Status
Current state: "Under Discussion"
Discussion thread:
JIRA: KAFKA-13602
Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).
Motivation
Lot of times, in Kafka Streams users want to send a record to more than one partition on the sink topic. Currently, if a user wants to replicate a message into N partitions, the only way of doing that is to replicate the message N times and then plug-in a custom partitioner to write the message N times into N different partitions. To give a more concrete example, let's say there's a sink topic with 10 partitions and a user wants to send records only to even numbered partitions.
In today's world, this is how the user can do the same:
final int numPartitions = 10; final KStream<String, String> inputStream = builder.stream("input-topic", Consumed.with(Serdes.String(), Serdes.String())); for (int i = 0; i < numPartitions; i += 2) { final int evenNumberedPartition = i; inputStream.to("output-topic", Produced.with(Serdes.String(), Serdes.String(), (topic, key, val, numPartitions) -> evenNumberedPartition)); }
This seems a little cumbersome way to broadcast. Also, there seems to be no way of dropping a record within the partitioner. This KIP aims to make this process simpler in Kafka Streams.
Public Interfaces
The StreamPartitioner method would have a new method added called partitions()
and the current partition()
method would be marked as deprecated. The partitions()
method would be marked as default within which it would invoke partition()
method and construct a singleton list out of it. Here's how the interface would look like now:
import java.util.List; public interface StreamPartitioner<K, V> { /** * Determine the partition number for a record with the given key and value and the current number of partitions. * * @param topic the topic name this record is sent to * @param key the key of the record * @param value the value of the record * @param numPartitions the total number of partitions * @return an integer between 0 and {@code numPartitions-1}, or {@code null} if the default partitioning logic should be used */ @Deprecated Integer partition(String topic, K key, V value, int numPartitions); /** * Determine the List of partition numbers to which a record, with the given key and value and the current number * of partitions, should be multi-casted to. Note that returning a single valued list with value -1 is a shorthand * for broadcasting the record to all the partitions of the topic. * * @param topic the topic name this record is sent to * @param key the key of the record * @param value the value of the record * @param numPartitions the total number of partitions * @return a List of integers between 0 and {@code numPartitions-1}, null would mean a broadcast to all partitions * while an empty list would mean the record would not be sent to any partitions. * the partitions of the topic. */ default List<Integer> partitions(String topic, K key, V value, int numPartitions) { return Collections.singletonList(partition(topic, key, value, numPartitions)); } }
Proposed Changes
- RecordCollector and it's implementation RecordCollectorImpl has the method which accepts a StreamPartitioner object and pushes it to the partitions returned by the partition method. This is the core piece of logic which would allow multi/broadcasting records. Here are the high level changes. Note that when the
partitions()
method return an empty list meaning we won't send the record to any partition, we would also be updating the droppedRecordsSensor.@Override public <K, V> void send(final String topic, final K key, final V value, final Headers headers, final Long timestamp, final Serializer<K> keySerializer, final Serializer<V> valueSerializer, final StreamPartitioner<? super K, ? super V> partitioner) { List<Integer> multicastPartitions; if (partitioner != null) { // Existing logic to fetch partitions // Changes to be made due to the KIP if (partitions.size() > 0) { multicastPartitions = partitioner.partitions(topic, key, value, partitions.size()); if (multicastPartitions.isEmpty()) { // New change. If a record is not to be sent to any partitions, mark it being // as a dropped record. log.info("Not sending the record to any partition"); droppedRecordsSensor.record(); } else { // broadcast if (multicastPartitions == null) { // Broadcast to all partitions multicastPartitions = partitions.stream().map(PartitionInfo::partition).collect(Collectors.toList()); } // iterate over the subset of partitions and send to those. for (int p: multicastPartitions) { send(topic, key, value, headers, p, timestamp, keySerializer, valueSerializer); } } } // Existing logic. }
- StreamPartitioner is also used in FK-join and Interactive queries. The invocation of
partition()
would be replaced bypartitions()
and a runtime check would be added to ensure that the returned list is singleton.
Example Usage
Continuing the example from the motivation section, with the new interface, here's how users can send to even number partitions:
// Define a partitioner class which sends to even numbered partitions public static class EvenPartitioner<K, V> implements StreamPartitioner<K, V> { @Override public Integer partition(String topic, K key, V value, int numPartitions) { return null; } @Override public List<Integer> partitions(String topic, K key, V value, int numPartitions) { final List<Integer> partitions = new ArrayList<>(); for (int i = 0; i < numPartitions; i += 2) { partitions.add(i); } return partitions; } } // Broadcasting public static class BroadcastingPartitioner<K, V> implements StreamPartitioner<K, V> { @Override public Integer partition(String topic, K key, V value, int numPartitions) { return null; } @Override public List<Integer> partitions(String topic, K key, V value, int numPartitions) { return null; } } // Don't send to any partitions public static class DroppingPartitioner<K, V> implements StreamPartitioner<K, V> { @Override public Integer partition(String topic, K key, V value, int numPartitions) { return null; } @Override public List<Integer> partitions(String topic, K key, V value, int numPartitions) { return Collections.emptyList(); } } // Build Stream. final KStream<String, String> inputStream = builder.stream("input-topic", Consumed.with(Serdes.String(), Serdes.String())); // Send to even numbered partitions inputStream.to("output-topic", Produced.with(Serdes.String(), Serdes.String(), (topic, key, val, numPartitions) -> new EvenPartitioner())); // Broadcase inputStream.to("output-topic", Produced.with(Serdes.String(), Serdes.String(), (topic, key, val, numPartitions) -> new BroadcastingPartitioner())); // Send to even numbered partitions inputStream.to("output-topic", Produced.with(Serdes.String(), Serdes.String(), (topic, key, val, numPartitions) -> new DroppingPartitioner())); }
This way users just need to define the partitioner and the internal routing mechanism would take care of sending the records to relevant or no partitions.
Compatibility, Deprecation, and Migration Plan
This KIP deprecates partition()
method in StreamPartitioner
. Here is what we would be doing as part of the same:
- StreamPartitioner is also used in FK-join and Interactive queries. The invocation of
partition()
would be replaced bypartitions()
and a runtime check would be added to ensure that the returned list is singleton. - Adding sufficient logging
Rejected Alternatives
The first 3 alternatives were rejected as they focussed only broadcasting to all partitions which seemed restrictive.
- extend to() / addSink() with a "broadcast" option/config in KafkaStreams.
- add toAllPartitions() / addBroadcastSink() methods in KafkaStreams.
- allow StreamPartitioner to return `-1` for "all partitions".
- Adding a separate class to handle multi casting. This was rejected in favour of enhancing the current
StreamPartitioner
interface.