Lattice is an open-source framework designed for fast and consistent data propagation in your CDC pipelines. Built to handle real-time data, you can have infrastructure running in a matter of minutes.

For more information, check out our case study: Lattice Platform Case Study

Lattice runs on a cluster of ECS instances in AWS and is composed of the following components:

• Apache Kafka — High-throughput distributed event streaming.
• Kafka Connect with Debezium — Change data capture from your source databases.
• Apicurio Schema Registry — Manages and stores schemas for Kafka messages, ensuring efficient data serialization and schema evolution.
• Lattice UI — Web interface for managing and monitoring your pipelines.
• OpenTofu Automation with CLI Wrapper — Infrastructure-as-code deployment tooling.

Local instance: Ensure that replication and the write-ahead log are accessible. In your postgresql.conf file, add or update the following settings:

wal_level = logical
max_replication_slots = 5
max_wal_senders = 5

Then, in your pg_hba.conf file, allow replication connections:

host replication all 0.0.0.0/0 md5

RDS instance: Enable logical replication via the RDS console by setting the following parameter:

name         = "rds.logical_replication"
value        = "1"
apply_method = "pending-reboot"

A reboot of your instance is required after making these changes.

Permissions: Inside psql, grant the appropriate permissions to your database user. The example below grants access to a user named inventory on the inventory database:

GRANT CONNECT ON DATABASE inventory TO inventory;
GRANT USAGE ON SCHEMA public TO inventory;
GRANT rds_replication TO inventory;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO inventory;
ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT SELECT ON TABLES TO inventory;

Run MongoDB with the --replSet flag and call rs.initiate() to initialize the replica set.

Ensure your AWS credentials have the permissions necessary to create ECS instances. You can use an admin account or a role with at minimum the following permissions:

iam:PassRole
iam:PutUserPolicy
iam:AttachUserPolicy
iam:PutRolePolicy

Contact your AWS administrator if you need these permissions granted.

Once permissions are in place:

1. Install and configure the AWS CLI.
2. Verify it's working by running:

   aws sts get-caller-identity
   

Follow the official OpenTofu installation guide. After installation, verify it's working:

tofu --help

Once the AWS CLI is configured and OpenTofu is installed, run npx lattice-platform-cli setup to provision your infrastructure. To teardown its as simple as npx lattice-platform-cli teardown

After deployment, locate the UI through the AWS console:

1. Navigate to your ECS cluster.
2. Find the service named lattice-ui.
3. Under Tasks, click the specific Task ID for lattice-ui.
4. Go to the Network tab to find the service's IP address.
5. Open that IP address in your browser to access the Lattice UI.

Alternatively, if you have access to your private VPC, navigate to lattice.local in your browser.

Full UI documentation is available in the how-to section of our website.

Sink connectors write data to its destination after it travels through the Kafka pipeline. Clickhouse is our primary tested and fully supported sink, though any JDBC-compatible connector can be used.

Tables in ClickHouse must follow a specific naming convention to map correctly from Kafka topics. For example, to sync the products table from PostgreSQL, create a table named fulfillment.public.products in ClickHouse. This follows the naming convention of the topic created in kafka. The Kafka topic will be mapped to the matching table name with what you setup in the GUI in step five. ensure that the rows in the schema are correctly mapped to the ClickHouse table columns.

Other JDBC-compatible databases are supported. Refer to the JDBC documentation for your specific database to confirm setup requirements and replication support.

Instructions for local/manual deployment can be found in the lattice-local repository.

Joshua Crane | Kessler Mentele | William Seipp