April 2026: Parallel Workers, DB2 Controls, and Analytics Enhancements
ELT & CDC
April's release cycle focused on accelerating initial syncs, expanding ad and analytics coverage, and tightening reliability across long-running CDC pipelines.
Introducing Parallel Workers for Initial Sync
Initial syncs from MySQL, PostgreSQL, Snowflake, and Redshift sources now run with parallel chunk workers, splitting large tables across multiple concurrent threads. The result is materially faster time-to-first-value when onboarding new sources or backfilling history, particularly on wide tables and large row counts where initial sync used to be the bottleneck before incremental CDC could take over.
New Features
- Custom Table List Fetch for DB2: Users can now select specific tables when configuring a DB2 source instead of pulling the full catalog at connection time, eliminating the schema-fetch timeouts that previously blocked setup against databases with very high table counts. This option is currently feature-flagged and off by default.
- ELT-to-ETL Initial Sync Script Generator for DB2 to BigQuery: The cross-product Initial Sync Script generator now supports DB2 (LUW) as a source and BigQuery as a destination, letting customers seed a BigQuery warehouse from DB2 through ETL and then hand off to ELT for ongoing CDC.
- Expanded Google Ads Coverage: The Google Ads connector now exposes additional fields on the ad_group_ads and keyword_view collections, including bidding strategy, negative keyword flags, and a broader set of performance metrics like clicks, impressions, conversions, and view-through conversions for richer paid-search reporting.
- Google Ads Incremental Reports: ACCOUNT_PERFORMANCE_REPORT, AD_PERFORMANCE_REPORT, and ADGROUP_PERFORMANCE_REPORT now sync incrementally instead of full-refresh, reducing API consumption and shortening sync windows for accounts with long reporting histories.
- Adobe Analytics OAuth Server-to-Server: The Adobe Analytics connector has migrated from the JWT authentication flow to OAuth Server-to-Server, keeping pipelines compliant with Adobe's authentication roadmap and avoiding the disruption of the JWT sunset.
- Trial Day Counter: Users without an active subscription now see remaining trial days in the dashboard, making it clearer how much evaluation time is left before a billing decision is needed.
Improvements
- The TikTok connector now consolidates all `pixel_instant_page_events` collections into a single collection, simplifying downstream modeling and removing the need to union near-duplicate streams.
- The MySQL connector's GTID handling has been reworked with a `gtidBaseline` value persisted in pipeline state, resolving connection failures when pipelines move to a new source server.
- The DB2 connector's Generate CDC SQL script generation has been parallelized and the timeout extended from one to five minutes, preventing setup timeouts on larger DB2 schemas.
- GA4 Export primary key uniqueness has been improved by including `user_pseudo_id`, `event_bundle_sequence_id`, and `batch_event_index` in the `event_custom_id` hash, eliminating duplicate-key collisions on high-volume GA4 streams.
Performance Enhancements
- DB2 to BigQuery initial sync has been parallelized end-to-end, cutting initial-load wall time on the path most commonly used to seed BigQuery warehouses from DB2 sources.
- The MySQL connector now uses a ten-second connection timeout by default, eliminating the TLS handshake failures that previously caused intermittent pipeline restarts on higher-latency source networks.
Take a look at what shipped this cycle and tell us which improvements you'd like to see go further.
about 1 month ago