Sending signals to a Debezium connector

Sending signals to a Debezium connector

Overview

The Debezium signaling mechanism provides a way to modify the behavior of a connector, or to trigger a one-time action, such as initiating an ad hoc snapshot of a table. To trigger a connector to perform a specified action, you issue a SQL command to add a signal message to a specialized signaling table, also referred to as a signaling data collection. The signaling table, which you create on the source database, is designated exclusively for communicating with Debezium. When Debezium detects that a new logging record or ad hoc snapshot record is added to the signaling table, it reads the signal, and initiates the requested operation.

Signaling is available for use with the following Debezium connectors:

Db2
MongoDB (Technology Preview)
MySQL
Oracle
PostgreSQL
SQL Server

Enabling signaling

By default, the Debezium signaling mechanism is disabled. You must explicitly enable signaling for each connector that you want to use it with.

Procedure

On the source database, create a signaling data collection table for sending signals to the connector. For information about the required structure of the signaling data collection, see Structure of a signaling data collection.
For source databases such as Db2 or SQL Server that implement a native change data capture (CDC) mechanism, enable CDC for the signaling table.
Add the name of the signaling data collection to the Debezium connector configuration.
In the connector configuration, add the property signal.data.collection, and set its value to the fully-qualified name of the signaling data collection that you created in Step 1.

For example, signal.data.collection = inventory.debezium_signals.

The format for the fully-qualified name of the signaling collection depends on the connector.
The following example shows the naming formats to use for each connector:

Db2

*<schemaName>*.*<tableName>*

MongoDB

*<databaseName>*.*<collectionName>*

MySQL

*<databaseName>*.*<tableName>*

Oracle

*<databaseName>*.*<schemaName>*.*<tableName>*

PostgreSQL

*<schemaName>*.*<tableName>*

SQL Server

*<databaseName>*.*<schemaName>*.*<tableName>*

For more information about setting the signal.data.collection property, see the table of configuration properties for your connector.

Structure of a signaling data collection

A signaling data collection, or signaling table, stores signals that you send to a connector to trigger a specified operation. The structure of the signaling table must conform to the following standard format.

Contains three fields (columns).
Fields are arranged in a specific order, as shown in Table 1.

Table 1. Required structure of a signaling data collection
Field	Type	Description
`id` (required)	`string`	An arbitrary unique string that identifies a signal instance. You assign an `id` to each signal that you submit to the signaling table. Typically the ID is a UUID string. You can use signal instances for logging, debugging, or de-duplication. When a signal triggers Debezium to perform an incremental snapshot, it generates a signal message with an arbitrary `id` string. The `id` string that the generated message contains is unrelated to the `id` string in the submitted signal.
`type` (required)	`string`	Specifies the type of signal to send. You can use some signal types with any connector for which signaling is available, while other signal types are available for specific connectors only.
`data` (optional)	`string`	Specifies JSON-formatted parameters to pass to a signal action. Each signal type requires a specific set of data.

The field names in a data collection are arbitrary. The preceding table provides suggested names. If you use a different naming convention, ensure that the values in each field are consistent with the expected content.

Creating a signaling data collection

You create a signaling table by submitting a standard SQL DDL query to the source database.

Prerequisites

You have sufficient access privileges to create a table on the target database.

Procedure

Submit a SQL query to the source database to create a table that is consistent with the required structure, as shown in the following example:

CREATE TABLE *<tableName>* (id VARCHAR(*<varcharValue>*) PRIMARY KEY, type VARCHAR(*<varcharValue>*) NOT NULL, data VARCHAR(*<varcharValue>*) NULL);

The amount of space that you allocate to the VARCHAR parameter of the id variable must be sufficient to accommodate the size of the ID strings of signals sent to the signaling table.
If the size of an ID exceeds the available space, the connector cannot process the signal.

The following example shows a CREATE TABLE command that creates a three-column debezium_signal table:

CREATE TABLE debezium_signal (id VARCHAR(42) PRIMARY KEY, type VARCHAR(32) NOT NULL, data VARCHAR(2048) NULL);

Signal actions

You can use signaling to initiate the following actions:

Add messages to the log.
Trigger ad hoc snapshots.
Stop execution of an ad hoc snapshot.
Pause incremental snapshots.
Resume incremenal snapshots.

Some signals are not compatible with all connectors.

Logging signals

You can request a connector to add an entry to the log by creating a signaling table entry with the log signal type. After processing the signal, the connector prints the specified message to the log. Optionally, you can configure the signal so that the resulting message includes the streaming coordinates.

Table 2. Example of a signaling record for adding a log message
Column	Value	Description
id	`924e3ff8-2245-43ca-ba77-2af9af02fa07`
type	`log`	The action type of the signal.
data	`{“message”: “Signal message at offset {}”}`	The `message` parameter specifies the string to print to the log. If you add a placeholder (`{}`) to the message, it is replaced with streaming coordinates.

Ad hoc snapshot signals

You can request a connector to initiate an ad hoc snapshot by creating a signaling table entry with the execute-snapshot signal type. After processing the signal, the connector runs the requested snapshot operation.

Unlike the initial snapshot that a connector runs after it first starts, an ad hoc snapshot occurs during runtime, after the connector has already begun to stream change events from a database. You can initiate ad hoc snapshots at any time.

Ad hoc snapshots are available for the following Debezium connectors:

Db2
MongoDB
MySQL
Oracle
PostgreSQL
SQL Server

Table 3. Example of an ad hoc snapshot signal record
Column	Value
id	`d139b9b7-7777-4547-917d-e1775ea61d41`
type	`execute-snapshot`
data	`{“data-collections”: [“public.MyFirstTable”, “public.MySecondTable”]}`

Currently, the execute-snapshot action triggers incremental snapshots only.

For more information about ad hoc snapshots, see the Snapshots topic in the documentation for your connector.

Additional resources

Ad hoc snapshot stop signals

You can request a connector to stop an in-progress ad hoc snapshot by creating a signal table entry with the stop-snapshot signal type. After processing the signal, the connector will stop the current in-progress snapshot operation.

You can stop ad hoc snapshots for the following Debezium connectors:

Db2
MongoDB
MySQL
Oracle
PostgreSQL
SQL Server

Table 4. Example of a stop ad hoc snapshot signal record
Column	Value
id	`d139b9b7-7777-4547-917d-e1775ea61d41`
type	`stop-snapshot`
data	`{“type”:”INCREMENTAL”, “data-collections”: [“public.MyFirstTable”]}`

You must specify the type of the signal. The data-collections field is optional. Leave the data-collections field blank to request the connector to stop all activity in the current snapshot. If you want the incremental snapshot to proceed, but you want to exclude specific collections from the snapshot, provide a comma-separated list of the names of the collections or regular expressions to exclude. After the connector processes the signal, the incremental snapshot proceeds, but it excludes data from the collections that you specify.

Incremental snapshots

Incremental snapshots are a specific type of ad hoc snapshot. In an incremental snapshot, the connector captures the baseline state of the tables that you specify, similar to an initial snapshot. However, unlike an initial snapshot, an incremental snapshot captures tables in chunks, rather than all at once. The connector uses a watermarking method to track the progress of the snapshot.

By capturing the initial state of the specified tables in chunks rather than in a single monolithic operation, incremental snapshots provide the following advantages over the initial snapshot process:

While the connector captures the baseline state of the specified tables, streaming of near real-time events from the transaction log continues uninterrupted.
If the incremental snapshot process is interrupted, it can be resumed from the point at which it stopped.
You can initiate an incremental snapshot at any time.

Incremental snapshot pause signals

You can request a connector to pause an in-progress incremental snapshot by creating a signal table entry with the pause-snapshot signal type. After processing the signal, the connector will stop pause current in-progress snapshot operation. Therefor it’s not possible to specify the data collection as the snapshot processing will be paused in position where it is in time of processing of the signal.

You can pause incremental snapshots for the following Debezium connectors:

Db2
MongoDB
MySQL
Oracle
PostgreSQL
SQL Server

Table 5. Example of a pause incremental snapshot signal record
Column	Value
id	`d139b9b7-7777-4547-917d-e1775ea61d41`
type	`pause-snapshot`

You must specify the type of the signal. The data field is ignored.

Incremental snapshot resume signals

You can request a connector to resume a paused incremental snapshot by creating a signal table entry with the resume-snapshot signal type. After processing the signal, the connector will resume previously paused snapshot operation.

You can resume incremental snapshots for the following Debezium connectors:

Db2
MongoDB
MySQL
Oracle
PostgreSQL
SQL Server

Table 6. Example of a resume incremental snapshot signal record
Column	Value
id	`d139b9b7-7777-4547-917d-e1775ea61d41`
type	`resume-snapshot`