Ingest Your MongoDB CDC Data to Snowflake

Learn how to ingest your MongoDB CDC data to Snowflake

In this guide, you will learn how to ingest data from your MongoDB database into a staging table in the data lake. Then you will understand how to monitor the snapshotting process, before creating a transformation job to load the data into Snowflake.


Before you ingest data into Upsolver, you must enable change data capture on your MongoDB database. If you are using a managed MongoDB service such as Atlas, CDC is most likely to be enabled, if not, please refer to the guide to Deploy a Replica Set for more information.

The steps for ingesting your CDC data are as follows:

  1. Connect to MongoDB

  2. Create a staging table to store the CDC data

  3. Create an ingestion job

  4. View the job status to check the snapshotting process

  5. View the CDC data in the staging table

  6. Connect to Snowflake

  7. Create a transformation job

Step 1

Connect to MongoDB

First, we will create a connection to the MongoDB database from which you want to ingest your CDC data. You will need the connection string to your database, and the username and password. Ensure your login has appropriate credentials for reading from the change data capture collections.

Here's the code:

CREATE MONGODB CONNECTION my_mongodb_connection
  CONNECTION_STRING = 'mongodb+srv://'
  USER_NAME = '<user_name>'
  PASSWORD = '<password>'
  COMMENT = 'Connection to MongoDB sales database';

Step 2

Create a staging table to store the CDC data

Now that you have connected to your source database, the next step is to create a table in the data lake to stage the CDC data.

Here's the code to create the staging table:

CREATE TABLE default_glue_catalog.upsolver_samples.sales_raw_data()
    PARTITIONED BY $event_date;

Let's understand what this code does.

Firstly, a table named sales_raw_data is created in the upsolver_samples database. The open brackets with no defined columns instruct Upsolver to infer the columns and types during data ingestion. This is helpful if you are unsure of the data in the source and want Upsolver to manage type changes and schema updates. In this example, we will ingest data from multiple collections into one staging table and don't need to worry about schemas, as Upsolver handles this for us.

Upsolver recommends partitioning by the system column $event_date or another date column in order to optimize your query performance. The $event_date column is added by default as a system column, along with $event_time, which will be used later when you create your transformation job. You can view all the system columns that Upsolver adds to the tables in your default glue catalog, by expanding the table name in the Entities tree in Upsolver, and then expanding SYSTEM COLUMNS.

Learn More

Step 3

Create an ingestion job

After defining the staging table, the next step is to create an ingestion job to copy in the CDC data.

Here's the code to create the ingestion job:

CREATE SYNC JOB load_raw_data_from_mongodb
    COMMENT = 'Ingest sales data from MongoDB'
AS COPY FROM MONGODB my_mongodb_connection
    COLLECTION_INCLUDE_LIST = ('sales.orders', 'sales.products', 'sales.customers')
INTO default_glue_catalog.upsolver_samples.sales_raw_data; 

Let's take a look at what this code does.

A job named load_raw_data_from_mongodb is created with an optional comment that you can use to describe the purpose of your job. Other users in your organization can see comments, so it is useful to provide a comment to describe the purpose of the job.

Ingestion jobs use the COPY FROM command to copy source data to the target, in this case, the sales_raw_data table in the AWS Glue Data Catalog, using the my_mongodb_connection connection.

In our example, the COLLECTION_INCLUDE_LIST source option instructs the job to ingest from the orders, products, and customers collections, and ignore everything else.

Step 4

View the job status

When you create your CDC job, Upsolver takes a snapshot of each of the included collections prior to the streaming process. Optionally you can use the SKIP_SNAPSHOTS job option to ignore this step, but usually, you want to ingest all historical data.

You can check the status of the snapshotting process by clicking on Jobs from the main menu on the left-hand side of the Upsolver UI. Then, click on the job you created, e.g. load_raw_data_from_mongodb, and the job page displays each collection and status, e.g. Pending, Snapshotting, or Streaming.

After the snapshot process has been completed and all collections are streaming, you can continue to use the job page to monitor and troubleshoot your job.

Learn More

Step 5

View the CDC data in the staging table

During the snapshotting process, Upsolver reads the column names and types from the CDC collections in the source database and creates a corresponding column in the staging table. Appended to your CDC columns are system information columns, including the source database and collection names. Additional columns include the binlog timestamp when the change was committed on the source database and an $is_delete column.

Prior to creating a transformation job to load data into the target, it is good practice to check the data in the staging table. Depending on the volume of data you are ingesting, it can take some time for the data to appear.

Here's the code:

FROM default_glue_catalog.upsolver_samples.sales_raw_data
LIMIT 50; -- return 50 rows to view a small sample of the ingested data 

Confirm your data is as expected, before moving on to the next steps of creating a transformation job to load the data into your target.

Step 6

Connect to Snowflake

Next, we want to create a connection to the target database, in this case, Snowflake. Create a persistent connection that is shared with other users in your organization as follows.

Here's the code:

CREATE SNOWFLAKE CONNECTION my_snowflake_connection
    USER_NAME = '<user_name>'
    PASSWORD = '<password>';

Step 7

Create a transformation job

Now that you have a connection to Snowflake, you can load your data using a transformation job. If you haven't already done so, create a target table in Snowflake.

Here's the code to create a table in Snowflake:

	ORDER_ID VARCHAR(16777216),

You can adapt this script to create additional tables for products and customers, or extend the columns in the table definition to load all data and transform it in Snowflake.

Next, create a transformation job to replicate your CDC data to the target table.

Here's the code:

CREATE SYNC JOB insert_web_orders_into_snowflake
        custid AS CUSTOMER_ID,
        orderid AS ORDER_ID,  
        first_name || ' ' || last_name AS CUSTOMER_NAME,
        nettotal AS NET_TOTAL,
        order_date AS ORDER_DATE           
    FROM default_glue_catalog.upsolver_samples.sales_raw_data 
    WHERE $event_time BETWEEN run_start_time() AND run_end_time();

Let's understand what this job does.

This code creates a job named insert_web_orders_into_snowflake and includes a couple of job options: START_FROM instructs the job to replicate all historical data by specifying the BEGINNING parameter, while RUN_INTERVAL tells Upsolver that this job should execute every 5 MINUTES.

The job inserts the data into the WEB_ORDERS table in the SALES schema in Snowflake. We don't need to specify the database (DEMO_DB) here because this is included in the connection string.

The MAP_COLUMNS_BY_NAME option maps each column in the SELECT statement to the column with the same name in the target table. This is helpful as the job, therefore, does not map the columns based on ordinality: if you compare the order of the columns in the script that creates the table with the order of the columns in the SELECT statement of the job, you'll notice that CUSTOMER_ID and ORDER_ID are in different positions.

The SELECT statement specifies which columns will be loaded into the target, and the alias names enable column mapping by name. A string function has been used to concatenate the customer's first and last names into the CUSTOMER_NAME column.

In the WHERE clause, all rows that have an $event_time that is between the start and end time interval of the job will be included in the load. The $event_time system column is populated with a timestamp when the data lands in the staging table.


In this guide you learned how to connect to your MongoDB database and ingest your change data capture collection into a staging table in the data lake. Then, you checked the status of your collections during the snapshotting process and saw how to view the ingested data in the staging table. Furthermore, you discovered how to write a transformation job to copy the data from your staging table to your target table in Snowflake.

Try it yourself

To ingest your CDC data from MongoDB:

  1. View the CDC data in the staging table

Learn More

Last updated