Flink Autopilot no longer pauses statements with Java UDFs while scaling
Autopilot keeps statements that use Java UDFs processing data while it scales them. Existing statements get the behavior after their next restart; Python UDFs are unaffected.
Autopilot keeps statements that use Java UDFs processing data while it scales them. Existing statements get the behavior after their next restart; Python UDFs are unaffected.
Single sign-on configuration now has its own page in Cloud Console, separate from Accounts & access, where group mapping configuration remains.
The fully managed Amazon S3 Sink and Snowflake Sink connectors now connect to destinations in a different region or cloud without contacting Confluent Support.
The AWS Lambda Sink connector supports STRING as an input.data.format value and a new input.key.format setting, defaulting to BYTES.
The Flink sqlv1 Statement and Materialized Table REST APIs make the Confluent-Identity-Pool-Id header optional, automatically mapping the OAuth token to matching identity pools.
Snapshot queries are GA on AWS, Azure, and Google Cloud. Run bounded, batch-mode Flink SQL queries that snapshot streaming data, compute a result, and exit.
The Java Table API is GA with plugin 2.3-3 on Flink 2.3.0, adding global API keys, OAuth 2.0 with identity pools, and statement lifecycle management. The Python Table API remains in Open Preview.
New regions: Azure italynorth, mexicocentral, southafricanorth, and Google Cloud europe-southwest1 (Madrid), all with public and private networking.
The XStream capture process can run on a separate downstream Oracle database, with downstream capture and cascading downstream capture topologies, in real-time and archived-log modes.
Gateway validates messages and enforces data governance with Schema Registry, with four policy types (schema ID, deep validation, field-level and full-payload encryption) and per-topic overrides. Docker only during EA.
Opt-in safeguard that blocks deletion of a Kafka cluster until an authorized user disables it. Available via Console, CLI, REST API, or Terraform.
The mTLS certificate revocation list size limit drops from 10 MB to 1.5 MB per CRL, and public CAs no longer support issuing client certificates.
The metric now reports lag for any group in the EMPTY state for less than one day, rolling out through late August 2026.
Dedicated clusters with Premier support or higher can enable release priority to receive updates before non-test clusters.