Sink data from RisingWave to Apache Iceberg
This guide describes how to sink data from RisingWave to Apache Iceberg using the Iceberg sink connector in RisingWave.
In RisingWave, you can stream data into Iceberg tables with the built-in Iceberg sink connector.
Apache Iceberg is a table format designed to support huge tables. For more information, see Apache Iceberg.
Prerequisites
- Ensure you already have an Iceberg table that you can sink data to. For additional guidance on creating a table and setting up Iceberg, refer to this quickstart guide on creating an Iceberg table.
- Ensure you have an upstream materialized view or source that you can sink data from.
Syntax
Parameters
For object storage configuration (S3, GCS, Azure Blob) and catalog configuration parameters, see:
Sink-specific parameters
Parameter name | Description |
---|---|
type | Required. Allowed values: append-only and upsert . |
force_append_only | Optional. If true, forces the sink to be append-only, even if it cannot be. |
database.name | Required. The database of the target Iceberg table. |
table.name | Required. The name of the target Iceberg table. |
primary_key | The primary key for an upsert sink. It is only applicable to the upsert mode. |
partition_by | Optional. Specify partitioning using column names or transformations. Supported formats include: column , transform(column) , transform(n,column) , and transform(n, column) . Transformations can include functions like bucket or truncate , where n is an optional parameter. Ensure that the specified partition fields exist in the schema. |
commit_checkpoint_interval | Optional. Commit every N checkpoints (N > 0). Default value is 10. The behavior of this field also depends on the sink_decouple setting:
time = barrier_interval_ms × checkpoint_frequency × commit_checkpoint_interval . barrier_interval_ms and checkpoint_frequency are system parameters that define the base checkpointing rate; commit_checkpoint_interval is configurable in the Iceberg sink. |
commit_retry_num | Optional. The number of times to retry a commit when an Iceberg commit fails. Default is 8. |
create_table_if_not_exists | Optional. When set to true , it will automatically create a table for the Iceberg sink. |
is_exactly_once | Optional. When set to true , enables exactly-once delivery semantics for the Iceberg sink. Default is false . |
Data type mapping
RisingWave converts RisingWave data types from/to Iceberg according to the following data type mapping table:
RisingWave Type | Iceberg Type |
---|---|
boolean | boolean |
int | int |
smallint | int |
bigint | long |
real | float |
float | float |
double | double |
varchar | string |
date | date |
timestamptz | timestamptz |
timestamp | timestamp |
map | map |
array | list |
struct | struct |
jsonb | string |
Iceberg table format
Currently, RisingWave only supports Iceberg tables in format v2.
Exactly-once delivery
RisingWave provides exactly-once delivery semantics for Iceberg sinks. This semantics guarantees that each data event is processed once and only once, even in the presence of failures such as retries or restarts. This level of delivery assurance is essential in scenarios where duplicate records can lead to incorrect analytics or data corruption in downstream systems.
Exactly-once delivery is achieved through a two-phase commit protocol involving a pre-commit phase and a commit phase. Iceberg’s commit operations are idempotent, which allows RisingWave to safely retry failed transactions without introducing duplicates.
By default, exactly-once semantics is disabled. To enable it for an Iceberg sink, include is_exactly_once = 'true'
in the WITH
clause of the sink definition. Note that enabling this option introduces additional coordination overhead due to metadata pre-commit, which may impact sink performance in high-throughput workloads.
Examples
This section includes several examples that you can use if you want to quickly experiment with sinking data to Iceberg.
Create an Iceberg table (if you do not already have one)
Set create_table_if_not_exists
to true
to automatically create an Iceberg table.
Alternatively, use Spark to create a table. For example, the following spark-sql
command creates an Iceberg table named table
under the database dev
in AWS S3. The table is in an S3 bucket named my-iceberg-bucket
in region ap-southeast-1
and under the path path/to/warehouse
. The table has the property format-version=2
, so it supports the upsert option. There should be a folder named s3://my-iceberg-bucket/path/to/warehouse/dev/table/metadata
.
Note that only S3-compatible object store is supported, such as AWS S3 or MinIO.
Create an upstream materialized view or source
The following query creates an append-only source. For more details on creating a source, see CREATE SOURCE .
Another option is to create an upsert table, which supports in-place updates. For more details on creating a table, see CREATE TABLE .
Append-only sink from append-only source
If you have an append-only source and want to create an append-only sink, set type = append-only
in the CREATE SINK
SQL query.
Append-only sink from upsert source
If you have an upsert source and want to create an append-only sink, set type = append-only
and force_append_only = true
. This will ignore delete messages in the upstream, and to turn upstream update messages into insert messages.
Upsert sink from upsert source
In RisingWave, you can directly sink data as upserts into Iceberg tables.