It seems very complicated to make a replicated broker work regarding consumers: it seems when stopping certain brokers, some consumers don't work anymore and, when the specific broker is up again, those consumers that didn't work receive all the "missing" messages.
I am using a 2 brokers scenario. Created a replicated topic like this:
$KAFKA_HOME/bin/kafka-topics.sh --create \ --zookeeper localhost:2181 \ --replication-factor 2 \ --partitions 3 \ --topic replicated_topic The excerpt from the server config looks like this ( notice it is the same for both servers except port, broker id and log dir):
# Licensed to the Apache Software Foundation (ASF) under one or more # contributor license agreements. See the NOTICE file distributed with # this work for additional information regarding copyright ownership. # The ASF licenses this file to You 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. # see kafka.server.KafkaConfig for additional details and defaults ############################# Server Basics ############################# # The id of the broker. This must be set to a unique integer for each broker. broker.id=0 ############################# Socket Server Settings ############################# # The address the socket server listens on. It will get the value returned from # java.net.InetAddress.getCanonicalHostName() if not configured. # FORMAT: # listeners = listener_name://host_name:port # EXAMPLE: # listeners = PLAINTEXT://your.host.name:9092 listeners=PLAINTEXT://:9092 # Hostname and port the broker will advertise to producers and consumers. If not set, # it uses the value for "listeners" if configured. Otherwise, it will use the value # returned from java.net.InetAddress.getCanonicalHostName(). #advertised.listeners=PLAINTEXT://your.host.name:9092 # Maps listener names to security protocols, the default is for them to be the same. See the config documentation for more details #listener.security.protocol.map=PLAINTEXT:PLAINTEXT,SSL:SSL,SASL_PLAINTEXT:SASL_PLAINTEXT,SASL_SSL:SASL_SSL # The number of threads that the server uses for receiving requests from the network and sending responses to the network num.network.threads=3 # The number of threads that the server uses for processing requests, which may include disk I/O num.io.threads=8 # The send buffer (SO_SNDBUF) used by the socket server socket.send.buffer.bytes=102400 # The receive buffer (SO_RCVBUF) used by the socket server socket.receive.buffer.bytes=102400 # The maximum size of a request that the socket server will accept (protection against OOM) socket.request.max.bytes=104857600 ############################# Log Basics ############################# # A comma seperated list of directories under which to store log files log.dirs=/tmp/kafka-logs0 # The default number of log partitions per topic. More partitions allow greater # parallelism for consumption, but this will also result in more files across # the brokers. num.partitions=1 # The number of threads per data directory to be used for log recovery at startup and flushing at shutdown. # This value is recommended to be increased for installations with data dirs located in RAID array. num.recovery.threads.per.data.dir=1 ############################# Internal Topic Settings ############################# # The replication factor for the group metadata internal topics "__consumer_offsets" and "__transaction_state" # For anything other than development testing, a value greater than 1 is recommended for to ensure availability such as 3. offsets.topic.replication.factor=2 transaction.state.log.replication.factor=1 transaction.state.log.min.isr=1 ############################# Log Flush Policy ############################# # Messages are immediately written to the filesystem but by default we only fsync() to sync # the OS cache lazily. The following configurations control the flush of data to disk. # There are a few important trade-offs here: # 1. Durability: Unflushed data may be lost if you are not using replication. # 2. Latency: Very large flush intervals may lead to latency spikes when the flush does occur as there will be a lot of data to flush. # 3. Throughput: The flush is generally the most expensive operation, and a small flush interval may lead to exceessive seeks. # The settings below allow one to configure the flush policy to flush data after a period of time or # every N messages (or both). This can be done globally and overridden on a per-topic basis. # The number of messages to accept before forcing a flush of data to disk #log.flush.interval.messages=10000 # The maximum amount of time a message can sit in a log before we force a flush #log.flush.interval.ms=1000 ############################# Log Retention Policy ############################# # The following configurations control the disposal of log segments. The policy can # be set to delete segments after a period of time, or after a given size has accumulated. # A segment will be deleted whenever *either* of these criteria are met. Deletion always happens # from the end of the log. # The minimum age of a log file to be eligible for deletion due to age log.retention.hours=168 # A size-based retention policy for logs. Segments are pruned from the log unless the remaining # segments drop below log.retention.bytes. Functions independently of log.retention.hours. #log.retention.bytes=1073741824 # The maximum size of a log segment file. When this size is reached a new log segment will be created. log.segment.bytes=1073741824 # The interval at which log segments are checked to see if they can be deleted according # to the retention policies log.retention.check.interval.ms=300000 ############################# Zookeeper ############################# # Zookeeper connection string (see zookeeper docs for details). # This is a comma separated host:port pairs, each corresponding to a zk # server. e.g. "127.0.0.1:3000,127.0.0.1:3001,127.0.0.1:3002". # You can also append an optional chroot string to the urls to specify the # root directory for all kafka znodes. zookeeper.connect=localhost:2181 # Timeout in ms for connecting to zookeeper zookeeper.connection.timeout.ms=6000 ############################# Group Coordinator Settings ############################# # The following configuration specifies the time, in milliseconds, that the GroupCoordinator will delay the initial consumer rebalance. # The rebalance will be further delayed by the value of group.initial.rebalance.delay.ms as new members join the group, up to a maximum of max.poll.interval.ms. # The default value for this is 3 seconds. # We override this to 0 here as it makes for a better out-of-the-box experience for development and testing. # However, in production environments the default value of 3 seconds is more suitable as this will help to avoid unnecessary, and potentially expensive, rebalances during application startup. group.initial.rebalance.delay.ms=0 Let's decribe my topic using 2 brokers:
Topic:replicated_topic PartitionCount:3 ReplicationFactor:2 Configs: Topic: replicated_topic Partition: 0 Leader: 1 Replicas: 1,0 Isr: 1,0 Topic: replicated_topic Partition: 1 Leader: 0 Replicas: 0,1 Isr: 1,0 Topic: replicated_topic Partition: 2 Leader: 1 Replicas: 1,0 Isr: 1,0 Let's see the code for the consumer: Consumer ( impl Callable )
@Override public Void call() throws Exception { final Properties props = new Properties(); props.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, bootstrapServers); props.put(ConsumerConfig.GROUP_ID_CONFIG, groupId); props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, IntegerDeserializer.class.getName()); props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName()); final Consumer<Integer, String> consumer = new KafkaConsumer<>(props); consumer.subscribe(Collections.singletonList(topicName)); ConsumerRecords<Integer, String> records = null; while (!Thread.currentThread().isInterrupted()) { records = consumer.poll(1000); if (records.isEmpty()) { continue; } records.forEach(rec -> LOGGER.debug("{}@{} consumed from topic {}, partition {} pair ({},{})", ConsumerCallable.class.getSimpleName(), hashCode(), rec.topic(), rec.partition(), rec.key(), rec.value())); consumer.commitAsync(); } consumer.close(); return null; } Producer and main code:
private static final String TOPIC_NAME = "replicated_topic"; private static final String BOOTSTRAP_SERVERS = "localhost:9092, localhost:9093"; private static final Logger LOGGER = LoggerFactory.getLogger(Main.class); public static void main(String[] args) { ExecutorService executor = Executors.newCachedThreadPool(); executor.submit(new ConsumerCallable(TOPIC_NAME, BOOTSTRAP_SERVERS, "group1")); executor.submit(new ConsumerCallable(TOPIC_NAME, BOOTSTRAP_SERVERS, "group2")); executor.submit(new ConsumerCallable(TOPIC_NAME, BOOTSTRAP_SERVERS, "group3")); try (Producer<Integer, String> producer = createProducer()) { Scanner scanner = new Scanner(System.in); String line = null; LOGGER.debug("Please enter 'k v' on the command line to send to Kafka or 'quit' to exit"); while (scanner.hasNextLine()) { line = scanner.nextLine(); if (line.trim().toLowerCase().equals("quit")) { break; } String[] elements = line.split(" "); Integer key = Integer.parseInt(elements[0]); String value = elements[1]; producer.send(new ProducerRecord<>(TOPIC_NAME, key, value)); producer.flush(); } } executor.shutdownNow(); } private static Producer<Integer, String> createProducer() { Properties props = new Properties(); props.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, BOOTSTRAP_SERVERS); props.put(ProducerConfig.CLIENT_ID_CONFIG, "KafkaExampleProducer"); props.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, IntegerSerializer.class.getName()); props.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, StringSerializer.class.getName()); return new KafkaProducer<>(props); } Now let's see the behaviour:
- All brokers are up:
Output of kafka topic:
Topic:replicated_topic PartitionCount:3 ReplicationFactor:2 Configs: Topic: replicated_topic Partition: 0 Leader: 1 Replicas: 1,0 Isr: 1,0 Topic: replicated_topic Partition: 1 Leader: 0 Replicas: 0,1 Isr: 1,0 Topic: replicated_topic Partition: 2 Leader: 1 Replicas: 1,0 Isr: 1,0 Output of program:
12:52:30.460 DEBUG Main - Please enter 'k v' on the command line to send to Kafka or 'quit' to exit 1 u 12:52:35.555 DEBUG ConsumerCallable - ConsumerCallable@1241910294 consumed from topic replicated_topic, partition 0 pair (1,u) 12:52:35.559 DEBUG ConsumerCallable - ConsumerCallable@1361430455 consumed from topic replicated_topic, partition 0 pair (1,u) 12:52:35.559 DEBUG ConsumerCallable - ConsumerCallable@186743616 consumed from topic replicated_topic, partition 0 pair (1,u) 2 d 12:52:38.096 DEBUG ConsumerCallable - ConsumerCallable@186743616 consumed from topic replicated_topic, partition 2 pair (2,d) 12:52:38.098 DEBUG ConsumerCallable - ConsumerCallable@1361430455 consumed from topic replicated_topic, partition 2 pair (2,d) 12:52:38.100 DEBUG ConsumerCallable - ConsumerCallable@1241910294 consumed from topic replicated_topic, partition 2 pair (2,d) Since the consumers are in different groups all messages are broadcasted to them, everything is ok.
2 Bring down broker 2:
Describe topic:
Topic:replicated_topic PartitionCount:3 ReplicationFactor:2 Configs: Topic: replicated_topic Partition: 0 Leader: 0 Replicas: 1,0 Isr: 0 Topic: replicated_topic Partition: 1 Leader: 0 Replicas: 0,1 Isr: 0 Topic: replicated_topic Partition: 2 Leader: 0 Replicas: 1,0 Isr: 0 Output of program:
3 t 12:57:03.898 DEBUG ConsumerCallable - ConsumerCallable@186743616 consumed from topic replicated_topic, partition 1 pair (3,t) 4 p 12:57:06.058 DEBUG ConsumerCallable - ConsumerCallable@186743616 consumed from topic replicated_topic, partition 1 pair (4,p) Now only 1 consumer receives data. Let's bring up broker 2 again: Now the other 2 consumers receive the missing data:
12:57:50.863 DEBUG ConsumerCallable - ConsumerCallable@1241910294 consumed from topic replicated_topic, partition 1 pair (3,t) 12:57:50.863 DEBUG ConsumerCallable - ConsumerCallable@1241910294 consumed from topic replicated_topic, partition 1 pair (4,p) 12:57:50.870 DEBUG ConsumerCallable - ConsumerCallable@1361430455 consumed from topic replicated_topic, partition 1 pair (3,t) 12:57:50.870 DEBUG ConsumerCallable - ConsumerCallable@1361430455 consumed from topic replicated_topic, partition 1 pair (4,p) - Bring down broker 1:
Now only 2 consumers receive data:
5 c 12:59:13.718 DEBUG ConsumerCallable - ConsumerCallable@1361430455 consumed from topic replicated_topic, partition 2 pair (5,c) 12:59:13.737 DEBUG ConsumerCallable - ConsumerCallable@1241910294 consumed from topic replicated_topic, partition 2 pair (5,c) 6 s 12:59:16.437 DEBUG ConsumerCallable - ConsumerCallable@1361430455 consumed from topic replicated_topic, partition 2 pair (6,s) 12:59:16.438 DEBUG ConsumerCallable - ConsumerCallable@1241910294 consumed from topic replicated_topic, partition 2 pair (6,s) If I bring it on the other consumer wil also receive missing data.
My point guys ( sorry for big write but I am trying to capture the context ), is how to make sure that no matter what broker I would stop, the consumers would work correctly? ( receive all messages normally )?
PS: I tried setting the offsets.topic.replication.factor=2 or 3, but it didn't have any effect.
1 Answers
Answers 1
Messages to that broker will not be ignored if the no. of alive brokers is lesser than the configured replicas. Whenever a new Kafka broker joins the cluster, the data gets replicated to that node. https://stackoverflow.com/a/38998062/6274525
So when your broker 2 goes down, the messages still get pushed to another alive broker because there is 1 live broker and replication factor is 2. Since your other 2 consumers are subscribed to broker 2 (which is down), they are unable to consume.
When your broker 2 is up again, the data gets duplicated to this new node and hence the consumers attached to this node receive the message (referred by you as "missing" messages).
0 comments:
Post a Comment