Spark on Amazon EMR with Apache Iceberg and OLake Go
Spark on Amazon EMR reads every Iceberg delete format. It can process the tables that OLake Go keeps in sync with an operational database.
Short answer. To read replicated tables on EMR, run OLake Go with AWS Glue as the Iceberg catalog, or with Amazon S3 Tables, Polaris, Lakekeeper or Hive Metastore. Point Spark on EMR at the same catalog. Deletion vectors need EMR 7.12 or later, which ships Iceberg 1.10.
Key takeaways
Last reviewed October 2026.
How OLake Go gets data into Spark on Amazon EMR
OLake Go uses change data capture (CDC) where the source has it and commits Iceberg tables to the catalog. Spark on EMR reads AWS Glue natively. S3 Tables works from EMR 7.5, and Polaris and Lakekeeper connect as REST (representational state transfer) catalogs.
No guide on this site runs EMR yet. The open-source Spark page covers the same catalog settings, and the Glue writer page covers the destination.
When to use it
- Spark jobs on AWS should transform or join the replicated tables.
- You already run EMR and want one copy of the data in Glue that Athena can also query.
When not to use it
- You only need SQL over the tables. Athena reads the same Glue tables with no cluster.
- The cluster runs a release before EMR 7.5 and the tables are in S3 Tables. The S3 Tables catalog needs EMR 7.5 or later.
Catalogs Spark on Amazon EMR reads
| Catalog | Spark on Amazon EMR reads | How |
|---|---|---|
| AWS Glue | Yes | native: Glue Data Catalog |
| Apache Polaris | Yes | REST catalog |
| Lakekeeper | Yes | REST catalog |
| Amazon S3 Tables | Yes | S3 Tables catalog (EMR 7.5 and later) |
| Hive Metastore | Yes | Hive Metastore |
Delete formats Spark on Amazon EMR reads
| Equality deletes (v2) | Positional deletes (v2) | Deletion vectors (v3) |
|---|---|---|
| Yes | Yes | Yes |
Spark on EMR reads every delete format. Deletion vectors need EMR 7.12 or later (Iceberg 1.10). Set Spark on Amazon EMR as a target query engine on the OLake Go job and OLake Go picks a format it reads; with Spark on Amazon EMR as the only reader that is positional deletes (v2).
Sources: EMR: Iceberg release history (checked 2026-10-12).
Tested guides for Spark on Amazon EMR
No guide ends in Spark on Amazon EMR yet.
Frequently asked questions
Which EMR release reads deletion vectors?
EMR 7.12 and later, which ship Iceberg 1.10.
Which catalogs can Spark on EMR use?
AWS Glue natively, Amazon S3 Tables from EMR 7.5, Hive Metastore, and Polaris or Lakekeeper as REST catalogs.
Is there a tested guide for EMR?
Not yet. The Spark checks on this site use the same catalogs with open-source Spark. See the Apache Spark page.