On 2026‑03‑18, Microsoft announced the public preview of Fabric Mirroring integration for Azure Database for MySQL, enabling near real-time replication into Fabric OneLake without building ETL pipelines . The architecture isolates transactional workloads while creating analytics-ready Delta tables in Fabric .
Background:
Operational MySQL databases are not optimized for analytics; classic solutions rely on ETL pipelines or batch exports. The 2026 preview leverages logical replication + CDC patterns and Fabric’s OneLake + Delta surface .
How it works (architecture):
Initial snapshot is shipped as Parquet to OneLake; Fabric Replicator ingests the snapshot to create Delta tables; subsequent changes are batched and applied continuously .
Architecture (Mermaid):
flowchart LR
S[Azure MySQL source] –>|Snapshot Parquet| L[OneLake landing zone]
S –>|CDC batches| L
L –> R[Fabric Replicator]
R –> D[Delta mirrored tables]
D –> BI[BI/Analytics/AI workloads]
Step-by-step enablement + troubleshooting:
Prerequisites include supported tiers (General Purpose, Business‑Critical), MySQL 8.0 LTS minor versions, binlog settings (binlog_row_image FULL/NOBLOB), proper user permissions, and active Fabric capacity .
Enable in Azure portal (Fabric Mirroring preview) and then create the mirrored database in Fabric portal; table selection supports up to 1,000 tables .
For issues, validate server parameters and query error logs:
SELECT * FROM fabric_info.error_logs;
Use cases:
Near real-time BI, AI/ML enrichment in Fabric, and reduced operational overhead via zero‑ETL replication .
Limitations (2026 docs):
Tier/HA/ports constraints, single-database limitation per server, no post-setup table add/remove, requirement for primary keys, DDL disruptions, network restrictions for VNet-enabled servers, and unsupported data types .
