September 11, 2026 - v3.5.0
Download the Spark Connector JAR.
[v3.5.0-v6.4.0]⚓︎
Added⚓︎
- Added the optional configuration setting
api_version. By default, the connector uses Config API v2 when the cluster serves v2 endpoints and falls back to v1. Set it to1to use v1. - Added the optional configuration setting
multi_org, off by default, for use with Hydrolix multi-org deployments. - Added read support for the
uuid,ip, andjsoncolumn types. All three surface to Spark as strings in the query engine's canonical text form, and comparisons,ORDER BY, andMIN/MAX/COUNTon these columns produce the same results whether pushed down or evaluated by Spark. Aggregations that group by a json column aren't pushed down; Spark evaluates them.
Changed⚓︎
- Bundled Turbine server updated to v6.4.0.
- The account used by the connector now needs
view_projectandview_hdxstorageRBAC permissions at the global level in addition toview_org. See the required user permissions for your deployment platform. - Table schemas and primary keys are now sourced from the query engine instead of the Config API.
Fixed⚓︎
- Fixed a legacy
api_urlending in/config/v1/failing with 404 errors against clusters that serve Config API v2 whenapi_versionis unset. The connector now adjusts the URL to the version served by the cluster. - Fixed the API session cache re-authenticating on every query in long-running sessions, which could surface as socket timeouts.
- Fixed aggregation pushdown accepting
GROUP BYexpressions it can't compile, which produced a query error. Such queries are now aggregated by Spark. - Updated the bundled Jackson and Log4j libraries to close dependency vulnerabilities.