Zerobus Ingest
Zerobus Ingest is a Databricks data engineering capability, introduced October 2025.
Push-based, serverless streaming API in Lakeflow Connect that writes records straight into Unity Catalog Delta tables over gRPC, REST, or OpenTelemetry, with no message bus to run.
- The name is the whole pitch: there is no bus. The docs sell it as creating a table and pushing data to it, with no brokers, partitions, or pipelines to manage.
- Ordering is guaranteed per stream and not globally across streams, so the obvious trick of spraying messages over several streams to go faster is also the way to lose your ordering.
- If it cannot flush buffered records into your table it holds them in durable storage for 28 days, then deletes them and emits a discard event logging exactly how many records and bytes it just dropped.
Limitations: Available only in select regions, with the workspace and target table both in a supported region. Delivery is at-least-once, it writes only to managed Delta tables and streaming tables, it does not auto-evolve your target table (evolve the table first, then send records against the new schema), it does not support recreating a target table, and partitioned tables allow at most 100 partitions in any 5-second interval. Records cap at 10,485,760 bytes. The default quotas (100 MB/s and 100,000 records/s per stream, 10 GB/s per target table, 10,000 REST requests/s) are adjustable through your Databricks account representative.
Open in REbricked →- Category
- Data engineering
- Also known as
- Zerobus, Zerobus Ingest connector
- Verified
- 2026-09-10
Sources
- Official Official Databricks / Microsoft docs
- Official Databricks docs: Use the Zerobus Ingest connector
- Official Databricks docs: Zerobus Ingest concepts (streams and ordering guarantees)
- Official Databricks docs: Zerobus Ingest connector limitations
- Official Databricks docs: Zerobus Ingest schema management (no auto-evolution)
- Official Databricks release notes: Zerobus Ingest Public Preview (October 2025)
- Official Databricks Blog: Zerobus Ingest GA (February 2026)
- Official Databricks release notes: Zerobus Ingest into streaming tables GA (August 2026)
- Official Databricks release notes: Apache Arrow support GA (September 1) and ingest into default storage Public Preview (September 7, 2026)