Skip to content

Latest commit

 

History

History
121 lines (90 loc) · 14.4 KB

databricks.md

File metadata and controls

121 lines (90 loc) · 14.4 KB
Raw

Databricks Lakehouse

Overview

This destination syncs data to Delta Lake on Databricks Lakehouse. Each stream is written to its own delta-table.

This connector requires a JDBC driver to connect to the Databricks cluster. By using the driver and the connector, you must agree to the JDBC ODBC driver license. This means that you can only use this connector to connect third party applications to Apache Spark SQL within a Databricks offering using the ODBC and/or JDBC protocols.

Currently, this connector requires 30+MB of memory for each stream. When syncing multiple streams, it may run into an out-of-memory error if the allocated memory is too small. This performance bottleneck is tracked in this issue. Once this issue is resolved, the connector should be able to sync an almost infinite number of streams with less than 500MB of memory.

Sync Mode

Feature Support Notes
Full Refresh Sync ✅ Warning: this mode deletes all previously synced data in the configured bucket path.
Incremental - Append Sync ✅
Incremental - Deduped History ❌
Namespaces ✅

Data Source

Databricks Delta Lake supports various cloud storage as the data source. Currently, only Amazon S3 is supported by this connector.

Configuration

Category Parameter Type Notes
Databricks Server Hostname string Required. Example: abc-12345678-wxyz.cloud.databricks.com. See documentation. Please note that this is the server for the Databricks Cluster. It is different from the SQL Endpoint Cluster.
HTTP Path string Required. Example: sql/protocolvx/o/1234567489/0000-1111111-abcd90. See documentation.
Port string Optional. Default to "443". See documentation.
Personal Access Token string Required. Example: dapi0123456789abcdefghij0123456789AB. See documentation.
General Database schema string Optional. Default to "public". Each data stream will be written to a table under this database schema.
Purge Staging Data boolean The connector creates staging files and tables on S3. By default, they will be purged when the data sync is complete. Set it to false for debugging purposes.
Data Source - S3 Bucket Name string Name of the bucket to sync data into.
Bucket Path string Subdirectory under the above bucket to sync the data into.
Region string See documentation for all region codes.
Access Key ID string AWS/Minio credential.
Secret Access Key string AWS/Minio credential.
S3 Filename pattern string The pattern allows you to set the file-name format for the S3 staging file(s), next placeholders combinations are currently supported: {date}, {date:yyyy_MM}, {timestamp}, {timestamp:millis}, {timestamp:micros}, {part_number}, {sync_id}, {format_extension}. Please, don't use empty space and not supportable placeholders, as they won't recognized.

⚠️ Please note that under "Full Refresh Sync" mode, data in the configured bucket and path will be wiped out before each sync. We recommend you provision a dedicated S3 resource for this sync to prevent unexpected data deletion from misconfiguration. ⚠️

Staging Parquet Files (Delta Format)

Data streams are first written as staging delta-table (Parquet + Transaction Log) files on S3, and then loaded into Databricks delta-tables. All the staging files will be deleted after the sync is done. For debugging purposes, here is the full path for a staging file:

s3://<bucket-name>/<bucket-path>/<uuid>/<stream-name>

For example:

s3://testing_bucket/data_output_path/98c450be-5b1c-422d-b8b5-6ca9903727d9/users/_delta_log
     ↑              ↑                ↑                                    ↑     ↑
     |              |                |                                    |     transaction log
     |              |                |                                    stream name
     |              |                database schema
     |              bucket path
     bucket name

Unmanaged Spark SQL Table

Currently, all streams are synced into unmanaged Spark SQL tables. See documentation for details. In summary, you have full control of the location of the data underlying an unmanaged table. The full path of each data stream is:

s3://<bucket-name>/<bucket-path>/<database-schema>/<stream-name>

For example:

s3://testing_bucket/data_output_path/public/users
     ↑              ↑                ↑      ↑
     |              |                |      stream name
     |              |                database schema
     |              bucket path
     bucket name

Please keep these data directories on S3. Otherwise, the corresponding tables will have no data in Databricks.

Output Schema

Each table will have the following columns:

Column Type Notes
_airbyte_ab_id string UUID.
_airbyte_emitted_at timestamp Data emission timestamp.
Data fields from the source stream various All fields in the staging Parquet files will be expanded in the table.

Under the hood, an Airbyte data stream in Json schema is first converted to an Avro schema, then the Json object is converted to an Avro record, and finally the Avro record is outputted to the Parquet format. Because the data stream can come from any data source, the Json to Avro conversion process has arbitrary rules and limitations. Learn more about how source data is converted to Avro and the current limitations here.

Getting started

Requirements

  1. Credentials for a Databricks cluster. See documentation.
  2. Credentials for an S3 bucket. See documentation.
  3. Grant the Databricks cluster full access to the S3 bucket. Or mount it as Databricks File System (DBFS). See documentation.

CHANGELOG

Version Date Pull Request Subject
(unpublished) 2022-09-01 #16243 Fix Json to Avro conversion when there is field name clash from combined restrictions (anyOf, oneOf, allOf fields)
0.2.6 2022-08-05 #14801 Fix multiply log bindings
0.2.5 2022-07-15 #14494 Make S3 output filename configurable.
0.2.4 2022-07-14 #14618 Removed additionalProperties: false from JDBC destination connectors
0.2.3 2022-06-16 #13852 Updated stacktrace format for any trace message errors
0.2.2 2022-06-13 #13722 Rename to "Databricks Lakehouse".
0.2.1 2022-06-08 #13630 Rename to "Databricks Delta Lake" and add field orders in the spec.
0.2.0 2022-05-15 #12861 Use new public Databricks JDBC driver, and open source the connector.
0.1.5 2022-05-04 #12578 In JSON to Avro conversion, log JSON field values that do not follow Avro schema for debugging.
0.1.4 2022-02-14 #10256 Add -XX:+ExitOnOutOfMemoryError JVM option
0.1.3 2022-01-06 #7622 #9153 Upgrade Spark JDBC driver to 2.6.21 to patch Log4j vulnerability; update connector fields title/description.
0.1.2 2021-11-03 #7288 Support Json additionalProperties.
0.1.1 2021-10-05 #6792 Require users to accept Databricks JDBC Driver Terms & Conditions.
0.1.0 2021-09-14 #5998 Initial private release.