Hazelcast Jet 4.5.4 is based on IMDG version 4.2.6.
[core] Fixed an issue where a Jet job was stuck when there is a Near Cache configured at the target map.
[dependencies] Updated Hazelcast IMDG to 4.2.6
[dependencies] Updated jackson to 2.14.1
[dependencies] Updated snakeyaml to 1.33
[dependencies] Updated hadoop to 3.3.4
[dependencies] Updated parquet to 1.12.3
[dependencies] Updated jetty to 9.4.49.v20220914
[dependencies] Updated hazelcast-aws to 3.4
[dependencies] Updated AWS SDK to 1.12.352
[dependencies] Updated hazelcast-kubernetes to 2.2.3
[dependencies] Updated hazelcast-azure to 2.1.2
[dependencies] Updated grpc to 1.48.2
[dependencies] Updated protobuf to 3.19.6
[dependencies] Updated postgresql driver to 42.5.1
[dependencies] Updated jackson-mapper-asl to 1.9.13
[dependencies] Updated google oauth client to 1.34.1
[dependencies] Updated lucene-queryparser to 6.6.5 (in elasticsearch-5 module)
[distribution] Removed the hazelcast-jet-cdc-mysql module from the Jet Enterprise distribution
Hazelcast Jet 4.5.2 is based on IMDG version 4.2.3 .
[dependencies] Update log4j to 2.16.0
[dependencies] Update slf4j to 1.7.32
[dependencies] Update jackson to 2.11.4
[dependencies] Update snakeyaml to 1.29
[dependencies] Update netty to 4.1.72.Final
[dependencies] Update mysql-connector-java to 8.0.25
[dependencies] Update aws-sdk to 1.12.128
[dependencies] Update avro to 1.11.0
Hazelcast Jet 4.5.1 is based on IMDG version 4.2.2 .
[core] Fix rebalance() nonfunctional before fused stages (#3051)
[hadoop] Update Apache Commons Compress to 1.21
[hadoop] Update JSON Small and Fast Parser to 2.4.7
[hadoop] Update Jetty to 9.4.43
[hadoop] Apache Commons IO to 2.7
[metrics] Update Prometheus JMX Exporter - Java Agent to 0.16.1
[elasticsearch] Update Elasticsearch6 to 6.8.17
- @hhromic added an option to configure the batch size for JDBC Sink.
batchLimit()method added to the
corresponding constructor arguments to the underlying implementation
class and meta-suppliers.
- Limit parallel partition reads: In Map/Cache source, Jet tries to
read all the partitions in parallel, which may lead to high memory
usage. Now the max number of parallel reads is limited.
- Security updates: updated several libraries with identified security
- SQL: pre-existing
ClassDefinitionis no longer required for
Portableserialization - if it does not exist, it's derived from the
[avro] Update Apache Commons Compress to 1.20 (#2953)
[avro] Update avro version to 1.10.1 (#2937)
[cdc] Update MySQL Connector to 8.0.20 (#2953)
[core] Update Jackson version to 2.11.4 (#2938)
[elasticsearch] Update ElasticSearch-6 to 6.8.14 (#2946)
[elasticsearch] Update ElasticSearch-7 to 7.10.2 (#2945, #2956)
[elasticsearch] Update Snakeyaml version to 1.26 (#2949)
[hadoop] Update Apache Httpclient to 4.5.13 (#2953)
[hadoop] Update Guava to 30.1-jre (#2953)
[kafka] Update Kafka to 2.2.2 (#2948)
[s3] Update Netty version to 4.1.59.Final (#2943)
- @hhromic improved the naming of source and sink stages across different
connectors, bringing them all in line with the same convention
- @TomaszGaweda added the
pipeline.isEmpty()method that tells whether
it contains any stage
- SQL Beta: submit jobs to Jet from the command-line SQL shell.
- Unified API to create sources and sinks from file-like resources: local
filesystem, Amazon S3, Azure Blob Storage and Data Lake Storage, Google
- Amazon Kinesis connector
- Prevent event reordering: by default Jet reorders data for performance,
now you can disable this to get strict event order where you need it.
- @TomaszGaweda added the
config property that disables the enforcement of the exact IMDG
dependency version. This allows adding IMDG quick fixes to the existing
- New packaging: download either the full package with all the extensions
enabled, or the minimal package and separately download the extensions
- Improved the behavior of
jet submit: now it waits for the job to
start and prints a message about it.
- Improved the error message when using a Python function but Python
is not installed.
- Improved the performance of the Kafka source by fine-tuning some
- @MohamedMandouh implemented distributed sorting:
- @guenter-hesse added a high-throughput source with precisely controlled
timing, to be used to benchmark Jet's throughput and latency.
- We added the ability for Jet to suspend an unbounded, fault-tolerant job
when it fails due to an exception in the pipeline code. This allows you
to diagnose the problem, remove its cause, and resume the pipeline without
- We have improved the consistency of reconnect behaviour across CDC sources and
introduced a uniform API to configure the reconnect strategy.
- We improved the behavior of streaming pipelines based on ingestion time: a
low event rate in any partition of the source used to stall the progress,
now the watermark advances even without events.
- We moved the
DAGtransformation from the client to the server,
in preparation for the upcoming feature of preventing event reordering in the
- We added Change Data Capture (CDC) support for PostgreSQL and MySql. New
connectors use Debezium to provide a high-level API for processing the
stream of changes from relational databases. A new map sink can be used to
automatically keep the map in-sync with changes from the database.
- We added new source and sink connectors for Elasticsearch 5, 6 and 7
with support for slicing and co-located reading for maximum read performance
and extended Elasticsearch's client retry mechanism for reliable writing.
- We improved JSON support through a new file source which can read JSON files
directly and also added built-in Jet methods for parsing and mapping JSON strings.
- We introduced .relabance() operator which can be used for distributing items
across the cluster.
- We made some improvements to Jet startup scripts:
- You can now use the PROMETHEUS_PORT to start Jet with prometheus metrics enabled.
- You can use the jvm.options file to edit the startup JVM parameters.
- You can now use the JET_MODULES environment variable to add modules when using docker
or jet-start script.
- We did several rounds of improvements to the documentation:
- Improvements to Docker and Kubernetes instructions
- New guidance on Garbage Collector choice and settings
- New and improved tutorials for CDC
- New documentation for Python module and early results support in windowing