def move_start_offset(self):
"""We move the start offset of the topic by writing really old messages
and waiting for them to be cleaned up.
"""
producer = VerifiableProducer(self.test_context, 1, self.kafka, self.topic,
throughput=-1, enable_idempotence=True,
create_time=1000)
producer.start()
wait_until(lambda: producer.num_acked > 0,
timeout_sec=30,
err_msg="Failed to get an acknowledgement for %ds" % 30)
# Wait 8 seconds to let the topic be seeded with messages that will
# be deleted. The 8 seconds is important, since we should get 2 deleted
# segments in this period based on the configured log roll time and the
# retention check interval.
time.sleep(8)
producer.stop()
self.logger.info("Seeded topic with %d messages which will be deleted" %\
producer.num_acked)
# Since the configured check interval is 5 seconds, we wait another
# 6 seconds to ensure that at least one more cleaning so that the last
# segment is deleted. An altenate to using timeouts is to poll each
# partition until the log start offset matches the end offset. The
# latter is more robust.
time.sleep(6)
class GetOffsetShellTest(Test):
"""
Tests GetOffsetShell tool
"""
def __init__(self, test_context):
super(GetOffsetShellTest, self).__init__(test_context)
self.num_zk = 1
self.num_brokers = 1
self.messages_received_count = 0
self.topics = {
TOPIC: {'partitions': NUM_PARTITIONS, 'replication-factor': REPLICATION_FACTOR}
}
self.zk = ZookeeperService(test_context, self.num_zk)
def setUp(self):
self.zk.start()
def start_kafka(self, security_protocol, interbroker_security_protocol):
self.kafka = KafkaService(
self.test_context, self.num_brokers,
self.zk, security_protocol=security_protocol,
interbroker_security_protocol=interbroker_security_protocol, topics=self.topics)
self.kafka.start()
def start_producer(self):
# This will produce to kafka cluster
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=TOPIC, throughput=1000, max_messages=MAX_MESSAGES)
self.producer.start()
current_acked = self.producer.num_acked
wait_until(lambda: self.producer.num_acked >= current_acked + MAX_MESSAGES, timeout_sec=10,
err_msg="Timeout awaiting messages to be produced and acked")
def start_consumer(self):
self.consumer = ConsoleConsumer(self.test_context, num_nodes=self.num_brokers, kafka=self.kafka, topic=TOPIC,
consumer_timeout_ms=1000)
self.consumer.start()
@cluster(num_nodes=4)
def test_get_offset_shell(self, security_protocol='PLAINTEXT'):
"""
Tests if GetOffsetShell is getting offsets correctly
:return: None
"""
self.start_kafka(security_protocol, security_protocol)
self.start_producer()
# Assert that offset fetched without any consumers consuming is 0
assert self.kafka.get_offset_shell(TOPIC, None, 1000, 1, -1), "%s:%s:%s" % (TOPIC, NUM_PARTITIONS - 1, 0)
self.start_consumer()
node = self.consumer.nodes[0]
wait_until(lambda: self.consumer.alive(node), timeout_sec=20, backoff_sec=.2, err_msg="Consumer was too slow to start")
# Assert that offset is correctly indicated by GetOffsetShell tool
wait_until(lambda: "%s:%s:%s" % (TOPIC, NUM_PARTITIONS - 1, MAX_MESSAGES) in self.kafka.get_offset_shell(TOPIC, None, 1000, 1, -1), timeout_sec=10,
err_msg="Timed out waiting to reach expected offset.")
def start_producer(self):
# This will produce to kafka cluster
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=TOPIC, throughput=1000, max_messages=MAX_MESSAGES)
self.producer.start()
current_acked = self.producer.num_acked
wait_until(lambda: self.producer.num_acked >= current_acked + MAX_MESSAGES, timeout_sec=10,
err_msg="Timeout awaiting messages to be produced and acked")
def test_producer_back_compatibility(self):
"""Run 0.9.X java producer against 0.8.X brokers.
This test documents the fact that java producer v0.9.0.0 and later won't run against 0.8.X brokers
the broker responds to a V1 produce request with a V0 fetch response; the client then tries to parse this V0
produce response as a V1 produce response, resulting in a BufferUnderflowException
"""
self.producer = VerifiableProducer(self.test_context,
self.num_producers,
self.kafka,
self.topic,
max_messages=100,
throughput=self.producer_throughput,
version=TRUNK)
node = self.producer.nodes[0]
try:
self.producer.start()
self.producer.wait()
raise Exception(
"0.9.X java producer should not run successfully against 0.8.X broker"
)
except:
# Expected
pass
finally:
self.producer.kill_node(node, clean_shutdown=False)
self.logger.info("Grepping producer log for expected error type")
node.account.ssh("egrep -m 1 %s %s" % (
"\"org\.apache\.kafka\.common\.protocol\.types\.SchemaException.*throttle_time_ms.*: java\.nio\.BufferUnderflowException\"",
self.producer.LOG_FILE),
allow_fail=False)
def __init__(self, test_context):
super(StreamsOptimizedTest, self).__init__(test_context)
self.topics = {
self.input_topic: {
'partitions': 6
},
self.aggregation_topic: {
'partitions': 6
},
self.reduce_topic: {
'partitions': 6
},
self.join_topic: {
'partitions': 6
}
}
self.zookeeper = ZookeeperService(self.test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context,
num_nodes=3,
zk=self.zookeeper,
topics=self.topics)
self.producer = VerifiableProducer(self.test_context,
1,
self.kafka,
self.input_topic,
throughput=1000,
acks=1)
def produce_and_consume(self, producer_version, consumer_version, group):
self.producer = VerifiableProducer(
self.test_context,
self.num_producers,
self.kafka,
self.topic,
throughput=self.producer_throughput,
message_validator=is_int,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context,
self.num_consumers,
self.kafka,
self.topic,
consumer_timeout_ms=30000,
message_validator=is_int,
version=KafkaVersion(consumer_version))
self.consumer.group_id = group
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(
self.messages_per_producer) == True,
timeout_sec=120,
backoff_sec=1,
err_msg=
"Producer did not produce all messages in reasonable amount of time"
))
def start_producer(self, max_messages, acks, timeout):
# This will produce to kafka cluster
current_acked = 0
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=TOPIC, throughput=1000, acks=acks, max_messages=max_messages)
self.producer.start()
wait_until(lambda: acks == 0 or self.producer.num_acked >= current_acked + max_messages, timeout_sec=timeout,
err_msg="Timeout awaiting messages to be produced and acked")
def test_consumer_back_compatibility(self):
"""Run the scala 0.8.X consumer against an 0.9.X cluster.
Expect 0.8.X scala consumer to fail with buffer underflow. This error is the same as when an 0.9.X producer
is run against an 0.8.X broker: the broker responds to a V1 fetch request with a V0 fetch response; the
client then tries to parse this V0 fetch response as a V1 fetch response, resulting in a BufferUnderflowException
"""
num_messages = 10
self.producer = VerifiableProducer(
self.test_context, self.num_producers, self.kafka, self.topic, max_messages=num_messages,
throughput=self.producer_throughput, version=LATEST_0_8_2)
self.consumer = ConsoleConsumer(
self.test_context, self.num_consumers, self.kafka, self.topic, group_id="consumer-09X",
consumer_timeout_ms=10000, message_validator=is_int, version=TRUNK)
self.old_consumer = ConsoleConsumer(
self.test_context, self.num_consumers, self.kafka, self.topic, group_id="consumer-08X",
consumer_timeout_ms=10000, message_validator=is_int, version=LATEST_0_8_2)
self.producer.run()
self.consumer.run()
self.old_consumer.run()
consumed = len(self.consumer.messages_consumed[1])
old_consumed = len(self.old_consumer.messages_consumed[1])
assert old_consumed == num_messages, "Expected 0.8.X scala consumer to consume %d, but only got %d" % (num_messages, old_consumed)
assert consumed == 0, "Expected 0.9.X scala consumer to fail to consume any messages, but got %d" % consumed
self.logger.info("Grepping consumer log for expected error type")
node = self.consumer.nodes[0]
node.account.ssh("egrep -m 1 %s %s" % ("\"java\.nio\.BufferUnderflowException\"", self.consumer.LOG_FILE), allow_fail=False)
def test_produce_consume(self, broker_version):
print("running producer_consumer_compat with broker_version = %s" %
broker_version)
self.kafka.set_version(KafkaVersion(broker_version))
self.kafka.security_protocol = "PLAINTEXT"
self.kafka.interbroker_security_protocol = self.kafka.security_protocol
self.producer = VerifiableProducer(
self.test_context,
self.num_producers,
self.kafka,
self.topic,
throughput=self.producer_throughput,
message_validator=is_int_with_prefix)
self.consumer = ConsoleConsumer(self.test_context,
self.num_consumers,
self.kafka,
self.topic,
consumer_timeout_ms=60000,
message_validator=is_int_with_prefix)
self.kafka.start()
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(
self.messages_per_producer) == True,
timeout_sec=120,
backoff_sec=1,
err_msg=
"Producer did not produce all messages in reasonable amount of time"
))
def create_producer(self, num_nodes=1, throughput=1000, **kwargs):
self.producer = VerifiableProducer(self.test_context,
num_nodes=num_nodes,
kafka=self.kafka,
topic=self.topic,
throughput=throughput,
**kwargs)
def test_compatibility(self, producer_version, consumer_version, compression_types, new_consumer=True, timestamp_type=None):
self.kafka = KafkaService(self.test_context, num_nodes=3, zk=self.zk, version=DEV_BRANCH, topics={self.topic: {
"partitions": 3,
"replication-factor": 3,
'configs': {"min.insync.replicas": 2}}})
for node in self.kafka.nodes:
if timestamp_type is not None:
node.config[config_property.MESSAGE_TIMESTAMP_TYPE] = timestamp_type
self.kafka.start()
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int,
compression_types=compression_types,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka,
self.topic, consumer_timeout_ms=30000, new_consumer=new_consumer,
message_validator=is_int, version=KafkaVersion(consumer_version))
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def run_with_failure(self, failure):
"""This is the top-level test template.
The steps are:
Produce messages in the background while driving some failure condition
When done driving failures, immediately stop producing
Consume all messages
Validate that messages acked by brokers were consumed
Note that consuming is a bit tricky, at least with console consumer. The goal is to consume all messages
(foreach partition) in the topic. In this case, waiting for the last message may cause the consumer to stop
too soon since console consumer is consuming multiple partitions from a single thread and therefore we lose
ordering guarantees.
Waiting on a count of consumed messages can be unreliable: if we stop consuming when num_consumed == num_acked,
we might exit early if some messages are duplicated (though not an issue here since producer retries==0)
Therefore rely here on the consumer.timeout.ms setting which times out on the interval between successively
consumed messages. Since we run the producer to completion before running the consumer, this is a reliable
indicator that nothing is left to consume.
"""
self.producer = VerifiableProducer(self.test_context,
self.num_producers,
self.kafka,
self.topic,
throughput=self.producer_throughput)
self.consumer = ConsoleConsumer(self.test_context,
self.num_consumers,
self.kafka,
self.topic,
consumer_timeout_ms=3000)
# Produce in a background thread while driving broker failures
self.producer.start()
if not wait_until(lambda: self.producer.num_acked > 5, timeout_sec=5):
raise RuntimeError(
"Producer failed to start in a reasonable amount of time.")
failure()
self.producer.stop()
self.acked = self.producer.acked
self.not_acked = self.producer.not_acked
self.logger.info("num not acked: %d" % self.producer.num_not_acked)
self.logger.info("num acked: %d" % self.producer.num_acked)
# Consume all messages
self.consumer.start()
self.consumer.wait()
self.consumed = self.consumer.messages_consumed[1]
self.logger.info("num consumed: %d" % len(self.consumed))
# Check produced vs consumed
success, msg = self.validate()
if not success:
self.mark_for_collect(self.producer)
assert success, msg
class TestVerifiableProducer(Test):
"""Sanity checks on verifiable producer service class."""
def __init__(self, test_context):
super(TestVerifiableProducer, self).__init__(test_context)
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(
test_context,
num_nodes=1,
zk=self.zk,
topics={self.topic: {
"partitions": 1,
"replication-factor": 1
}})
self.num_messages = 1000
# This will produce to source kafka cluster
self.producer = VerifiableProducer(test_context,
num_nodes=1,
kafka=self.kafka,
topic=self.topic,
max_messages=self.num_messages,
throughput=1000)
def setUp(self):
self.zk.start()
self.kafka.start()
@cluster(num_nodes=3)
@parametrize(producer_version=str(LATEST_0_8_2))
@parametrize(producer_version=str(LATEST_0_9))
@parametrize(producer_version=str(DEV_BRANCH))
def test_simple_run(self, producer_version=DEV_BRANCH):
"""
Test that we can start VerifiableProducer on the current branch snapshot version or against the 0.8.2 jar, and
verify that we can produce a small number of messages.
"""
node = self.producer.nodes[0]
node.version = KafkaVersion(producer_version)
self.producer.start()
wait_until(
lambda: self.producer.num_acked > 5,
timeout_sec=5,
err_msg="Producer failed to start in a reasonable amount of time.")
# using version.vstring (distutils.version.LooseVersion) is a tricky way of ensuring
# that this check works with DEV_BRANCH
# When running VerifiableProducer 0.8.X, both the current branch version and 0.8.X should show up because of the
# way verifiable producer pulls in some development directories into its classpath
if node.version <= LATEST_0_8_2:
assert is_version(node, [node.version.vstring, DEV_BRANCH.vstring])
else:
assert is_version(node, [node.version.vstring])
self.producer.wait()
num_produced = self.producer.num_acked
assert num_produced == self.num_messages, "num_produced: %d, num_messages: %d" % (
num_produced, self.num_messages)
def create_producer(self):
# This will produce to source kafka cluster
self.producer = VerifiableProducer(self.test_context,
num_nodes=1,
kafka=self.kafka,
topic=self.topic,
max_messages=self.num_messages,
throughput=self.num_messages // 10)
def create_producer_and_consumer(self):
self.producer = VerifiableProducer(
self.test_context, self.num_producers, self.kafka, self.topic,
throughput=self.producer_throughput)
self.consumer = ConsoleConsumer(
self.test_context, self.num_consumers, self.kafka, self.topic,
consumer_timeout_ms=60000, message_validator=is_int)
self.consumer.group_id = self.group
class TestVerifiableProducer(Test):
"""Sanity checks on verifiable producer service class."""
def __init__(self, test_context):
super(TestVerifiableProducer, self).__init__(test_context)
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(test_context, num_nodes=1, zk=self.zk,
topics={self.topic: {"partitions": 1, "replication-factor": 1}})
self.num_messages = 1000
# This will produce to source kafka cluster
self.producer = VerifiableProducer(test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
max_messages=self.num_messages, throughput=self.num_messages/5)
def setUp(self):
self.zk.start()
self.kafka.start()
@cluster(num_nodes=3)
@parametrize(producer_version=str(LATEST_0_8_2))
@parametrize(producer_version=str(LATEST_0_9))
@parametrize(producer_version=str(LATEST_0_10_0))
@parametrize(producer_version=str(LATEST_0_10_1))
@parametrize(producer_version=str(DEV_BRANCH))
def test_simple_run(self, producer_version=DEV_BRANCH):
"""
Test that we can start VerifiableProducer on the current branch snapshot version or against the 0.8.2 jar, and
verify that we can produce a small number of messages.
"""
node = self.producer.nodes[0]
node.version = KafkaVersion(producer_version)
self.producer.start()
wait_until(lambda: self.producer.num_acked > 5, timeout_sec=5,
err_msg="Producer failed to start in a reasonable amount of time.")
# using version.vstring (distutils.version.LooseVersion) is a tricky way of ensuring
# that this check works with DEV_BRANCH
# When running VerifiableProducer 0.8.X, both the current branch version and 0.8.X should show up because of the
# way verifiable producer pulls in some development directories into its classpath
#
# If the test fails here because 'ps .. | grep' couldn't find the process it means
# the login and grep that is_version() performs is slower than
# the time it takes the producer to produce its messages.
# Easy fix is to decrease throughput= above, the good fix is to make the producer
# not terminate until explicitly killed in this case.
if node.version <= LATEST_0_8_2:
assert is_version(node, [node.version.vstring, DEV_BRANCH.vstring], logger=self.logger)
else:
assert is_version(node, [node.version.vstring], logger=self.logger)
self.producer.wait()
num_produced = self.producer.num_acked
assert num_produced == self.num_messages, "num_produced: %d, num_messages: %d" % (num_produced, self.num_messages)
def __init__(self, test_context):
super(DelegationTokenTest, self).__init__(test_context)
self.test_context = test_context
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(
self.test_context,
num_nodes=1,
zk=self.zk,
zk_chroot="/kafka",
topics={self.topic: {
"partitions": 1,
"replication-factor": 1
}},
server_prop_overides=[[
config_property.DELEGATION_TOKEN_MAX_LIFETIME_MS, "604800000"
], [config_property.DELEGATION_TOKEN_EXPIRY_TIME_MS, "86400000"
], [config_property.DELEGATION_TOKEN_SECRET_KEY, "test12345"],
[
config_property.SASL_ENABLED_MECHANISMS,
"GSSAPI,SCRAM-SHA-256"
]])
self.jaas_deleg_conf_path = "/tmp/jaas_deleg.conf"
self.jaas_deleg_conf = ""
self.client_properties_content = """
security.protocol=SASL_PLAINTEXT
sasl.mechanism=SCRAM-SHA-256
sasl.kerberos.service.name=kafka
client.id=console-consumer
"""
self.client_kafka_opts = ' -Djava.security.auth.login.config=' + self.jaas_deleg_conf_path
self.producer = VerifiableProducer(
self.test_context,
num_nodes=1,
kafka=self.kafka,
topic=self.topic,
max_messages=1,
throughput=1,
kafka_opts_override=self.client_kafka_opts,
client_prop_file_override=self.client_properties_content)
self.consumer = ConsoleConsumer(
self.test_context,
num_nodes=1,
kafka=self.kafka,
topic=self.topic,
kafka_opts_override=self.client_kafka_opts,
client_prop_file_override=self.client_properties_content)
self.kafka.security_protocol = 'SASL_PLAINTEXT'
self.kafka.client_sasl_mechanism = 'GSSAPI,SCRAM-SHA-256'
self.kafka.interbroker_sasl_mechanism = 'GSSAPI'
def __init__(self, test_context):
super(TestVerifiableProducer, self).__init__(test_context)
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1) if quorum.for_test(test_context) == quorum.zk else None
self.kafka = KafkaService(test_context, num_nodes=1, zk=self.zk,
topics={self.topic: {"partitions": 1, "replication-factor": 1}})
self.num_messages = 1000
# This will produce to source kafka cluster
self.producer = VerifiableProducer(test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
max_messages=self.num_messages, throughput=self.num_messages // 10)
class SimpleConsumerShellTest(Test):
"""
Tests SimpleConsumerShell tool
"""
def __init__(self, test_context):
super(SimpleConsumerShellTest, self).__init__(test_context)
self.num_zk = 1
self.num_brokers = 1
self.messages_received_count = 0
self.topics = {TOPIC: {"partitions": NUM_PARTITIONS, "replication-factor": REPLICATION_FACTOR}}
self.zk = ZookeeperService(test_context, self.num_zk)
def setUp(self):
self.zk.start()
def start_kafka(self):
self.kafka = KafkaService(self.test_context, self.num_brokers, self.zk, topics=self.topics)
self.kafka.start()
def run_producer(self):
# This will produce to kafka cluster
self.producer = VerifiableProducer(
self.test_context, num_nodes=1, kafka=self.kafka, topic=TOPIC, throughput=1000, max_messages=MAX_MESSAGES
)
self.producer.start()
wait_until(
lambda: self.producer.num_acked == MAX_MESSAGES,
timeout_sec=10,
err_msg="Timeout awaiting messages to be produced and acked",
)
def start_simple_consumer_shell(self):
self.simple_consumer_shell = SimpleConsumerShell(self.test_context, 1, self.kafka, TOPIC)
self.simple_consumer_shell.start()
def test_simple_consumer_shell(self):
"""
Tests if SimpleConsumerShell is fetching expected records
:return: None
"""
self.start_kafka()
self.run_producer()
self.start_simple_consumer_shell()
# Assert that SimpleConsumerShell is fetching expected number of messages
wait_until(
lambda: self.simple_consumer_shell.get_output().count("\n") == (MAX_MESSAGES + 1),
timeout_sec=10,
err_msg="Timed out waiting to receive expected number of messages.",
)
def seed_messages(self, topic, num_seed_messages):
seed_timeout_sec = 10000
seed_producer = VerifiableProducer(context=self.test_context,
num_nodes=1,
kafka=self.kafka,
topic=topic,
message_validator=is_int,
max_messages=num_seed_messages,
enable_idempotence=True)
seed_producer.start()
wait_until(lambda: seed_producer.num_acked >= num_seed_messages,
timeout_sec=seed_timeout_sec,
err_msg="Producer failed to produce messages %d in %ds." %\
(self.num_seed_messages, seed_timeout_sec))
return seed_producer.acked
def seed_messages(self, topic, num_seed_messages):
seed_timeout_sec = 10000
seed_producer = VerifiableProducer(context=self.test_context,
num_nodes=1,
kafka=self.kafka,
topic=topic,
message_validator=is_int,
max_messages=num_seed_messages,
enable_idempotence=True)
seed_producer.start()
wait_until(lambda: seed_producer.num_acked >= num_seed_messages,
timeout_sec=seed_timeout_sec,
err_msg="Producer failed to produce messages %d in %ds." %\
(self.num_seed_messages, seed_timeout_sec))
return seed_producer.acked
def create_producer(self, num_nodes=1, throughput=1000, **kwargs):
self.producer = VerifiableProducer(self.test_context,
num_nodes=num_nodes,
kafka=self.kafka,
topic=self.topic,
throughput=throughput,
**kwargs)
def __init__(self, test_context):
super(DelegationTokenTest, self).__init__(test_context)
self.test_context = test_context
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=1, zk=self.zk, zk_chroot="/kafka",
topics={self.topic: {"partitions": 1, "replication-factor": 1}},
server_prop_overides=[
[config_property.DELEGATION_TOKEN_MAX_LIFETIME_MS, "604800000"],
[config_property.DELEGATION_TOKEN_EXPIRY_TIME_MS, "86400000"],
[config_property.DELEGATION_TOKEN_MASTER_KEY, "test12345"],
[config_property.SASL_ENABLED_MECHANISMS, "GSSAPI,SCRAM-SHA-256"]
])
self.jaas_deleg_conf_path = "/tmp/jaas_deleg.conf"
self.jaas_deleg_conf = ""
self.client_properties_content = """
security.protocol=SASL_PLAINTEXT
sasl.mechanism=SCRAM-SHA-256
sasl.kerberos.service.name=kafka
client.id=console-consumer
"""
self.client_kafka_opts=' -Djava.security.auth.login.config=' + self.jaas_deleg_conf_path
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic, max_messages=1,
throughput=1, kafka_opts_override=self.client_kafka_opts,
client_prop_file_override=self.client_properties_content)
self.consumer = ConsoleConsumer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
kafka_opts_override=self.client_kafka_opts,
client_prop_file_override=self.client_properties_content)
self.kafka.security_protocol = 'SASL_PLAINTEXT'
self.kafka.client_sasl_mechanism = 'GSSAPI,SCRAM-SHA-256'
self.kafka.interbroker_sasl_mechanism = 'GSSAPI'
def test_reassign_partitions(self, bounce_brokers,
reassign_from_offset_zero):
"""Reassign partitions tests.
Setup: 1 zk, 4 kafka nodes, 1 topic with partitions=20, replication-factor=3,
and min.insync.replicas=3
- Produce messages in the background
- Consume messages in the background
- Reassign partitions
- If bounce_brokers is True, also bounce a few brokers while partition re-assignment is in progress
- When done reassigning partitions and bouncing brokers, stop producing, and finish consuming
- Validate that every acked message was consumed
"""
self.kafka.start()
if not reassign_from_offset_zero:
self.move_start_offset()
self.producer = VerifiableProducer(self.test_context,
self.num_producers,
self.kafka,
self.topic,
throughput=self.producer_throughput,
enable_idempotence=True)
self.consumer = ConsoleConsumer(self.test_context,
self.num_consumers,
self.kafka,
self.topic,
consumer_timeout_ms=60000,
message_validator=is_int)
self.enable_idempotence = True
self.run_produce_consume_validate(
core_test_action=lambda: self.reassign_partitions(bounce_brokers))
def test_compatibility(self, producer_version, consumer_version, compression_types, new_consumer=True, timestamp_type=None):
self.kafka = KafkaService(self.test_context, num_nodes=3, zk=self.zk, version=TRUNK, topics={self.topic: {
"partitions": 3,
"replication-factor": 3,
'configs': {"min.insync.replicas": 2}}})
for node in self.kafka.nodes:
if timestamp_type is not None:
node.config[config_property.MESSAGE_TIMESTAMP_TYPE] = timestamp_type
self.kafka.start()
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int,
compression_types=compression_types,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka,
self.topic, consumer_timeout_ms=30000, new_consumer=new_consumer,
message_validator=is_int, version=KafkaVersion(consumer_version))
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def start_producer(self, max_messages, acks, timeout):
# This will produce to kafka cluster
current_acked = 0
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=TOPIC, throughput=1000, acks=acks, max_messages=max_messages)
self.producer.start()
wait_until(lambda: acks == 0 or self.producer.num_acked >= current_acked + max_messages, timeout_sec=timeout,
err_msg="Timeout awaiting messages to be produced and acked")
def __init__(self, test_context):
super(StreamsStaticMembershipTest, self).__init__(test_context)
self.topics = {
self.input_topic: {'partitions': 18},
}
self.zookeeper = ZookeeperService(self.test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=3,
zk=self.zookeeper, topics=self.topics)
self.producer = VerifiableProducer(self.test_context,
1,
self.kafka,
self.input_topic,
throughput=1000,
acks=1)
def test_replication_with_broker_failure(self, failure_mode, security_protocol, broker_type,
client_sasl_mechanism="GSSAPI", interbroker_sasl_mechanism="GSSAPI",
compression_type=None, enable_idempotence=False):
"""Replication tests.
These tests verify that replication provides simple durability guarantees by checking that data acked by
brokers is still available for consumption in the face of various failure scenarios.
Setup: 1 zk, 3 kafka nodes, 1 topic with partitions=3, replication-factor=3, and min.insync.replicas=2
- Produce messages in the background
- Consume messages in the background
- Drive broker failures (shutdown, or bounce repeatedly with kill -15 or kill -9)
- When done driving failures, stop producing, and finish consuming
- Validate that every acked message was consumed
"""
self.kafka.security_protocol = security_protocol
self.kafka.interbroker_security_protocol = security_protocol
self.kafka.client_sasl_mechanism = client_sasl_mechanism
self.kafka.interbroker_sasl_mechanism = interbroker_sasl_mechanism
new_consumer = False if self.kafka.security_protocol == "PLAINTEXT" else True
self.enable_idempotence = enable_idempotence
compression_types = None if not compression_type else [compression_type] * self.num_producers
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka, self.topic,
throughput=self.producer_throughput, compression_types=compression_types,
enable_idempotence=enable_idempotence)
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka, self.topic, new_consumer=new_consumer, consumer_timeout_ms=60000, message_validator=is_int)
self.kafka.start()
self.run_produce_consume_validate(core_test_action=lambda: failures[failure_mode](self, broker_type))
def test_consumer_back_compatibility(self):
"""Run the scala 0.8.X consumer against an 0.9.X cluster.
Expect 0.8.X scala consumer to fail with buffer underflow. This error is the same as when an 0.9.X producer
is run against an 0.8.X broker: the broker responds to a V1 fetch request with a V0 fetch response; the
client then tries to parse this V0 fetch response as a V1 fetch response, resulting in a BufferUnderflowException
"""
num_messages = 10
self.producer = VerifiableProducer(
self.test_context, self.num_producers, self.kafka, self.topic, max_messages=num_messages,
throughput=self.producer_throughput, version=LATEST_0_8_2)
self.consumer = ConsoleConsumer(
self.test_context, self.num_consumers, self.kafka, self.topic, group_id="consumer-09X",
consumer_timeout_ms=10000, message_validator=is_int, version=TRUNK)
self.old_consumer = ConsoleConsumer(
self.test_context, self.num_consumers, self.kafka, self.topic, group_id="consumer-08X",
consumer_timeout_ms=10000, message_validator=is_int, version=LATEST_0_8_2)
self.producer.run()
self.consumer.run()
self.old_consumer.run()
consumed = len(self.consumer.messages_consumed[1])
old_consumed = len(self.old_consumer.messages_consumed[1])
assert old_consumed == num_messages, "Expected 0.8.X scala consumer to consume %d, but only got %d" % (num_messages, old_consumed)
assert consumed == 0, "Expected 0.9.X scala consumer to fail to consume any messages, but got %d" % consumed
self.logger.info("Grepping consumer log for expected error type")
node = self.consumer.nodes[0]
node.account.ssh("egrep -m 1 %s %s" % ("\"java\.nio\.BufferUnderflowException\"", self.consumer.LOG_FILE), allow_fail=False)
def test_reassign_partitions(self, bounce_brokers, security_protocol):
"""Reassign partitions tests.
Setup: 1 zk, 3 kafka nodes, 1 topic with partitions=3, replication-factor=3, and min.insync.replicas=2
- Produce messages in the background
- Consume messages in the background
- Reassign partitions
- If bounce_brokers is True, also bounce a few brokers while partition re-assignment is in progress
- When done reassigning partitions and bouncing brokers, stop producing, and finish consuming
- Validate that every acked message was consumed
"""
self.kafka.security_protocol = security_protocol
self.kafka.interbroker_security_protocol = security_protocol
new_consumer = False if self.kafka.security_protocol == "PLAINTEXT" else True
self.producer = VerifiableProducer(self.test_context,
self.num_producers,
self.kafka,
self.topic,
throughput=self.producer_throughput)
self.consumer = ConsoleConsumer(self.test_context,
self.num_consumers,
self.kafka,
self.topic,
new_consumer=new_consumer,
consumer_timeout_ms=60000,
message_validator=is_int)
self.kafka.start()
self.run_produce_consume_validate(
core_test_action=lambda: self.reassign_partitions(bounce_brokers))
def get_producer(self, num_messages):
return VerifiableProducer(self.test_context,
1,
self.kafka,
self.inputTopic,
max_messages=num_messages,
acks=1)
def test_compressed_topic(self, compression_types):
"""Test produce => consume => validate for compressed topics
Setup: 1 zk, 1 kafka node, 1 topic with partitions=10, replication-factor=1
compression_types parameter gives a list of compression types (or no compression if
"none"). Each producer in a VerifiableProducer group (num_producers = 4) will use a
compression type from the list based on producer's index in the group.
- Produce messages in the background
- Consume messages in the background
- Stop producing, and finish consuming
- Validate that every acked message was consumed
"""
self.kafka.security_protocol = "PLAINTEXT"
self.kafka.interbroker_security_protocol = self.kafka.security_protocol
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int_with_prefix,
compression_types=compression_types)
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka, self.topic,
consumer_timeout_ms=60000, message_validator=is_int_with_prefix)
self.kafka.start()
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def test_replication_with_broker_failure(self, failure_mode,
security_protocol):
"""Replication tests.
These tests verify that replication provides simple durability guarantees by checking that data acked by
brokers is still available for consumption in the face of various failure scenarios.
Setup: 1 zk, 3 kafka nodes, 1 topic with partitions=3, replication-factor=3, and min.insync.replicas=2
- Produce messages in the background
- Consume messages in the background
- Drive broker failures (shutdown, or bounce repeatedly with kill -15 or kill -9)
- When done driving failures, stop producing, and finish consuming
- Validate that every acked message was consumed
"""
self.kafka.security_protocol = 'PLAINTEXT'
self.kafka.interbroker_security_protocol = security_protocol
self.producer = VerifiableProducer(self.test_context,
self.num_producers,
self.kafka,
self.topic,
throughput=self.producer_throughput)
self.consumer = ConsoleConsumer(self.test_context,
self.num_consumers,
self.kafka,
self.topic,
consumer_timeout_ms=60000,
message_validator=is_int)
self.kafka.start()
self.run_produce_consume_validate(
core_test_action=lambda: failures[failure_mode](self))
class ClientCompatibilityTestNewBroker(ProduceConsumeValidateTest):
def __init__(self, test_context):
super(ClientCompatibilityTestNewBroker, self).__init__(test_context=test_context)
def setUp(self):
self.topic = "test_topic"
self.zk = ZookeeperService(self.test_context, num_nodes=1)
self.zk.start()
# Producer and consumer
self.producer_throughput = 10000
self.num_producers = 1
self.num_consumers = 1
self.messages_per_producer = 1000
@cluster(num_nodes=6)
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(DEV_BRANCH), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(DEV_BRANCH), compression_types=["none"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(LATEST_0_9), compression_types=["none"], new_consumer=False, timestamp_type=None)
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(LATEST_0_9), compression_types=["snappy"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_2_2), consumer_version=str(LATEST_2_2), compression_types=["none"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_2_1), consumer_version=str(LATEST_2_1), compression_types=["zstd"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_2_0), consumer_version=str(LATEST_2_0), compression_types=["snappy"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_1_1), consumer_version=str(LATEST_1_1), compression_types=["lz4"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_1_0), consumer_version=str(LATEST_1_0), compression_types=["none"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_0_11_0), consumer_version=str(LATEST_0_11_0), compression_types=["gzip"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_0_10_2), consumer_version=str(LATEST_0_10_2), compression_types=["lz4"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_0_10_1), consumer_version=str(LATEST_0_10_1), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(LATEST_0_10_0), consumer_version=str(LATEST_0_10_0), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(DEV_BRANCH), compression_types=["none"], timestamp_type=None)
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(DEV_BRANCH), compression_types=["snappy"], timestamp_type=None)
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(LATEST_0_9), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(LATEST_0_8_2), consumer_version=str(LATEST_0_8_2), compression_types=["none"], new_consumer=False, timestamp_type=None)
def test_compatibility(self, producer_version, consumer_version, compression_types, new_consumer=True, timestamp_type=None):
self.kafka = KafkaService(self.test_context, num_nodes=3, zk=self.zk, version=DEV_BRANCH, topics={self.topic: {
"partitions": 3,
"replication-factor": 3,
'configs': {"min.insync.replicas": 2}}})
for node in self.kafka.nodes:
if timestamp_type is not None:
node.config[config_property.MESSAGE_TIMESTAMP_TYPE] = timestamp_type
self.kafka.start()
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int,
compression_types=compression_types,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka,
self.topic, consumer_timeout_ms=30000, new_consumer=new_consumer,
message_validator=is_int, version=KafkaVersion(consumer_version))
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def setup_producer(self, topic, max_messages=-1):
return VerifiableProducer(self.test_context,
self.num_producers,
self.kafka,
topic,
max_messages=max_messages,
throughput=500)
def __init__(self, test_context):
super(TestMirrorMakerService, self).__init__(test_context)
self.topic = "topic"
self.source_zk = ZookeeperService(test_context, num_nodes=1)
self.target_zk = ZookeeperService(test_context, num_nodes=1)
self.source_kafka = KafkaService(
test_context,
num_nodes=1,
zk=self.source_zk,
topics={self.topic: {
"partitions": 1,
"replication-factor": 1
}})
self.target_kafka = KafkaService(
test_context,
num_nodes=1,
zk=self.target_zk,
topics={self.topic: {
"partitions": 1,
"replication-factor": 1
}})
self.num_messages = 1000
# This will produce to source kafka cluster
self.producer = VerifiableProducer(test_context,
num_nodes=1,
kafka=self.source_kafka,
topic=self.topic,
max_messages=self.num_messages,
throughput=1000)
# Use a regex whitelist to check that the start command is well-formed in this case
self.mirror_maker = MirrorMaker(test_context,
num_nodes=1,
source=self.source_kafka,
target=self.target_kafka,
whitelist=".*",
consumer_timeout_ms=2000)
# This will consume from target kafka cluster
self.consumer = ConsoleConsumer(test_context,
num_nodes=1,
kafka=self.target_kafka,
topic=self.topic,
consumer_timeout_ms=1000)
class CompressionTest(ProduceConsumeValidateTest):
"""
These tests validate produce / consume for compressed topics.
"""
def __init__(self, test_context):
""":type test_context: ducktape.tests.test.TestContext"""
super(CompressionTest, self).__init__(test_context=test_context)
self.topic = "test_topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(test_context, num_nodes=1, zk=self.zk, topics={self.topic: {
"partitions": 10,
"replication-factor": 1}})
self.num_partitions = 10
self.timeout_sec = 60
self.producer_throughput = 1000
self.num_producers = 4
self.messages_per_producer = 1000
self.num_consumers = 1
def setUp(self):
self.zk.start()
def min_cluster_size(self):
# Override this since we're adding services outside of the constructor
return super(CompressionTest, self).min_cluster_size() + self.num_producers + self.num_consumers
@parametrize(compression_types=["snappy","gzip","lz4","none"], new_consumer=True)
@parametrize(compression_types=["snappy","gzip","lz4","none"], new_consumer=False)
def test_compressed_topic(self, compression_types, new_consumer):
"""Test produce => consume => validate for compressed topics
Setup: 1 zk, 1 kafka node, 1 topic with partitions=10, replication-factor=1
compression_types parameter gives a list of compression types (or no compression if
"none"). Each producer in a VerifiableProducer group (num_producers = 4) will use a
compression type from the list based on producer's index in the group.
- Produce messages in the background
- Consume messages in the background
- Stop producing, and finish consuming
- Validate that every acked message was consumed
"""
self.kafka.security_protocol = "PLAINTEXT"
self.kafka.interbroker_security_protocol = self.kafka.security_protocol
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int_with_prefix,
compression_types=compression_types)
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka, self.topic,
new_consumer=new_consumer, consumer_timeout_ms=60000,
message_validator=is_int_with_prefix)
self.kafka.start()
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def __init__(self, test_context):
super(StreamsCooperativeRebalanceUpgradeTest, self).__init__(test_context)
self.topics = {
self.source_topic: {'partitions': 9},
self.sink_topic: {'partitions': 9}
}
self.zookeeper = ZookeeperService(self.test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=3,
zk=self.zookeeper, topics=self.topics)
self.producer = VerifiableProducer(self.test_context,
1,
self.kafka,
self.source_topic,
throughput=1000,
acks=1)
def get_producer(self, topic, num_messages, repeating_keys=None):
return VerifiableProducer(self.test_context,
1,
self.kafka,
topic,
max_messages=num_messages,
acks=1,
repeating_keys=repeating_keys)
class ClientCompatibilityProduceConsumeTest(ProduceConsumeValidateTest):
"""
These tests validate that we can use a new client to produce and consume from older brokers.
"""
def __init__(self, test_context):
""":type test_context: ducktape.tests.test.TestContext"""
super(ClientCompatibilityProduceConsumeTest, self).__init__(test_context=test_context)
self.topic = "test_topic"
self.zk = ZookeeperService(test_context, num_nodes=3)
self.kafka = KafkaService(test_context, num_nodes=3, zk=self.zk, topics={self.topic:{
"partitions": 10,
"replication-factor": 2}})
self.num_partitions = 10
self.timeout_sec = 60
self.producer_throughput = 1000
self.num_producers = 2
self.messages_per_producer = 1000
self.num_consumers = 1
def setUp(self):
self.zk.start()
def min_cluster_size(self):
# Override this since we're adding services outside of the constructor
return super(ClientCompatibilityProduceConsumeTest, self).min_cluster_size() + self.num_producers + self.num_consumers
@parametrize(broker_version=str(DEV_BRANCH))
@parametrize(broker_version=str(LATEST_0_10_0))
@parametrize(broker_version=str(LATEST_0_10_1))
@parametrize(broker_version=str(LATEST_0_10_2))
@parametrize(broker_version=str(LATEST_0_11_0))
@parametrize(broker_version=str(LATEST_1_0))
@parametrize(broker_version=str(LATEST_1_1))
@parametrize(broker_version=str(LATEST_2_0))
@parametrize(broker_version=str(LATEST_2_1))
@parametrize(broker_version=str(LATEST_2_2))
@parametrize(broker_version=str(LATEST_2_3))
@parametrize(broker_version=str(LATEST_2_4))
@parametrize(broker_version=str(LATEST_2_5))
def test_produce_consume(self, broker_version):
print("running producer_consumer_compat with broker_version = %s" % broker_version, flush=True)
self.kafka.set_version(KafkaVersion(broker_version))
self.kafka.security_protocol = "PLAINTEXT"
self.kafka.interbroker_security_protocol = self.kafka.security_protocol
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int_with_prefix)
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka, self.topic,
consumer_timeout_ms=60000,
message_validator=is_int_with_prefix)
self.kafka.start()
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
class TestVerifiableProducer(Test):
"""Sanity checks on verifiable producer service class."""
def __init__(self, test_context):
super(TestVerifiableProducer, self).__init__(test_context)
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(test_context, num_nodes=1, zk=self.zk,
topics={self.topic: {"partitions": 1, "replication-factor": 1}})
self.num_messages = 1000
# This will produce to source kafka cluster
self.producer = VerifiableProducer(test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
max_messages=self.num_messages, throughput=1000)
def setUp(self):
self.zk.start()
self.kafka.start()
@parametrize(producer_version=str(LATEST_0_8_2))
@parametrize(producer_version=str(LATEST_0_9))
@parametrize(producer_version=str(TRUNK))
def test_simple_run(self, producer_version=TRUNK):
"""
Test that we can start VerifiableProducer on trunk or against the 0.8.2 jar, and
verify that we can produce a small number of messages.
"""
node = self.producer.nodes[0]
node.version = KafkaVersion(producer_version)
self.producer.start()
wait_until(lambda: self.producer.num_acked > 5, timeout_sec=5,
err_msg="Producer failed to start in a reasonable amount of time.")
# using version.vstring (distutils.version.LooseVersion) is a tricky way of ensuring
# that this check works with TRUNK
# When running VerifiableProducer 0.8.X, both trunk version and 0.8.X should show up because of the way
# verifiable producer pulls in some trunk directories into its classpath
if node.version <= LATEST_0_8_2:
assert is_version(node, [node.version.vstring, TRUNK.vstring])
else:
assert is_version(node, [node.version.vstring])
self.producer.wait()
num_produced = self.producer.num_acked
assert num_produced == self.num_messages, "num_produced: %d, num_messages: %d" % (num_produced, self.num_messages)
def test_version(self):
"""Check that console consumer v0.8.2.X successfully starts and consumes messages."""
self.kafka.start()
num_messages = 1000
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
max_messages=num_messages, throughput=1000)
self.producer.start()
self.producer.wait()
self.consumer.nodes[0].version = LATEST_0_8_2
self.consumer.consumer_timeout_ms = 1000
self.consumer.start()
self.consumer.wait()
num_consumed = len(self.consumer.messages_consumed[1])
num_produced = self.producer.num_acked
assert num_produced == num_consumed, "num_produced: %d, num_consumed: %d" % (num_produced, num_consumed)
class ClientCompatibilityTestNewBroker(ProduceConsumeValidateTest):
def __init__(self, test_context):
super(ClientCompatibilityTestNewBroker, self).__init__(test_context=test_context)
def setUp(self):
self.topic = "test_topic"
self.zk = ZookeeperService(self.test_context, num_nodes=1)
self.zk.start()
# Producer and consumer
self.producer_throughput = 10000
self.num_producers = 1
self.num_consumers = 1
self.messages_per_producer = 1000
@cluster(num_nodes=6)
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(DEV_BRANCH), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(DEV_BRANCH), compression_types=["none"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(LATEST_0_9), compression_types=["none"], new_consumer=False, timestamp_type=None)
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(LATEST_0_9), compression_types=["snappy"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_1_1), consumer_version=str(LATEST_1_1), compression_types=["lz4"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_1_0), consumer_version=str(LATEST_1_0), compression_types=["none"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_0_11_0), consumer_version=str(LATEST_0_11_0), compression_types=["gzip"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_0_10_2), consumer_version=str(LATEST_0_10_2), compression_types=["lz4"], timestamp_type=str("CreateTime"))
@parametrize(producer_version=str(LATEST_0_10_1), consumer_version=str(LATEST_0_10_1), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(LATEST_0_10_0), consumer_version=str(LATEST_0_10_0), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(DEV_BRANCH), compression_types=["none"], timestamp_type=None)
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(DEV_BRANCH), compression_types=["snappy"], timestamp_type=None)
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(LATEST_0_9), compression_types=["snappy"], timestamp_type=str("LogAppendTime"))
@parametrize(producer_version=str(LATEST_0_8_2), consumer_version=str(LATEST_0_8_2), compression_types=["none"], new_consumer=False, timestamp_type=None)
def test_compatibility(self, producer_version, consumer_version, compression_types, new_consumer=True, timestamp_type=None):
self.kafka = KafkaService(self.test_context, num_nodes=3, zk=self.zk, version=DEV_BRANCH, topics={self.topic: {
"partitions": 3,
"replication-factor": 3,
'configs': {"min.insync.replicas": 2}}})
for node in self.kafka.nodes:
if timestamp_type is not None:
node.config[config_property.MESSAGE_TIMESTAMP_TYPE] = timestamp_type
self.kafka.start()
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int,
compression_types=compression_types,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka,
self.topic, consumer_timeout_ms=30000, new_consumer=new_consumer,
message_validator=is_int, version=KafkaVersion(consumer_version))
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def __init__(self, test_context):
super(TestVerifiableProducer, self).__init__(test_context)
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(test_context, num_nodes=1, zk=self.zk,
topics={self.topic: {"partitions": 1, "replication-factor": 1}})
self.num_messages = 1000
# This will produce to source kafka cluster
self.producer = VerifiableProducer(test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
max_messages=self.num_messages, throughput=1000)
class ClientCompatibilityProduceConsumeTest(ProduceConsumeValidateTest):
"""
These tests validate that we can use a new client to produce and consume from older brokers.
"""
def __init__(self, test_context):
""":type test_context: ducktape.tests.test.TestContext"""
super(ClientCompatibilityProduceConsumeTest, self).__init__(test_context=test_context)
self.topic = "test_topic"
self.zk = ZookeeperService(test_context, num_nodes=3)
self.kafka = KafkaService(test_context, num_nodes=3, zk=self.zk, topics={self.topic:{
"partitions": 10,
"replication-factor": 2}})
self.num_partitions = 10
self.timeout_sec = 60
self.producer_throughput = 1000
self.num_producers = 2
self.messages_per_producer = 1000
self.num_consumers = 1
def setUp(self):
self.zk.start()
def min_cluster_size(self):
# Override this since we're adding services outside of the constructor
return super(ClientCompatibilityProduceConsumeTest, self).min_cluster_size() + self.num_producers + self.num_consumers
@parametrize(broker_version=str(DEV_BRANCH))
@parametrize(broker_version=str(LATEST_0_10_0))
@parametrize(broker_version=str(LATEST_0_10_1))
@parametrize(broker_version=str(LATEST_0_10_2))
@parametrize(broker_version=str(LATEST_0_11_0))
@parametrize(broker_version=str(LATEST_1_0))
@parametrize(broker_version=str(LATEST_1_1))
def test_produce_consume(self, broker_version):
print("running producer_consumer_compat with broker_version = %s" % broker_version)
self.kafka.set_version(KafkaVersion(broker_version))
self.kafka.security_protocol = "PLAINTEXT"
self.kafka.interbroker_security_protocol = self.kafka.security_protocol
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int_with_prefix)
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka, self.topic,
consumer_timeout_ms=60000,
message_validator=is_int_with_prefix)
self.kafka.start()
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def produce_and_consume(self, producer_version, consumer_version, group):
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic,
throughput=self.producer_throughput,
message_validator=is_int,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka,
self.topic, consumer_timeout_ms=30000,
message_validator=is_int, version=KafkaVersion(consumer_version))
self.consumer.group_id = group
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def test_produce_consume(self, broker_version):
print("running producer_consumer_compat with broker_version = %s" % broker_version)
self.kafka.set_version(KafkaVersion(broker_version))
self.kafka.security_protocol = "PLAINTEXT"
self.kafka.interbroker_security_protocol = self.kafka.security_protocol
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic, throughput=self.producer_throughput,
message_validator=is_int_with_prefix)
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka, self.topic,
consumer_timeout_ms=60000,
message_validator=is_int_with_prefix)
self.kafka.start()
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
def test_version(self):
"""Check that console consumer v0.8.2.X successfully starts and consumes messages."""
self.kafka.start()
num_messages = 1000
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
max_messages=num_messages, throughput=1000)
self.producer.start()
self.producer.wait()
self.consumer.nodes[0].version = LATEST_0_8_2
self.consumer.consumer_timeout_ms = 1000
self.consumer.start()
self.consumer.wait()
num_consumed = len(self.consumer.messages_consumed[1])
num_produced = self.producer.num_acked
assert num_produced == num_consumed, "num_produced: %d, num_consumed: %d" % (num_produced, num_consumed)
def __init__(self, test_context):
super(StreamsOptimizedTest, self).__init__(test_context)
self.topics = {
self.input_topic: {'partitions': 6},
self.aggregation_topic: {'partitions': 6},
self.reduce_topic: {'partitions': 6},
self.join_topic: {'partitions': 6}
}
self.zookeeper = ZookeeperService(self.test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=3,
zk=self.zookeeper, topics=self.topics)
self.producer = VerifiableProducer(self.test_context,
1,
self.kafka,
self.input_topic,
throughput=1000,
acks=1)
def test_producer_back_compatibility(self):
"""Run 0.9.X java producer against 0.8.X brokers.
This test documents the fact that java producer v0.9.0.0 and later won't run against 0.8.X brokers
the broker responds to a V1 produce request with a V0 fetch response; the client then tries to parse this V0
produce response as a V1 produce response, resulting in a BufferUnderflowException
"""
self.producer = VerifiableProducer(
self.test_context, self.num_producers, self.kafka, self.topic, max_messages=100,
throughput=self.producer_throughput, version=TRUNK)
node = self.producer.nodes[0]
try:
self.producer.start()
self.producer.wait()
raise Exception("0.9.X java producer should not run successfully against 0.8.X broker")
except:
# Expected
pass
finally:
self.producer.kill_node(node, clean_shutdown=False)
self.logger.info("Grepping producer log for expected error type")
node.account.ssh("egrep -m 1 %s %s" % ("\"org\.apache\.kafka\.common\.protocol\.types\.SchemaException.*throttle_time_ms.*: java\.nio\.BufferUnderflowException\"", self.producer.LOG_FILE), allow_fail=False)
class StreamsOptimizedTest(Test):
"""
Test doing upgrades of a Kafka Streams application
that is un-optimized initially then optimized
"""
input_topic = 'inputTopic'
aggregation_topic = 'aggregationTopic'
reduce_topic = 'reduceTopic'
join_topic = 'joinTopic'
operation_pattern = 'AGGREGATED\|REDUCED\|JOINED'
def __init__(self, test_context):
super(StreamsOptimizedTest, self).__init__(test_context)
self.topics = {
self.input_topic: {'partitions': 6},
self.aggregation_topic: {'partitions': 6},
self.reduce_topic: {'partitions': 6},
self.join_topic: {'partitions': 6}
}
self.zookeeper = ZookeeperService(self.test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=3,
zk=self.zookeeper, topics=self.topics)
self.producer = VerifiableProducer(self.test_context,
1,
self.kafka,
self.input_topic,
throughput=1000,
acks=1)
def test_upgrade_optimized_topology(self):
self.zookeeper.start()
self.kafka.start()
processor1 = StreamsOptimizedUpgradeTestService(self.test_context, self.kafka)
processor2 = StreamsOptimizedUpgradeTestService(self.test_context, self.kafka)
processor3 = StreamsOptimizedUpgradeTestService(self.test_context, self.kafka)
processors = [processor1, processor2, processor3]
# produce records continually during the test
self.producer.start()
# start all processors unoptimized
for processor in processors:
self.set_topics(processor)
processor.CLEAN_NODE_ENABLED = False
self.verify_running_repartition_topic_count(processor, 4)
self.verify_processing(processors, verify_individual_operations=False)
self.stop_processors(processors)
# start again with topology optimized
for processor in processors:
processor.OPTIMIZED_CONFIG = 'all'
self.verify_running_repartition_topic_count(processor, 1)
self.verify_processing(processors, verify_individual_operations=True)
self.stop_processors(processors)
self.producer.stop()
self.kafka.stop()
self.zookeeper.stop()
@staticmethod
def verify_running_repartition_topic_count(processor, repartition_topic_count):
node = processor.node
with node.account.monitor_log(processor.STDOUT_FILE) as monitor:
processor.start()
monitor.wait_until('REBALANCING -> RUNNING with REPARTITION TOPIC COUNT=%s' % repartition_topic_count,
timeout_sec=120,
err_msg="Never saw 'REBALANCING -> RUNNING with REPARTITION TOPIC COUNT=%s' message "
% repartition_topic_count + str(processor.node.account))
@staticmethod
def verify_stopped(processor):
node = processor.node
with node.account.monitor_log(processor.STDOUT_FILE) as monitor:
processor.stop()
monitor.wait_until('OPTIMIZE_TEST Streams Stopped',
timeout_sec=60,
err_msg="'OPTIMIZE_TEST Streams Stopped' message" + str(processor.node.account))
def verify_processing(self, processors, verify_individual_operations):
for processor in processors:
if not self.all_source_subtopology_tasks(processor):
if verify_individual_operations:
for operation in self.operation_pattern.split('\|'):
self.do_verify(processor, operation)
else:
self.do_verify(processor, self.operation_pattern)
else:
self.logger.info("Skipping processor %s with all source tasks" % processor.node.account)
def do_verify(self, processor, pattern):
self.logger.info("Verifying %s processing pattern in STDOUT_FILE" % pattern)
with processor.node.account.monitor_log(processor.STDOUT_FILE) as monitor:
monitor.wait_until(pattern,
timeout_sec=60,
err_msg="Never saw processing of %s " % pattern + str(processor.node.account))
def all_source_subtopology_tasks(self, processor):
retries = 0
while retries < 5:
found = list(processor.node.account.ssh_capture("sed -n 's/.*current active tasks: \[\(\(0_[0-9], \)\{3\}0_[0-9]\)\].*/\1/p' %s" % processor.LOG_FILE, allow_fail=True))
self.logger.info("Returned %s from assigned task check" % found)
if len(found) > 0:
return True
retries += 1
time.sleep(1)
return False
def stop_processors(self, processors):
for processor in processors:
self.verify_stopped(processor)
def set_topics(self, processor):
processor.INPUT_TOPIC = self.input_topic
processor.AGGREGATION_TOPIC = self.aggregation_topic
processor.REDUCE_TOPIC = self.reduce_topic
processor.JOIN_TOPIC = self.join_topic
class EndToEndTest(Test):
"""This class provides a shared template for tests which follow the common pattern of:
- produce to a topic in the background
- consume from that topic in the background
- run some logic, e.g. fail topic leader etc.
- perform validation
"""
DEFAULT_TOPIC_CONFIG = {"partitions": 2, "replication-factor": 1}
def __init__(self, test_context, topic="test_topic", topic_config=DEFAULT_TOPIC_CONFIG):
super(EndToEndTest, self).__init__(test_context=test_context)
self.topic = topic
self.topic_config = topic_config
self.records_consumed = []
self.last_consumed_offsets = {}
def create_zookeeper(self, num_nodes=1, **kwargs):
self.zk = ZookeeperService(self.test_context, num_nodes=num_nodes, **kwargs)
def create_kafka(self, num_nodes=1, **kwargs):
group_metadata_config = {
"partitions": num_nodes,
"replication-factor": min(num_nodes, 3),
"configs": {"cleanup.policy": "compact"}
}
topics = {
self.topic: self.topic_config,
"__consumer_offsets": group_metadata_config
}
self.kafka = KafkaService(self.test_context, num_nodes=num_nodes,
zk=self.zk, topics=topics, **kwargs)
def create_consumer(self, num_nodes=1, group_id="test_group", **kwargs):
self.consumer = VerifiableConsumer(self.test_context,
num_nodes=num_nodes,
kafka=self.kafka,
topic=self.topic,
group_id=group_id,
on_record_consumed=self.on_record_consumed,
**kwargs)
def create_producer(self, num_nodes=1, throughput=1000, **kwargs):
self.producer = VerifiableProducer(self.test_context,
num_nodes=num_nodes,
kafka=self.kafka,
topic=self.topic,
throughput=throughput,
**kwargs)
def on_record_consumed(self, record, node):
partition = TopicPartition(record["topic"], record["partition"])
record_id = int(record["value"])
offset = record["offset"]
self.last_consumed_offsets[partition] = offset
self.records_consumed.append(record_id)
def await_consumed_offsets(self, last_acked_offsets, timeout_sec):
def has_finished_consuming():
for partition, offset in last_acked_offsets.iteritems():
if not partition in self.last_consumed_offsets:
return False
if self.last_consumed_offsets[partition] < offset:
return False
return True
wait_until(has_finished_consuming,
timeout_sec=timeout_sec,
err_msg="Consumer failed to consume up to offsets %s after waiting %ds." %\
(str(last_acked_offsets), timeout_sec))
def _collect_all_logs(self):
for s in self.test_context.services:
self.mark_for_collect(s)
def await_startup(self, min_records=5, timeout_sec=30):
try:
wait_until(lambda: self.consumer.total_consumed() >= min_records,
timeout_sec=timeout_sec,
err_msg="Timed out after %ds while awaiting initial record delivery of %d records" %\
(timeout_sec, min_records))
except BaseException:
self._collect_all_logs()
raise
def run_validation(self, min_records=5000, producer_timeout_sec=30,
consumer_timeout_sec=30, enable_idempotence=False):
try:
wait_until(lambda: self.producer.num_acked > min_records,
timeout_sec=producer_timeout_sec,
err_msg="Producer failed to produce messages for %ds." %\
producer_timeout_sec)
self.logger.info("Stopping producer after writing up to offsets %s" %\
str(self.producer.last_acked_offsets))
self.producer.stop()
self.await_consumed_offsets(self.producer.last_acked_offsets, consumer_timeout_sec)
self.consumer.stop()
self.validate(enable_idempotence)
except BaseException:
self._collect_all_logs()
raise
def validate(self, enable_idempotence):
self.logger.info("Number of acked records: %d" % len(self.producer.acked))
self.logger.info("Number of consumed records: %d" % len(self.records_consumed))
def check_lost_data(missing_records):
return self.kafka.search_data_files(self.topic, missing_records)
succeeded, error_msg = validate_delivery(self.producer.acked, self.records_consumed,
enable_idempotence, check_lost_data)
# Collect all logs if validation fails
if not succeeded:
self._collect_all_logs()
assert succeeded, error_msg
class ConsoleConsumerTest(Test):
"""Sanity checks on console consumer service class."""
def __init__(self, test_context):
super(ConsoleConsumerTest, self).__init__(test_context)
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=1, zk=self.zk, zk_chroot="/kafka",
topics={self.topic: {"partitions": 1, "replication-factor": 1}})
self.consumer = ConsoleConsumer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic)
def setUp(self):
self.zk.start()
@cluster(num_nodes=3)
@matrix(security_protocol=['PLAINTEXT', 'SSL'])
@cluster(num_nodes=4)
@matrix(security_protocol=['SASL_SSL'], sasl_mechanism=['PLAIN', 'SCRAM-SHA-256', 'SCRAM-SHA-512'])
@matrix(security_protocol=['SASL_PLAINTEXT', 'SASL_SSL'])
def test_lifecycle(self, security_protocol, sasl_mechanism='GSSAPI'):
"""Check that console consumer starts/stops properly, and that we are capturing log output."""
self.kafka.security_protocol = security_protocol
self.kafka.client_sasl_mechanism = sasl_mechanism
self.kafka.interbroker_sasl_mechanism = sasl_mechanism
self.kafka.start()
self.consumer.security_protocol = security_protocol
t0 = time.time()
self.consumer.start()
node = self.consumer.nodes[0]
wait_until(lambda: self.consumer.alive(node),
timeout_sec=20, backoff_sec=.2, err_msg="Consumer was too slow to start")
self.logger.info("consumer started in %s seconds " % str(time.time() - t0))
# Verify that log output is happening
wait_until(lambda: file_exists(node, ConsoleConsumer.LOG_FILE), timeout_sec=10,
err_msg="Timed out waiting for consumer log file to exist.")
wait_until(lambda: line_count(node, ConsoleConsumer.LOG_FILE) > 0, timeout_sec=1,
backoff_sec=.25, err_msg="Timed out waiting for log entries to start.")
# Verify no consumed messages
assert line_count(node, ConsoleConsumer.STDOUT_CAPTURE) == 0
self.consumer.stop_node(node)
@cluster(num_nodes=4)
def test_version(self):
"""Check that console consumer v0.8.2.X successfully starts and consumes messages."""
self.kafka.start()
num_messages = 1000
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
max_messages=num_messages, throughput=1000)
self.producer.start()
self.producer.wait()
self.consumer.nodes[0].version = LATEST_0_8_2
self.consumer.new_consumer = False
self.consumer.consumer_timeout_ms = 1000
self.consumer.start()
self.consumer.wait()
num_consumed = len(self.consumer.messages_consumed[1])
num_produced = self.producer.num_acked
assert num_produced == num_consumed, "num_produced: %d, num_consumed: %d" % (num_produced, num_consumed)
class MessageFormatChangeTest(ProduceConsumeValidateTest):
def __init__(self, test_context):
super(MessageFormatChangeTest, self).__init__(test_context=test_context)
def setUp(self):
self.topic = "test_topic"
self.zk = ZookeeperService(self.test_context, num_nodes=1)
self.zk.start()
# Producer and consumer
self.producer_throughput = 10000
self.num_producers = 1
self.num_consumers = 1
self.messages_per_producer = 100
def produce_and_consume(self, producer_version, consumer_version, group):
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic,
throughput=self.producer_throughput,
message_validator=is_int,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka,
self.topic, consumer_timeout_ms=30000,
message_validator=is_int, version=KafkaVersion(consumer_version))
self.consumer.group_id = group
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
@cluster(num_nodes=12)
@parametrize(producer_version=str(DEV_BRANCH), consumer_version=str(DEV_BRANCH))
@parametrize(producer_version=str(LATEST_0_10), consumer_version=str(LATEST_0_10))
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(LATEST_0_9))
def test_compatibility(self, producer_version, consumer_version):
""" This tests performs the following checks:
The workload is a mix of 0.9.x, 0.10.x and 0.11.x producers and consumers
that produce to and consume from a DEV_BRANCH cluster
1. initially the topic is using message format 0.9.0
2. change the message format version for topic to 0.10.0 on the fly.
3. change the message format version for topic to 0.11.0 on the fly.
4. change the message format version for topic back to 0.10.0 on the fly (only if the client version is 0.11.0 or newer)
- The producers and consumers should not have any issue.
Note regarding step number 4. Downgrading the message format version is generally unsupported as it breaks
older clients. More concretely, if we downgrade a topic from 0.11.0 to 0.10.0 after it contains messages with
version 0.11.0, we will return the 0.11.0 messages without down conversion due to an optimisation in the
handling of fetch requests. This will break any consumer that doesn't support 0.11.0. So, in practice, step 4
is similar to step 2 and it didn't seem worth it to increase the cluster size to in order to add a step 5 that
would change the message format version for the topic back to 0.9.0.0.
"""
self.kafka = KafkaService(self.test_context, num_nodes=3, zk=self.zk, version=DEV_BRANCH, topics={self.topic: {
"partitions": 3,
"replication-factor": 3,
'configs': {"min.insync.replicas": 2}}})
self.kafka.start()
self.logger.info("First format change to 0.9.0")
self.kafka.alter_message_format(self.topic, str(LATEST_0_9))
self.produce_and_consume(producer_version, consumer_version, "group1")
self.logger.info("Second format change to 0.10.0")
self.kafka.alter_message_format(self.topic, str(LATEST_0_10))
self.produce_and_consume(producer_version, consumer_version, "group2")
self.logger.info("Third format change to 0.11.0")
self.kafka.alter_message_format(self.topic, str(LATEST_0_11))
self.produce_and_consume(producer_version, consumer_version, "group3")
if producer_version == str(DEV_BRANCH) and consumer_version == str(DEV_BRANCH):
self.logger.info("Fourth format change back to 0.10.0")
self.kafka.alter_message_format(self.topic, str(LATEST_0_10))
self.produce_and_consume(producer_version, consumer_version, "group4")
class ClientCompatibilityTest(Test):
def __init__(self, test_context):
super(ClientCompatibilityTest, self).__init__(test_context=test_context)
def setUp(self):
self.topic = "test_topic"
self.zk = ZookeeperService(self.test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=3, zk=self.zk, version=LATEST_0_8_2, topics={self.topic: {
"partitions": 3,
"replication-factor": 3,
'configs': {"min.insync.replicas": 2}}})
self.zk.start()
self.kafka.start()
# Producer and consumer
self.producer_throughput = 10000
self.num_producers = 1
self.num_consumers = 1
def test_producer_back_compatibility(self):
"""Run 0.9.X java producer against 0.8.X brokers.
This test documents the fact that java producer v0.9.0.0 and later won't run against 0.8.X brokers
the broker responds to a V1 produce request with a V0 fetch response; the client then tries to parse this V0
produce response as a V1 produce response, resulting in a BufferUnderflowException
"""
self.producer = VerifiableProducer(
self.test_context, self.num_producers, self.kafka, self.topic, max_messages=100,
throughput=self.producer_throughput, version=TRUNK)
node = self.producer.nodes[0]
try:
self.producer.start()
self.producer.wait()
raise Exception("0.9.X java producer should not run successfully against 0.8.X broker")
except:
# Expected
pass
finally:
self.producer.kill_node(node, clean_shutdown=False)
self.logger.info("Grepping producer log for expected error type")
node.account.ssh("egrep -m 1 %s %s" % ("\"org\.apache\.kafka\.common\.protocol\.types\.SchemaException.*throttle_time_ms.*: java\.nio\.BufferUnderflowException\"", self.producer.LOG_FILE), allow_fail=False)
def test_consumer_back_compatibility(self):
"""Run the scala 0.8.X consumer against an 0.9.X cluster.
Expect 0.8.X scala consumer to fail with buffer underflow. This error is the same as when an 0.9.X producer
is run against an 0.8.X broker: the broker responds to a V1 fetch request with a V0 fetch response; the
client then tries to parse this V0 fetch response as a V1 fetch response, resulting in a BufferUnderflowException
"""
num_messages = 10
self.producer = VerifiableProducer(
self.test_context, self.num_producers, self.kafka, self.topic, max_messages=num_messages,
throughput=self.producer_throughput, version=LATEST_0_8_2)
self.consumer = ConsoleConsumer(
self.test_context, self.num_consumers, self.kafka, self.topic, group_id="consumer-09X",
consumer_timeout_ms=10000, message_validator=is_int, version=TRUNK)
self.old_consumer = ConsoleConsumer(
self.test_context, self.num_consumers, self.kafka, self.topic, group_id="consumer-08X",
consumer_timeout_ms=10000, message_validator=is_int, version=LATEST_0_8_2)
self.producer.run()
self.consumer.run()
self.old_consumer.run()
consumed = len(self.consumer.messages_consumed[1])
old_consumed = len(self.old_consumer.messages_consumed[1])
assert old_consumed == num_messages, "Expected 0.8.X scala consumer to consume %d, but only got %d" % (num_messages, old_consumed)
assert consumed == 0, "Expected 0.9.X scala consumer to fail to consume any messages, but got %d" % consumed
self.logger.info("Grepping consumer log for expected error type")
node = self.consumer.nodes[0]
node.account.ssh("egrep -m 1 %s %s" % ("\"java\.nio\.BufferUnderflowException\"", self.consumer.LOG_FILE), allow_fail=False)
class MessageFormatChangeTest(ProduceConsumeValidateTest):
def __init__(self, test_context):
super(MessageFormatChangeTest, self).__init__(test_context=test_context)
def setUp(self):
self.topic = "test_topic"
self.zk = ZookeeperService(self.test_context, num_nodes=1)
self.zk.start()
# Producer and consumer
self.producer_throughput = 10000
self.num_producers = 1
self.num_consumers = 1
self.messages_per_producer = 100
def produce_and_consume(self, producer_version, consumer_version, group):
self.producer = VerifiableProducer(self.test_context, self.num_producers, self.kafka,
self.topic,
throughput=self.producer_throughput,
message_validator=is_int,
version=KafkaVersion(producer_version))
self.consumer = ConsoleConsumer(self.test_context, self.num_consumers, self.kafka,
self.topic, consumer_timeout_ms=30000,
message_validator=is_int, version=KafkaVersion(consumer_version))
self.consumer.group_id = group
self.run_produce_consume_validate(lambda: wait_until(
lambda: self.producer.each_produced_at_least(self.messages_per_producer) == True,
timeout_sec=120, backoff_sec=1,
err_msg="Producer did not produce all messages in reasonable amount of time"))
@parametrize(producer_version=str(TRUNK), consumer_version=str(TRUNK))
@parametrize(producer_version=str(LATEST_0_9), consumer_version=str(LATEST_0_9))
def test_compatibility(self, producer_version, consumer_version):
""" This tests performs the following checks:
The workload is a mix of 0.9.x and 0.10.x producers and consumers
that produce to and consume from a 0.10.x cluster
1. initially the topic is using message format 0.9.0
2. change the message format version for topic to 0.10.0 on the fly.
3. change the message format version for topic back to 0.9.0 on the fly.
- The producers and consumers should not have any issue.
- Note that for 0.9.x consumers/producers we only do steps 1 and 2
"""
self.kafka = KafkaService(self.test_context, num_nodes=3, zk=self.zk, version=TRUNK, topics={self.topic: {
"partitions": 3,
"replication-factor": 3,
'configs': {"min.insync.replicas": 2}}})
self.kafka.start()
self.logger.info("First format change to 0.9.0")
self.kafka.alter_message_format(self.topic, str(LATEST_0_9))
self.produce_and_consume(producer_version, consumer_version, "group1")
self.logger.info("Second format change to 0.10.0")
self.kafka.alter_message_format(self.topic, str(LATEST_0_10))
self.produce_and_consume(producer_version, consumer_version, "group2")
if producer_version == str(TRUNK) and consumer_version == str(TRUNK):
self.logger.info("Third format change back to 0.9.0")
self.kafka.alter_message_format(self.topic, str(LATEST_0_9))
self.produce_and_consume(producer_version, consumer_version, "group3")
class DelegationTokenTest(Test):
def __init__(self, test_context):
super(DelegationTokenTest, self).__init__(test_context)
self.test_context = test_context
self.topic = "topic"
self.zk = ZookeeperService(test_context, num_nodes=1)
self.kafka = KafkaService(self.test_context, num_nodes=1, zk=self.zk, zk_chroot="/kafka",
topics={self.topic: {"partitions": 1, "replication-factor": 1}},
server_prop_overides=[
[config_property.DELEGATION_TOKEN_MAX_LIFETIME_MS, "604800000"],
[config_property.DELEGATION_TOKEN_EXPIRY_TIME_MS, "86400000"],
[config_property.DELEGATION_TOKEN_MASTER_KEY, "test12345"],
[config_property.SASL_ENABLED_MECHANISMS, "GSSAPI,SCRAM-SHA-256"]
])
self.jaas_deleg_conf_path = "/tmp/jaas_deleg.conf"
self.jaas_deleg_conf = ""
self.client_properties_content = """
security.protocol=SASL_PLAINTEXT
sasl.mechanism=SCRAM-SHA-256
sasl.kerberos.service.name=kafka
client.id=console-consumer
"""
self.client_kafka_opts=' -Djava.security.auth.login.config=' + self.jaas_deleg_conf_path
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic, max_messages=1,
throughput=1, kafka_opts_override=self.client_kafka_opts,
client_prop_file_override=self.client_properties_content)
self.consumer = ConsoleConsumer(self.test_context, num_nodes=1, kafka=self.kafka, topic=self.topic,
kafka_opts_override=self.client_kafka_opts,
client_prop_file_override=self.client_properties_content)
self.kafka.security_protocol = 'SASL_PLAINTEXT'
self.kafka.client_sasl_mechanism = 'GSSAPI,SCRAM-SHA-256'
self.kafka.interbroker_sasl_mechanism = 'GSSAPI'
def setUp(self):
self.zk.start()
def tearDown(self):
self.producer.nodes[0].account.remove(self.jaas_deleg_conf_path)
self.consumer.nodes[0].account.remove(self.jaas_deleg_conf_path)
def generate_delegation_token(self):
self.logger.debug("Request delegation token")
self.delegation_tokens.generate_delegation_token()
self.jaas_deleg_conf = self.delegation_tokens.create_jaas_conf_with_delegation_token()
def expire_delegation_token(self):
self.kafka.client_sasl_mechanism = 'GSSAPI,SCRAM-SHA-256'
token_hmac = self.delegation_tokens.token_hmac()
self.delegation_tokens.expire_delegation_token(token_hmac)
def produce_with_delegation_token(self):
self.producer.acked_values = []
self.producer.nodes[0].account.create_file(self.jaas_deleg_conf_path, self.jaas_deleg_conf)
self.logger.debug(self.jaas_deleg_conf)
self.producer.start()
def consume_with_delegation_token(self):
self.logger.debug("Consume messages with delegation token")
self.consumer.nodes[0].account.create_file(self.jaas_deleg_conf_path, self.jaas_deleg_conf)
self.logger.debug(self.jaas_deleg_conf)
self.consumer.consumer_timeout_ms = 5000
self.consumer.start()
self.consumer.wait()
def get_datetime_ms(self, input_date):
return int(time.mktime(datetime.strptime(input_date,"%Y-%m-%dT%H:%M").timetuple()) * 1000)
def renew_delegation_token(self):
dt = self.delegation_tokens.parse_delegation_token_out()
orig_expiry_date_ms = self.get_datetime_ms(dt["expirydate"])
new_expirydate_ms = orig_expiry_date_ms + 1000
self.delegation_tokens.renew_delegation_token(dt["hmac"], new_expirydate_ms)
def test_delegation_token_lifecycle(self):
self.kafka.start()
self.delegation_tokens = DelegationTokens(self.kafka, self.test_context)
self.generate_delegation_token()
self.renew_delegation_token()
self.produce_with_delegation_token()
wait_until(lambda: self.producer.num_acked > 0, timeout_sec=30,
err_msg="Expected producer to still be producing.")
assert 1 == self.producer.num_acked, "number of acked messages: %d" % self.producer.num_acked
self.consume_with_delegation_token()
num_consumed = len(self.consumer.messages_consumed[1])
assert 1 == num_consumed, "number of consumed messages: %d" % num_consumed
self.expire_delegation_token()
self.produce_with_delegation_token()
assert 0 == self.producer.num_acked, "number of acked messages: %d" % self.producer.num_acked
class ReplicaVerificationToolTest(Test):
"""
Tests ReplicaVerificationTool
"""
def __init__(self, test_context):
super(ReplicaVerificationToolTest, self).__init__(test_context)
self.num_zk = 1
self.num_brokers = 2
self.messages_received_count = 0
self.topics = {
TOPIC: {'partitions': 1, 'replication-factor': 2}
}
self.zk = ZookeeperService(test_context, self.num_zk)
self.kafka = None
self.producer = None
self.replica_verifier = None
def setUp(self):
self.zk.start()
def start_kafka(self, security_protocol, interbroker_security_protocol):
self.kafka = KafkaService(
self.test_context, self.num_brokers,
self.zk, security_protocol=security_protocol,
interbroker_security_protocol=interbroker_security_protocol, topics=self.topics)
self.kafka.start()
def start_replica_verification_tool(self, security_protocol):
self.replica_verifier = ReplicaVerificationTool(self.test_context, 1, self.kafka, TOPIC, report_interval_ms=REPORT_INTERVAL_MS, security_protocol=security_protocol)
self.replica_verifier.start()
def start_producer(self, max_messages, acks, timeout):
# This will produce to kafka cluster
self.producer = VerifiableProducer(self.test_context, num_nodes=1, kafka=self.kafka, topic=TOPIC, throughput=1000, acks=acks, max_messages=max_messages)
current_acked = self.producer.num_acked
self.logger.info("current_acked = %s" % current_acked)
self.producer.start()
wait_until(lambda: acks == 0 or self.producer.num_acked >= current_acked + max_messages, timeout_sec=timeout,
err_msg="Timeout awaiting messages to be produced and acked")
def stop_producer(self):
self.producer.stop()
def test_replica_lags(self, security_protocol='PLAINTEXT'):
"""
Tests ReplicaVerificationTool
:return: None
"""
self.start_kafka(security_protocol, security_protocol)
self.start_replica_verification_tool(security_protocol)
self.start_producer(max_messages=10, acks=-1, timeout=15)
# Verify that there is no lag in replicas and is correctly reported by ReplicaVerificationTool
wait_until(lambda: self.replica_verifier.get_lag_for_partition(TOPIC, 0) == 0, timeout_sec=10,
err_msg="Timed out waiting to reach zero replica lags.")
self.stop_producer()
self.start_producer(max_messages=1000, acks=0, timeout=5)
# Verify that there is lag in replicas and is correctly reported by ReplicaVerificationTool
wait_until(lambda: self.replica_verifier.get_lag_for_partition(TOPIC, 0) > 0, timeout_sec=10,
err_msg="Timed out waiting to reach non-zero number of replica lags.")
