SQL Server Change Stream sample using Debezium. A change feed or change stream allow applications to access real-time data changes, using standard technologies and well-known API, to create modern applications using the full power of database like SQL Server.

Debezium make use of Change Data Capture, so it can be used with On-Premises SQL Servers, SQL Servers running on VMs in any cloud, and Azure SQL MI.

With Debezium and SQL Server you can not only create more modern and reactive applications that handle data changes in near real time with a minium impact on the database, but you can also use it to implement your Hybrid IT strategy, still using On-Prem SQL Server but relying on Azure for all your computing needs, taking advantage of PaaS offerings like EventHubs and Azure Functions. This sample will show how to do that.

For those who want more details, Debezium uses Apache Kafka, which is not natively available on Azure as a PaaS offer. But luckily EventHubs offer almost 100% support, so we can use it instead of Kafka and make maintenance and scalability easier.

Event Hubs for Kafka

This step by step guide uses Wide World Importers sample database from here:

https://github.com/Microsoft/sql-server-samples/releases/tag/wide-world-importers-v1.0

Make sure you download the OLTP database if you want to follow this guide without having to change a thing.

Debezium needs to query the database so a dedicated user is used throughout the sample. User has db_owner access to make the script simpler. In real world you may want to tighten security a bit more.

To create the login and user run the script /sql/00-setup-database-user.sql on the server where you have restored Wide World Importers database.

Debezium uses Change Data Capture to capture all the changes done to selected tables.

In this samples only two tables are monitored:

• Sales.Orders
• Warehouse.StockItems

The script /sql/01-enable-cdc.sql enable Change Data Capture on the aforementioned tables.

All data gathered by Change Data Capture will be send to Event Hubs, so create an Azure Event Hubs in your Azure Subscription. Using the Azure Cloud Shell Bash:

# create group
az group create -n debezium -l eastus

# create eventhuvbs with kafka enabled
az eventhubs namespace create -n debezium -g debezium -l eastus --enable-kafka

Later in the configuration process you'll need the EventHubs connection string, so grab it and store it somewhere:

az eventhubs namespace authorization-rule keys list -g debezium --namespace-name debezium -n RootManageSharedAccessKey --query "primaryConnectionString" -o tsv

In order to run Debezium you have to install and configure Apache Kafka, Apache Zookeper and Kafka Connect. If you already know how to do that, or your already have a testing or development environment, well, perfect. Go and install Debezium SQL Server Connector: Installing a Debezium Connector.

If prefer a more lean and quick easy to start using Debezium, you can just use the Debezium Docker Image, that provides anything you need to run a test instance of Debezium. Just make sure you have Docker and Docker Compose installed.

Docker Compose will use .env to get the environment variables values used in the .yaml configuration file. The provided .env.template file look like the following:

DEBEZIUM_VERSION=0.10
EH_NAME=debezium
EH_CONNECTION_STRING=

Copy it and create a new .env file. Leave the version set to 0.10. Change the EH_NAME to the EventHubs name you created before. Also set EH_CONNECTION_STRING to hold the EventHubs connection string you got before. Make sure not to use any additional quotes or double quotes.

If you are just interested in testing Debezium you can safely skip this section and move to the next one to start Debezium. If you want to understand how to make Debezium work with Eventhubs, read on.

Debezium needs Apache Kafka to run, NOT EventHubs. Luckily for us, EventHubs exposes a Kafka-Compatible endpoint, so we can still enjoy Kafka with all the comfort of a PaaS offering. There are a few tweeks needed in order to make Debezium working with EventHubs.

First of all EventHubs requires authentication. This part is taken care from the configuration settings that looks like the following:

- *_SECURITY_PROTOCOL=SASL_SSL
- *_SASL_MECHANISM=PLAIN
- *_SASL_JAAS_CONFIG=[...]

Documentation on EventHubs Kafka Authentication and Kafka Connect is available here:

Integrate Apache Kafka Connect support on Azure Event Hubs

Since we're running a Docker Image, we cannot really change the configuration file, but Debezium allows pass-through configurations:

Debezium Connect-Base

There is additional caveat to keep in mind. EventHubs security uses the string $ConnectionString as username. In order to avoid to have Docker Compose to treat it as a variable instead, a double dollar sign $$ needs to be used:

Docker Compose Config File Variable Substitution

Two other options useful for running Debezium on EventHubs are the following:

- CONNECT_KEY_CONVERTER_SCHEMAS_ENABLE=false
- CONNECT_VALUE_CONVERTER_SCHEMAS_ENABLE=true

They control if the schema is sent with the data or not. Since the EventHub only support values, as opposed to Apache Kafka, which everything is actually a key-value pair, the schema generation for the key section can be safely turned off. While this can also be done for the value part, it is not recommended as some data type are serialized in a Kafka-specific way and you need to know their "sematic" type in order to recreate the correct value.

Debezium SQL Server Connector Data Types

Here's a sample of a schema for a "create" (INSERT) event:

Debezium SQL Server Connector Create Event Sample

Debezium can now be started. If you're using the Docker Images you can just do this by running debezium/on-prem/start-debezium.ps1 (or the .sh file if you're on Linux/WSL)

Once the startup has finished, you'll see something like

[Worker clientId=connect-1, groupId=1] Finished starting connectors and tasks   [org.apache.kafka.connect.runtime.distributed.DistributedHerder]

you will see three topics (or EventHub to use the Azure EventHubs nomenclature):

az eventhubs eventhub list -g debezium --namespace debezium -o table

and the result will show:

• debezium_configs
• debezium_offsets
• debezium_statuses

to explore EventHubs is strongly suggest to download and use Service Bus Explorer

Now that Debezium is running, the SQL Server Connector can be registered. Before doing that, make sure to specify the correct connection for your SQL Server instance in the debezium/register-sqlserver-eh.json file.

If you are using the Wide World Importers database, the only values you have to change are:

"database.hostname" : "192.168.0.80",
"database.port" : "1433",

If you are following the step-by-step guide using a database of yours, make sure to also correctly set values for

"database.user" : "debezium-wwi",
"database.password" : "debezium-WWI-P@ssw0rd!",
"database.dbname" : "WideWorldImporters",

All the other values used are explained in detail here:

SQL Server Connector Configuration Values

Once the configuration file is set, just register that using debezium/on-prem/register-connector.ps1.

Depending on how big your tables are, it make take a while (more on this later). Once you see the following message:

Snapshot step 8 - Finalizing   [io.debezium.relational.HistorizedRelationalSnapshotChangeEventSource]

and no other errors or exception before that, you'll know that the SQL Server Connector is correctly running.

Now that Debezium is running and fully configured, you can generate a new Sales Order and insert, update and delete some data in the Stock table. You can use the following scripts:

./sql/02-create-new-sales-order.sql
./sql/03-modify-warehouse-stock.sql

After running the script you can use Service Bus Explorer or VS Code Event Hub Explorer to consume the stream of changes sent to EventHubs. You'll notice a new topic named wwi. That's where we instructed Debezium to send all the changes detected to the monitored tables.

One way to quickly react to the Change Stream data coming from Debezium is to use Azure Functions. A sample is available in folder azure-function. The easiest way to run the sample is to open it from VS Code. It will automatically recognize it as an Azure Function and download everything needed to run it.

Make sure you have a local.setting.json that looks like the provided template. Copy the EventHubs connection string you got at the beginning into the Debezium configuration option.

Start the function. As soon as the Azure Function runtime is running, the code will start to process the changes already available in EventHubs and you'll see something like this:

Event from Change Feed received:
- Object: Sales.Orders
- Operation: Insert
- Captured At: 2019-08-04T22:35:59.0100000Z
> OrderID = 73625
> CustomerID = 941
> SalespersonPersonID = 3
> PickedByPersonID =
> ContactPersonID = 3141
> BackorderOrderID =
> OrderDate = 8/4/2019 12:00:00 AM
> ExpectedDeliveryDate = 8/5/2019 12:00:00 AM
> CustomerPurchaseOrderNumber = 4923
> IsUndersupplyBackordered = False
> Comments = Auto-generated
> DeliveryInstructions = Unit 17, 1466 Deilami Road
> InternalComments =
> PickingCompletedWhen =
> LastEditedBy = 3
> LastEditedWhen = 8/4/2019 10:35:58 PM
Executed 'ProcessDebeziumPayload' (Succeeded, Id=ee9d1080-64ff-4039-83af-69c4b12fa85f)

Congratulations, you now have a working Change Stream from SQL Server. This opens up a whole new set of possibilities! Have fun!

As described in the documentation, Debezium take an initial snapshot of the selected tables, in order to send all the existing data into the change stream. As you can imagine this can take a long time if tables are big. Also keep in mind that, by default, Debezium will do a

SELECT * FROM <table> WITH (TABLOCKX)

and will do the same connection for all the configured tables so it will lock pretty much everything. Be aware! The lock used during the initial snapshot can be configured using dedicated options (see the "Advanced" section):

Connector Properties

If you don't want Debezium to take the snapshot, for example because you're doing it on your own, using Database Snapshots and Bulk Load, then you can set the option snapshot.mode to initial_schema_only, to make sure only schema is snapshotted and not data. In that case, starting from version 0.10.beta4 of Debezium, the following statement will be executed:

SELECT TOP (0) * FROM <table> WITH (TABLOCKX)

And that won't cause any harm to your system.

If you're using Debezium with Azure SQL MI, you may want to run Debezium on Azure. Sample script to run the Debezium container on Azure Container Instances are available in the debezium/azure folder.

More details on SQL Server and Event Hubs specific configuration here:

SQL Server Connector Configuration Values