default.client.supplier that allows to use a custom KafkaClientSupplier without any code changes.
+ + RocksDB offers many metrics which are critical to monitor and tune its performance. Kafka Streams started to make RocksDB metrics accessible + like any other Kafka metric via KIP-471 in 2.4.0 release. + However, the KIP was only partially implemented, and is now completed with the 3.2.0 release. + For a full list of available RocksDB metrics, please consult the monitoring documentation. +
+ +
+ Kafka Streams ships with RocksDB and in-memory store implementations and users can pick which one to use.
+ However, for the DSL, the choice is a per-operator one, making it cumbersome to switch from the default RocksDB
+ store to in-memory store for all operators, especially for larger topologies.
+ KIP-591
+ adds a new config default.dsl.store that enables setting the default store for all DSL operators globally.
+ Note that it is required to pass TopologyConfig to the StreamsBuilder constructor to make use of this new config.
+
+ For multi-AZ deployments, it is desired to assign StandbyTasks to a KafkaStreams instance running in a different
+ AZ than the corresponding active StreamTask.
+ KIP-708
+ enables configuring Kafka Streams instances with a rack-aware StandbyTask assignment strategy, by using the new added configs
+ rack.aware.assignment.tags and corresponding client.tag.<myTag>.
+
+ KIP-791
+ adds a new method Optional<RecordMetadata> StateStoreContext.recordMetadata() to expose
+ record metadata. This helps for example to provide read-your-writes consistency guarantees in interactive queries.
+
+ Interactive Queries allow users to
+ tap into the operational state of Kafka Streams processor nodes. The existing API is tightly coupled with the
+ actual state store interfaces and thus the internal implementation of state store. To break up this tight coupling
+ and allow for building more advanced IQ features,
+ KIP-796 introduces
+ a completely new IQv2 API, via StateQueryRequest and StateQueryResult classes,
+ as well as Query and QueryResult interfaces (plus additional helper classes).
+
+ In addition, multiple built-in query types were added: KeyQuery for key lookups and
+ RangeQuery (via KIP-805)
+ for key-range queries on key-value stores, as well as WindowKeyQuery and WindowRangeQuery
+ (via KIP-806)
+ for key and range lookup into windowed stores.
+
+ The Kafka Streams DSL may insert so-called repartition topics for certain DSL operators to ensure correct partitioning
+ of data. These topics are configured with infinite retention time, and Kafka Streams purges old data explicitly
+ via "delete record" requests, when commiting input topic offsets.
+ KIP-811
+ adds a new config repartition.purge.interval.ms allowing you to configure the purge interval independently of the commit interval.
+
The semantics of left/outer stream-stream join got improved via