[FLINK-27320] When using KafkaSink EXACTLY_ONCE semantics, frequent OutOfOrderSequenceException anomalies

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Done
• Affects Version/s: 1.14.4
• Component/s: Connectors / Kafka
• Flags: Important
• Assignee: Unassigned
• Reporter: Zhengqi Zhang
• Created: 20/Apr/22 09:58
• Updated: 24/Nov/22 08:26
• Resolved: 24/Nov/22 08:26

Description

This problem does not occur when using EXACTLY_ONCE semantics in FlinkKafkaProducer, but occurs frequently when using KafkaSink.

This is ProducerConfig when using KafkaSink:

    acks = 1
    batch.size = 16384
    bootstrap.servers = [localhost:9092]
    buffer.memory = 33554432
    client.dns.lookup = default
    client.id = 
    compression.type = none
    connections.max.idle.ms = 540000
    delivery.timeout.ms = 120000
    enable.idempotence = false
    interceptor.classes = []
    key.serializer = class org.apache.kafka.common.serialization.ByteArraySerializer
    linger.ms = 0
    max.block.ms = 60000
    max.in.flight.requests.per.connection = 5
    max.request.size = 1048576
    metadata.max.age.ms = 300000
    metric.reporters = []
    metrics.num.samples = 2
    metrics.recording.level = INFO
    metrics.sample.window.ms = 30000
    partitioner.class = class org.apache.kafka.clients.producer.internals.DefaultPartitioner
    receive.buffer.bytes = 32768
    reconnect.backoff.max.ms = 1000
    reconnect.backoff.ms = 50
    request.timeout.ms = 30000
    retries = 2147483647
    retry.backoff.ms = 100
    sasl.client.callback.handler.class = null
    sasl.jaas.config = null
    sasl.kerberos.kinit.cmd = /usr/bin/kinit
    sasl.kerberos.min.time.before.relogin = 60000
    sasl.kerberos.service.name = null
    sasl.kerberos.ticket.renew.jitter = 0.05
    sasl.kerberos.ticket.renew.window.factor = 0.8
    sasl.login.callback.handler.class = null
    sasl.login.class = null
    sasl.login.refresh.buffer.seconds = 300
    sasl.login.refresh.min.period.seconds = 60
    sasl.login.refresh.window.factor = 0.8
    sasl.login.refresh.window.jitter = 0.05
    sasl.mechanism = GSSAPI
    security.protocol = PLAINTEXT
    security.providers = null
    send.buffer.bytes = 131072
    ssl.cipher.suites = null
    ssl.enabled.protocols = [TLSv1.2, TLSv1.1, TLSv1]
    ssl.endpoint.identification.algorithm = https
    ssl.key.password = null
    ssl.keymanager.algorithm = SunX509
    ssl.keystore.location = null
    ssl.keystore.password = null
    ssl.keystore.type = JKS
    ssl.protocol = TLS
    ssl.provider = null
    ssl.secure.random.implementation = null
    ssl.trustmanager.algorithm = PKIX
    ssl.truststore.location = null
    ssl.truststore.password = null
    ssl.truststore.type = JKS
    transaction.timeout.ms = 300000
    transactional.id = kafka-sink-0-36
    value.serializer = class org.apache.kafka.common.serialization.ByteArraySerializer

This is ProducerConfig when using FlinkKafkaProducer:

 

    acks = 1
    batch.size = 16384
    bootstrap.servers = [localhost:9092]
    buffer.memory = 33554432
    client.dns.lookup = default
    client.id = 
    compression.type = none
    connections.max.idle.ms = 540000
    delivery.timeout.ms = 120000
    enable.idempotence = false
    interceptor.classes = []
    key.serializer = class org.apache.kafka.common.serialization.ByteArraySerializer
    linger.ms = 0
    max.block.ms = 60000
    max.in.flight.requests.per.connection = 5
    max.request.size = 1048576
    metadata.max.age.ms = 300000
    metric.reporters = []
    metrics.num.samples = 2
    metrics.recording.level = INFO
    metrics.sample.window.ms = 30000
    partitioner.class = class org.apache.kafka.clients.producer.internals.DefaultPartitioner
    receive.buffer.bytes = 32768
    reconnect.backoff.max.ms = 1000
    reconnect.backoff.ms = 50
    request.timeout.ms = 30000
    retries = 2147483647
    retry.backoff.ms = 100
    sasl.client.callback.handler.class = null
    sasl.jaas.config = null
    sasl.kerberos.kinit.cmd = /usr/bin/kinit
    sasl.kerberos.min.time.before.relogin = 60000
    sasl.kerberos.service.name = null
    sasl.kerberos.ticket.renew.jitter = 0.05
    sasl.kerberos.ticket.renew.window.factor = 0.8
    sasl.login.callback.handler.class = null
    sasl.login.class = null
    sasl.login.refresh.buffer.seconds = 300
    sasl.login.refresh.min.period.seconds = 60
    sasl.login.refresh.window.factor = 0.8
    sasl.login.refresh.window.jitter = 0.05
    sasl.mechanism = GSSAPI
    security.protocol = PLAINTEXT
    security.providers = null
    send.buffer.bytes = 131072
    ssl.cipher.suites = null
    ssl.enabled.protocols = [TLSv1.2, TLSv1.1, TLSv1]
    ssl.endpoint.identification.algorithm = https
    ssl.key.password = null
    ssl.keymanager.algorithm = SunX509
    ssl.keystore.location = null
    ssl.keystore.password = null
    ssl.keystore.type = JKS
    ssl.protocol = TLS
    ssl.provider = null
    ssl.secure.random.implementation = null
    ssl.trustmanager.algorithm = PKIX
    ssl.truststore.location = null
    ssl.truststore.password = null
    ssl.truststore.type = JKS
    transaction.timeout.ms = 300000
    transactional.id = Sink: Unnamed-0a448493b4782967b150582570326227-2
    value.serializer = class org.apache.kafka.common.serialization.ByteArraySerializer