Package com.linkedin.davinci.kafka.consumer

Class LeaderProducedRecordContext

java.lang.Object

com.linkedin.davinci.kafka.consumer.LeaderProducedRecordContext

public class LeaderProducedRecordContext
extends java.lang.Object

This class holds all the necessary context information which is passed from kafka consumer thread -> kafka producer callback thread -> drainer thread. All constructors are private by intention. This object should be created through the static utility function based on usecase. Internally these utility function fills up the consumedOffset, messageType, keyBytes, valueUnion. consumer thread and drainer thread creates this object by calling the appropriate static utility function based on message type. kafka producer callback thread fills up the producedOffset. drainer thread completes the persistedToDBFuture.

Method Summary

Modifier and Type	Method	Description
void	completePersistedToDBFuture(java.lang.Exception e)
int	getConsumedKafkaClusterId()
long	getConsumedOffset()
byte[]	getKeyBytes()
MessageType	getMessageType()
java.util.concurrent.CompletableFuture<java.lang.Void>	getPersistedToDBFuture()
long	getProducedOffset()
long	getProducedTimestampMs()
java.lang.Object	getValueUnion()
boolean	hasCorrespondingUpstreamMessage()	Some bookkeeping operations are intended to be performed only on messages produced by the leader which have a directly corresponding upstream message, and should be skipped for messages that are generated by the leader in the absence of a directly corresponding upstream.
static LeaderProducedRecordContext	newChunkDeleteRecord(byte[] keyBytes, Delete valueUnion)
static LeaderProducedRecordContext	newChunkPutRecord(byte[] keyBytes, Put valueUnion)
static LeaderProducedRecordContext	newControlMessageRecord(byte[] keyBytes, ControlMessage valueUnion)
static LeaderProducedRecordContext	newControlMessageRecord(int consumedKafkaClusterId, long consumedOffset, byte[] keyBytes, ControlMessage valueUnion)
static LeaderProducedRecordContext	newDeleteRecord(int consumedKafkaClusterId, long consumedOffset, byte[] keyBytes, Delete valueUnion)
static LeaderProducedRecordContext	newPutRecord(int consumedKafkaClusterId, long consumedOffset, byte[] keyBytes, Put valueUnion)
static LeaderProducedRecordContext	newPutRecordWithFuture(int consumedKafkaClusterId, long consumedOffset, byte[] keyBytes, Put valueUnion, java.util.concurrent.CompletableFuture<java.lang.Void> persistedToDBFuture)
void	setKeyBytes(byte[] keyBytes)
void	setProducedOffset(long producerOffset)
void	setProducedTimestampMs(long timeMs)
java.lang.String	toString()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Method Detail

newControlMessageRecord

public static LeaderProducedRecordContext newControlMessageRecord(int consumedKafkaClusterId,
                                                                  long consumedOffset,
                                                                  byte[] keyBytes,
                                                                  ControlMessage valueUnion)

newControlMessageRecord

public static LeaderProducedRecordContext newControlMessageRecord(byte[] keyBytes,
                                                                  ControlMessage valueUnion)

newPutRecord

public static LeaderProducedRecordContext newPutRecord(int consumedKafkaClusterId,
                                                       long consumedOffset,
                                                       byte[] keyBytes,
                                                       Put valueUnion)

newChunkPutRecord

public static LeaderProducedRecordContext newChunkPutRecord(byte[] keyBytes,
                                                            Put valueUnion)

newChunkDeleteRecord

public static LeaderProducedRecordContext newChunkDeleteRecord(byte[] keyBytes,
                                                               Delete valueUnion)

newPutRecordWithFuture

public static LeaderProducedRecordContext newPutRecordWithFuture(int consumedKafkaClusterId,
                                                                 long consumedOffset,
                                                                 byte[] keyBytes,
                                                                 Put valueUnion,
                                                                 java.util.concurrent.CompletableFuture<java.lang.Void> persistedToDBFuture)

newDeleteRecord

public static LeaderProducedRecordContext newDeleteRecord(int consumedKafkaClusterId,
                                                          long consumedOffset,
                                                          byte[] keyBytes,
                                                          Delete valueUnion)

setKeyBytes

public void setKeyBytes(byte[] keyBytes)

setProducedOffset

public void setProducedOffset(long producerOffset)

getConsumedKafkaClusterId

public int getConsumedKafkaClusterId()

getConsumedOffset

public long getConsumedOffset()

getMessageType

public MessageType getMessageType()

getKeyBytes

public byte[] getKeyBytes()

getValueUnion

public java.lang.Object getValueUnion()

getProducedOffset

public long getProducedOffset()

setProducedTimestampMs

public void setProducedTimestampMs(long timeMs)

getProducedTimestampMs

public long getProducedTimestampMs()

getPersistedToDBFuture

public java.util.concurrent.CompletableFuture<java.lang.Void> getPersistedToDBFuture()

completePersistedToDBFuture

public void completePersistedToDBFuture(java.lang.Exception e)

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object

hasCorrespondingUpstreamMessage

public boolean hasCorrespondingUpstreamMessage()

Some bookkeeping operations are intended to be performed only on messages produced by the leader which have a directly corresponding upstream message, and should be skipped for messages that are generated by the leader in the absence of a directly corresponding upstream. This function helps disambiguate these cases.

Returns:

true if the message produced by the leader has a directly corresponding upstream message false if the message does not (e.g. happens in cases of leader-generated chunks or TopicSwitch)