Write to Apache Iceberg Tables with SQL in Spice

Data Acceleration

SQL Federation

Spice AI Team

Editorial Team at Spice AINovember 18, 2025
Write to Apache Iceberg tables with SQL in Spice blog post header image

TL;DR: Spice v1.8 adds native Apache Iceberg write support using standard SQL INSERT INTO statements. Write query results, transformed data, or new records directly to Iceberg tables from the same runtime used for federation and acceleration, no separate ETL pipeline required.

With the release of Spice v1.8, developers can now write directly to Apache Iceberg tables and catalogs using standard SQL INSERT INTO statements. 

This feature extends Spice's SQL federation capabilities beyond reads, enabling data ingestion, transformation, and pipeline workloads to write results back into Iceberg directly from the same runtime used for queries and acceleration.

What sets Spice apart from other query engines is its broader, application-focused feature set designed for modern data and AI workloads. Spice brings together federation, hybrid search, embedded LLM inference, and now native writes in one unified runtime - enabling teams to build complete, end-to-end workflows without the management overhead and performance concessions of using multiple systems. 

Iceberg write support is available in preview, with append-only operations and schema validation for secure and predictable data management.

From Read-Only Federation to Full Data Workflows

Data teams are standardizing on open table formats like Apache Iceberg to unify analytical and operational data across systems; Iceberg offers a consistent way to store, version, and manage data across different engines and clouds, helping teams avoid vendor lock-in while maintaining strong governance and interoperability.

Supporting Iceberg writes natively inside Spice means development teams can:

  • Direct Writes to Iceberg without ETL: Insert data directly into Iceberg from SQL queries.
  • Simplify ingestion paths: Load transformed or federated data into Iceberg without separate tools.
  • Enforce governance: Maintain schema validation and secure access through read_write permissions.

Paired with Spice's built-in SQL federation and acceleration, these write capabilities make it easier to use Iceberg not just as a storage solution, but as a queryable data layer for both operational and AI workloads

How It Works

Spice supports INSERT_INTO statements on Iceberg tables and catalogs explicitly marked as read_write.

Example Spicepod configuration:

catalogs:
  - from: iceberg:http://localhost:8181/v1/namespaces
    access: read_write  # Uncomment this line
    name: ice
    params:
      iceberg_s3_endpoint: http://localhost:9000
      iceberg_s3_access_key_id: admin
      iceberg_s3_secret_access_key: password
      iceberg_s3_region: us-east-1

And, here's an example SQL query:

-- Insert from another table
INSERT INTO iceberg_table
SELECT * FROM existing_table;

-- Insert with values
INSERT INTO iceberg_table (id, name, amount)
VALUES (1, 'John', 100.0), (2, 'Jane', 200.0);

-- Insert into catalog table
INSERT INTO ice.sales.transactions
VALUES (1001, '2025-01-15', 299.99, 'completed');

Support for updates, deletes, and merges will be added in future releases.

Now, let's walk through an end-to-end workflow demonstrating how to execute Iceberg writes in Spice.

Write to Iceberg Tables with Spice Cookbook

Prerequisites: 

  • Access to an Iceberg catalog, or Docker to run an Iceberg catalog locally.
  • Spice is installed (see the Getting Started documentation).

Step 1: Create a new directory and initialize a Spicepod

mkdir iceberg-catalog-recipe
cd iceberg-catalog-recipe
spice init

Step 2. Run the Docker container for the Iceberg catalog

In a separate terminal, clone the cookbook repository and run the Docker container for the Iceberg catalog.

git clone https://github.com/spiceai/cookbook.git
cd cookbook/catalogs/iceberg
docker compose up -d

Step 3. Add the Iceberg Catalog Connector to your Spicepod

catalogs:
  - from: iceberg:http://localhost:8181/v1/namespaces
    # access: read_write
    name: ice
    params:
      iceberg_s3_endpoint: http://localhost:9000
      iceberg_s3_access_key_id: admin
      iceberg_s3_secret_access_key: password
      iceberg_s3_region: us-east-1

Step 4. Run Spice

spice run

2025/01/27 11:08:36 INFO Checking for latest Spice runtime release...
2025/01/27 11:08:37 INFO Spice.ai runtime starting...
2025-01-27T19:08:37.494155Z  INFO runtime::init::dataset: No datasets were configured. If this is unexpected, check the Spicepod configuration.
2025-01-27T19:08:37.494905Z  INFO runtime::init::catalog: Registering catalog 'ice' for iceberg
2025-01-27T19:08:37.499162Z  INFO runtime::metrics_server: Spice Runtime Metrics listening on 127.0.0.1:9090
2025-01-27T19:08:37.499174Z  INFO runtime::flight: Spice Runtime Flight listening on 127.0.0.1:50051
2025-01-27T19:08:37.500689Z  INFO runtime::http: Spice Runtime HTTP listening on 127.0.0.1:8090
2025-01-27T19:08:37.503376Z  INFO runtime::opentelemetry: Spice Runtime OpenTelemetry listening on 127.0.0.1:50052
2025-01-27T19:08:37.696469Z  INFO runtime::init::results_cache: Initialized results cache; max size: 128.00 MiB, item ttl: 1s
2025-01-27T19:08:37.697178Z  INFO runtime::init::catalog: Registered catalog 'ice' with 1 schema and 8 tables

Step 5. Query the Iceberg catalog

spice sql
sql> show tables;
+---------------+--------------+--------------+------------+
| table_catalog | table_schema | table_name   | table_type |
+---------------+--------------+--------------+------------+
| ice           | tpch_sf1     | lineitem     | BASE TABLE |
| ice           | tpch_sf1     | nation       | BASE TABLE |
| ice           | tpch_sf1     | orders       | BASE TABLE |
| ice           | tpch_sf1     | supplier     | BASE TABLE |
| ice           | tpch_sf1     | customer     | BASE TABLE |
| ice           | tpch_sf1     | partsupp     | BASE TABLE |
| ice           | tpch_sf1     | region       | BASE TABLE |
| ice           | tpch_sf1     | part         | BASE TABLE |
| spice         | runtime      | task_history | BASE TABLE |
| spice         | runtime      | metrics      | BASE TABLE |
+---------------+--------------+--------------+------------+

Run Pricing Summary Report Query (Q1). More information about TPC-H and all the queries involved can be found in the official TPC Benchmark H Standard Specification.

select
  l_returnflag,
  l_linestatus,
  sum(l_quantity) as sum_qty,
  sum(l_extendedprice) as sum_base_price,
  sum(l_extendedprice * (1 - l_discount)) as sum_disc_price,
  sum(l_extendedprice * (1 - l_discount) * (1 + l_tax)) as sum_charge,
  avg(l_quantity) as avg_qty,
  avg(l_extendedprice) as avg_price,
  avg(l_discount) as avg_disc,
  count(*) as count_order
from
  ice.tpch_sf1.lineitem
where
  l_shipdate <= date '1998-12-01' - interval '110' day
group by
  l_returnflag,
  l_linestatus
order by
  l_returnflag,
  l_linestatus
;

Output:

+--------------+--------------+-------------+-----------------+-------------------+---------------------+-----------+--------------+----------+-------------+
| l_returnflag | l_linestatus | sum_qty     | sum_base_price  | sum_disc_price    | sum_charge          | avg_qty   | avg_price    | avg_disc | count_order |
+--------------+--------------+-------------+-----------------+-------------------+---------------------+-----------+--------------+----------+-------------+
| A            | F            | 37734107.00 | 56586554400.73  | 53758257134.8700  | 55909065222.827692  | 25.522005 | 38273.129734 | 0.049985 | 1478493     |
| N            | F            | 991417.00   | 1487504710.38   | 1413082168.0541   | 1469649223.194375   | 25.516471 | 38284.467760 | 0.050093 | 38854       |
| N            | O            | 73416597.00 | 110112303006.41 | 104608220776.3836 | 108796375788.183317 | 25.502437 | 38249.282778 | 0.049996 | 2878807     |
| R            | F            | 37719753.00 | 56568041380.90  | 53741292684.6040  | 55889619119.831932  | 25.505793 | 38250.854626 | 0.050009 | 1478870     |
+--------------+--------------+-------------+-----------------+-------------------+---------------------+-----------+--------------+----------+-------------+

Time: 0.186233833 seconds. 10 rows.

Step 6. Write to Iceberg tables

To enable write operations to Iceberg tables, uncomment the access: read_write configuration and restart Spice.

6.1. Update the Spicepod configuration

Edit the spicepod.yaml file to uncomment the access line:

catalogs:
  - from: iceberg:http://localhost:8181/v1/namespaces
    access: read_write  # Uncomment this line
    name: ice
    params:
      iceberg_s3_endpoint: http://localhost:9000
      iceberg_s3_access_key_id: admin
      iceberg_s3_secret_access_key: password
      iceberg_s3_region: us-east-1

6.2. Restart Spice

Stop the current Spice instance (Ctrl+C) and restart it:

spice run

6.3. Insert data into Iceberg tables

Now you can write data to the Iceberg tables using SQL INSERT statements:

spice sql

‍Example: Insert a new region into the region table:

INSERT INTO ice.tpch_sf1.region (r_regionkey, r_name, r_comment)
VALUES (5, 'ANTARCTICA', 'A cold and remote region');

+-------+
| count |
+-------+
| 1     |
+-------+

Example: Insert a new nation into the nation table:

INSERT INTO ice.tpch_sf1.nation (n_nationkey, n_name, n_regionkey, n_comment)
VALUES (25, 'PENGUINIA', 5, 'A vibrant home for brave penguins in Antarctica');

+-------+
| count |
+-------+
| 1     |
+-------+

‍Verify the inserts by querying the tables:

SELECT * FROM ice.tpch_sf1.region WHERE r_regionkey = 5;
SELECT * FROM ice.tpch_sf1.nation WHERE n_nationkey = 25;

Step 7. View the Iceberg tables in MinIO

Navigate to http://localhost:9001 and login with admin and password. View the iceberg bucket to see the created Iceberg tables.

Step 8. Clean up

docker compose down --volumes --rmi local

Next steps with Iceberg writes in Spice

Iceberg write support is available in preview. See the Iceberg connector docs for configuration details and try the Iceberg Catalog Connector recipe to get started. 

Feedback is welcome as we round out support for Iceberg writes in upcoming releases! 

SQL writes complete the operational data lakehouse loop: read, accelerate, and write back with one engine. To try Iceberg writes on your lakehouse, get a demo.

Frequently Asked Questions

Do you need Apache Spark to write to Apache Iceberg tables?

No. Apache Iceberg is engine-agnostic, and Spice writes to Iceberg tables with standard SQL INSERT INTO statements from a single runtime, so teams that only need SQL ingestion can skip running a Spark cluster for writes. For background on how the table format itself works, see the guide to Apache Iceberg.

Can Spice write data from other databases into Iceberg tables?

Yes. An INSERT INTO ... SELECT statement can read from any connected data source and write the results into an Iceberg table, because writes run on the same SQL federation engine that handles reads. This moves data from operational databases or object storage into Iceberg without a separate ingestion tool.

Can other query engines read the data Spice writes to Iceberg?

Yes. Spice commits writes through the standard Iceberg protocol: new Parquet data files land in object storage, and the catalog metadata pointer updates atomically. Any Iceberg-compatible engine reading from the same catalog sees the new snapshot. The Apache Iceberg at Spice AI deep dive covers the commit path in detail.

Can you update or delete rows in an Iceberg table with Spice?

Not yet. Iceberg write support in Spice v1.8 is in preview and append-only, so INSERT INTO is the supported operation. Support for updates, deletes, and merges is planned for future releases.

What happens if an INSERT does not match the Iceberg table schema?

Spice validates every insert against the target table's schema before writing. Statements with mismatched columns fail validation instead of writing malformed data, which keeps tables predictable for downstream readers.

Why write to Iceberg tables instead of plain Parquet files on object storage?

Writing through an Iceberg catalog adds ACID transactions, schema evolution, and snapshot-based time travel on top of the same Parquet files, so concurrent readers and writers see consistent results. It also keeps table metadata accurate, which lets query engines prune data files from metadata statistics instead of scanning entire directories.

Can you query data immediately after writing it to an Iceberg table?

Yes. Rows committed with INSERT INTO are queryable from the same runtime as soon as the statement completes, so a SELECT issued right after the insert returns the new records. The walkthrough in this post verifies each insert exactly this way.

Explore more Spice resources

Tutorials, docs, and blog posts to help you go deeper with Spice.

Blog

Real-Time Control Plane Acceleration with DynamoDB Streams 

How to sync DynamoDB data to thousands of nodes with sub-second latency using a two-tier architecture with DynamoDB Streams and Spice acceleration.

Real-Time Control Plane Acceleration with DynamoDB Streams 
Blog

How we use Apache DataFusion at Spice AI

Why we chose to build on DataFusion and how we extended it with custom TableProviders, optimizer rules, and UDFs for federated SQL

How we use Apache DataFusion at Spice AI

See Spice in action

Walk through your use case with an engineer and see how Spice handles federation, acceleration, and AI integration for production workloads.

Talk to an engineer

Share
twitter logolinkedin logomailto logo