('Interrupted: {}-{}-{}'.format(self.name, self.sid, e))
else:
logger.info('Finished: {}-{}'.format(self.name, self.sid))