If your analytics warehouse on Greenplum is hitting disk-scaling walls or growing painful to maintain, moving to a modern Lakehouse architecture with Apache Iceberg has probably come up in architecture reviews more than once. Between the clean pitch of open table formats and live production, however, sits messy reality: terabytes of critical data, dozens of Greenplum-specific functions, and a hard business requirement not to interrupt data mart builds for a single minute.
I picked this webinar from the VK Tech team because its agenda skips vague marketing promises. Instead, it walks through the hands-on engineering mechanics of migrating: translating SQL dialects with SQLGlot, splitting stored procedures between Apache Spark and Airflow, verifying checksums across old and new engines, and switching production safely via a shadow run.
Greenplum → Lakehouse: zero-downtime migration
- Organizer: VK Tech.
- Speakers: Nikita Surovegin (Data Services Analyst, VK Tech) and Pavel Savchenko (Head of Data Services Presales, VK Tech).
- When: Wednesday, November 25, 2026, at 16:00 Moscow time (MSK, UTC+3).
- Format & terms: online webinar, free attendance, advance registration required on the organizer's platform.
- Who it's for: data engineers, database architects, DWH developers, and engineering leaders (CTO, Head of Data, CDO) planning to modernize an analytics stack.
- Agenda:
- Architectural gap: how Greenplum works under the hood (coordinator, segments, mirrors) and why segment distribution doesn't map directly to a data lake — requiring a shift to Apache Iceberg partitioning and sorting.
- Data preservation & transformation: converting heap, append-optimized, and columnar tables to Parquet, replacing external tables and loaders with ELT pipelines, and reading directly from Greenplum segments via the native CedrusData Engine connector.
- Migrating logic & pipelines: converting declarative SQL and views with SQLGlot, splitting procedures across the SQL engine, Apache Spark jobs, and Apache Airflow DAGs, migrating cron jobs, and replacing specialized extensions (MADlib, PostGIS, GPText).
- Safe zero-downtime cutover: running a shadow run, comparing checksums and schemas on old vs. new engines, benchmarking query performance, and setting up CDC plus incremental catch-up loads before final switchover.
- Downstream integration: reconnecting Apache Superset dashboards and analytics tools via Trino connectors and configuring role-based access controls.
- Registration: you can sign up for the broadcast on the official VK Tech webinar page.
