For installation instructions check out the getting started guide.
Added
- kafka: Added optional tombstone field to franz-based Kafka outputs (redpanda, kafka_franz) to support producing compacted topic deletes and null values. (@squiidz, #4959)
- oracledb_cdc: Exposed rs_id and ssn metadata from Oracle LogMiner to uniquely identify row changes in transactions, following Debezium's field names. (@Leward, #4826)
- oracledb_cdc: Added row_seq metadata to disambiguate array DELETE rows that share the same rs_id and ssn, making the (rs_id, ssn, row_seq) triple unique per row change. (@Leward, #4827)
- redpanda: Added partition_assignment_strategy field to redpanda input to control consumer group partition balancing across cooperative-sticky, sticky, range, and roundrobin strategies. (@alextreichler, #4923)
- sap_hana: Added SAP HANA polling input with bulk, incrementing, timestamp, and query modes, supporting schema metadata discovery, high-water mark checkpointing, and LOB/binary column handling. (@squiidz, #4775)
Fixed
- couchbase: Couchbase output now correctly fails only messages whose individual operations failed, instead of acking all messages regardless of operation errors. (@Leward, #4969)
- redpanda_migrator: Schema registry sync no longer fails the entire output on per-subject registration errors; partial sync failures are now retried on the next interval. (@rjustice-rp, #4895)
- schema_registry_encode: Optional fields in json_schema mode now accept null values, matching Avro conversion behavior for NULL values from CDC inputs. (@JakeSCahill, #4950)
- oracledb_cdc: replay open LOB transactions at window boundaries. (@josephwoodward, #4964)
Changed
- general: Shared field descriptions from benthos internals as exported constants and gave component-specific fields their own descriptions. (@JakeSCahill, #4877)
- general: Updated Benthos dependency to v4.83.0 to pull in credential redaction for URLs and httpclient retry fixes. (@Jeffail, #4972)
- kafka: Clarified compression field behavior when unset and set default to snappy to match franz-go's actual default preference. (@JakeSCahill, #4951)
- oracledb_cdc: Improved oracledb_cdc parsing and message processing with better row type handling, inline read-ahead scaling, and cleaner error handling. (@josephwoodward, #4941)
- prometheus: Added documentation for mixed use_histogram_timing settings across fleets and histogram_timing_seconds_suffix workaround for remote-write receivers. (@squiidz, #4946)
- sql: SQL integration tests now run in parallel with shared database containers, reducing overall test package execution time. (@Leward, #4874)
The full change log can be found here.