# Copyright 2013-2017 DataStax, Inc. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. from cProfile import Profile import logging import os.path import sys from threading import Thread import time from optparse import OptionParser import uuid from greplin import scales dirname = os.path.dirname(os.path.abspath(__file__)) sys.path.append(dirname) sys.path.append(os.path.join(dirname, '..')) import cassandra from cassandra.cluster import Cluster from cassandra.io.asyncorereactor import AsyncoreConnection log = logging.getLogger() handler = logging.StreamHandler() handler.setFormatter(logging.Formatter("%(asctime)s [%(levelname)s] %(name)s: %(message)s")) log.addHandler(handler) logging.getLogger('cassandra').setLevel(logging.WARN) _log_levels = { 'CRITICAL': logging.CRITICAL, 'ERROR': logging.ERROR, 'WARN': logging.WARNING, 'WARNING': logging.WARNING, 'INFO': logging.INFO, 'DEBUG': logging.DEBUG, 'NOTSET': logging.NOTSET, } have_libev = False supported_reactors = [AsyncoreConnection] try: from cassandra.io.libevreactor import LibevConnection have_libev = True supported_reactors.append(LibevConnection) except ImportError as exc: pass have_twisted = False try: from cassandra.io.twistedreactor import TwistedConnection have_twisted = True supported_reactors.append(TwistedConnection) except ImportError as exc: log.exception("Error importing twisted") pass KEYSPACE = "testkeyspace" + str(int(time.time())) TABLE = "testtable" COLUMN_VALUES = { 'int': 42, 'text': "'42'", 'float': 42.0, 'uuid': uuid.uuid4(), 'timestamp': "'2016-02-03 04:05+0000'" } def setup(options): log.info("Using 'cassandra' package from %s", cassandra.__path__) cluster = Cluster(options.hosts, schema_metadata_enabled=False, token_metadata_enabled=False) try: session = cluster.connect() log.debug("Creating keyspace...") try: session.execute(""" CREATE KEYSPACE %s WITH replication = { 'class': 'SimpleStrategy', 'replication_factor': '2' } """ % options.keyspace) log.debug("Setting keyspace...") except cassandra.AlreadyExists: log.debug("Keyspace already exists") session.set_keyspace(options.keyspace) log.debug("Creating table...") create_table_query = """ CREATE TABLE {0} ( thekey text, """ for i in range(options.num_columns): create_table_query += "col{0} {1},\n".format(i, options.column_type) create_table_query += "PRIMARY KEY (thekey))" try: session.execute(create_table_query.format(TABLE)) except cassandra.AlreadyExists: log.debug("Table already exists.") finally: cluster.shutdown() def teardown(options): cluster = Cluster(options.hosts, schema_metadata_enabled=False, token_metadata_enabled=False) session = cluster.connect() if not options.keep_data: session.execute("DROP KEYSPACE " + options.keyspace) cluster.shutdown() def benchmark(thread_class): options, args = parse_options() for conn_class in options.supported_reactors: setup(options) log.info("==== %s ====" % (conn_class.__name__,)) kwargs = {'metrics_enabled': options.enable_metrics, 'connection_class': conn_class} if options.protocol_version: kwargs['protocol_version'] = options.protocol_version cluster = Cluster(options.hosts, **kwargs) session = cluster.connect(options.keyspace) log.debug("Sleeping for two seconds...") time.sleep(2.0) # Generate the query if options.read: query = "SELECT * FROM {0} WHERE thekey = '{{key}}'".format(TABLE) else: query = "INSERT INTO {0} (thekey".format(TABLE) for i in range(options.num_columns): query += ", col{0}".format(i) query += ") VALUES ('{key}'" for i in range(options.num_columns): query += ", {0}".format(COLUMN_VALUES[options.column_type]) query += ")" values = None # we don't use that anymore. Keeping it in case we go back to prepared statements. per_thread = options.num_ops // options.threads threads = [] log.debug("Beginning {0}...".format('reads' if options.read else 'inserts')) start = time.time() try: for i in range(options.threads): thread = thread_class( i, session, query, values, per_thread, cluster.protocol_version, options.profile) thread.daemon = True threads.append(thread) for thread in threads: thread.start() for thread in threads: while thread.is_alive(): thread.join(timeout=0.5) end = time.time() finally: cluster.shutdown() teardown(options) total = end - start log.info("Total time: %0.2fs" % total) log.info("Average throughput: %0.2f/sec" % (options.num_ops / total)) if options.enable_metrics: stats = scales.getStats()['cassandra'] log.info("Connection errors: %d", stats['connection_errors']) log.info("Write timeouts: %d", stats['write_timeouts']) log.info("Read timeouts: %d", stats['read_timeouts']) log.info("Unavailables: %d", stats['unavailables']) log.info("Other errors: %d", stats['other_errors']) log.info("Retries: %d", stats['retries']) request_timer = stats['request_timer'] log.info("Request latencies:") log.info(" min: %0.4fs", request_timer['min']) log.info(" max: %0.4fs", request_timer['max']) log.info(" mean: %0.4fs", request_timer['mean']) log.info(" stddev: %0.4fs", request_timer['stddev']) log.info(" median: %0.4fs", request_timer['median']) log.info(" 75th: %0.4fs", request_timer['75percentile']) log.info(" 95th: %0.4fs", request_timer['95percentile']) log.info(" 98th: %0.4fs", request_timer['98percentile']) log.info(" 99th: %0.4fs", request_timer['99percentile']) log.info(" 99.9th: %0.4fs", request_timer['999percentile']) def parse_options(): parser = OptionParser() parser.add_option('-H', '--hosts', default='127.0.0.1', help='cassandra hosts to connect to (comma-separated list) [default: %default]') parser.add_option('-t', '--threads', type='int', default=1, help='number of threads [default: %default]') parser.add_option('-n', '--num-ops', type='int', default=10000, help='number of operations [default: %default]') parser.add_option('--asyncore-only', action='store_true', dest='asyncore_only', help='only benchmark with asyncore connections') parser.add_option('--libev-only', action='store_true', dest='libev_only', help='only benchmark with libev connections') parser.add_option('--twisted-only', action='store_true', dest='twisted_only', help='only benchmark with Twisted connections') parser.add_option('-m', '--metrics', action='store_true', dest='enable_metrics', help='enable and print metrics for operations') parser.add_option('-l', '--log-level', default='info', help='logging level: debug, info, warning, or error') parser.add_option('-p', '--profile', action='store_true', dest='profile', help='Profile the run') parser.add_option('--protocol-version', type='int', dest='protocol_version', default=4, help='Native protocol version to use') parser.add_option('-c', '--num-columns', type='int', dest='num_columns', default=2, help='Specify the number of columns for the schema') parser.add_option('-k', '--keyspace', type='str', dest='keyspace', default=KEYSPACE, help='Specify the keyspace name for the schema') parser.add_option('--keep-data', action='store_true', dest='keep_data', default=False, help='Keep the data after the benchmark') parser.add_option('--column-type', type='str', dest='column_type', default='text', help='Specify the column type for the schema (supported: int, text, float, uuid, timestamp)') parser.add_option('--read', action='store_true', dest='read', default=False, help='Read mode') options, args = parser.parse_args() options.hosts = options.hosts.split(',') level = options.log_level.upper() try: log.setLevel(_log_levels[level]) except KeyError: log.warn("Unknown log level specified: %s; specify one of %s", options.log_level, _log_levels.keys()) if options.asyncore_only: options.supported_reactors = [AsyncoreConnection] elif options.libev_only: if not have_libev: log.error("libev is not available") sys.exit(1) options.supported_reactors = [LibevConnection] elif options.twisted_only: if not have_twisted: log.error("Twisted is not available") sys.exit(1) options.supported_reactors = [TwistedConnection] else: options.supported_reactors = supported_reactors if not have_libev: log.warning("Not benchmarking libev reactor because libev is not available") return options, args class BenchmarkThread(Thread): def __init__(self, thread_num, session, query, values, num_queries, protocol_version, profile): Thread.__init__(self) self.thread_num = thread_num self.session = session self.query = query self.values = values self.num_queries = num_queries self.protocol_version = protocol_version self.profiler = Profile() if profile else None def start_profile(self): if self.profiler: self.profiler.enable() def run_query(self, key, **kwargs): return self.session.execute_async(self.query.format(key=key), **kwargs) def finish_profile(self): if self.profiler: self.profiler.disable() self.profiler.dump_stats('profile-%d' % self.thread_num)