Debezium Connector for Oracle

Typically, the redo logs on an Oracle server are configured to not retain the complete history of the database. As a result, the Debezium Oracle connector cannot retrieve the entire history of the database from the logs. To enable the connector to establish a baseline for the current state of the database, the first time that the connector starts, it performs an initial consistent snapshot of the database.

By default, a connector runs an initial snapshot operation only after it starts for the first time. Following this initial snapshot, under normal circumstances, the connector does not repeat the snapshot process. Any future change event data that the connector captures comes in through the streaming process only.

However, in some situations the data that the connector obtained during the initial snapshot might become stale, lost, or incomplete. To provide a mechanism for recapturing table data, Debezium includes an option to perform ad hoc snapshots. You might want to perform an ad hoc snapshot after any of the following changes occur in your Debezium environment:

You can re-run a snapshot for a table for which you previously captured a snapshot by initiating a so-called ad-hoc snapshot. Ad hoc snapshots require the use of signaling tables. You initiate an ad hoc snapshot by sending a signal request to the Debezium signaling table.

When you initiate an ad hoc snapshot of an existing table, the connector appends content to the topic that already exists for the table. If a previously existing topic was removed, Debezium can create a topic automatically if automatic topic creation is enabled.

Ad hoc snapshot signals specify the tables to include in the snapshot. The snapshot can capture the entire contents of the database, or capture only a subset of the tables in the database. Also, the snapshot can capture a subset of the contents of the table(s) in the database.

You specify the tables to capture by sending an execute-snapshot message to the signaling table. Set the type of the execute-snapshot signal to incremental or blocking, and provide the names of the tables to include in the snapshot, as described in the following table:

You initiate an ad hoc incremental snapshot by adding an entry with the execute-snapshot signal type to the signaling table, or by sending a signal message to a Kafka signaling topic. After the connector processes the message, it begins the snapshot operation. The snapshot process reads the first and last primary key values and uses those values as the start and end point for each table. Based on the number of entries in the table, and the configured chunk size, Debezium divides the table into chunks, and proceeds to snapshot each chunk, in succession, one at a time.

For more information, see Incremental snapshots.

You initiate an ad hoc blocking snapshot by adding an entry with the execute-snapshot signal type to the signaling table or signaling topic. After the connector processes the message, it begins the snapshot operation. The connector temporarily stops streaming, and then initiates a snapshot of the specified table, following the same process that it uses during an initial snapshot. After the snapshot completes, the connector resumes streaming.

For more information, see Blocking snapshots.

To provide flexibility in managing snapshots, Debezium includes a supplementary snapshot mechanism, known as incremental snapshotting. Incremental snapshots rely on the Debezium mechanism for sending signals to a Debezium connector. Incremental snapshots are based on the DDD-3 design document.

In an incremental snapshot, instead of capturing the full state of a database all at once, as in an initial snapshot, Debezium captures each table in phases, in a series of configurable chunks. You can specify the tables that you want the snapshot to capture and the size of each chunk. The chunk size determines the number of rows that the snapshot collects during each fetch operation on the database. The default chunk size for incremental snapshots is 1024 rows.

As an incremental snapshot proceeds, Debezium uses watermarks to track its progress, maintaining a record of each table row that it captures. This phased approach to capturing data provides the following advantages over the standard initial snapshot process:

When you run an incremental snapshot, Debezium sorts each table by primary key and then splits the table into chunks based on the configured chunk size. Working chunk by chunk, it then captures each table row in a chunk. For each row that it captures, the snapshot emits a READ event. That event represents the value of the row when the snapshot for the chunk began.

As a snapshot proceeds, it’s likely that other processes continue to access the database, potentially modifying table records. To reflect such changes, INSERT, UPDATE, or DELETE operations are committed to the transaction log as per usual. Similarly, the ongoing Debezium streaming process continues to detect these change events and emits corresponding change event records to Kafka.

In some cases, the UPDATE or DELETE events that the streaming process emits are received out of sequence. That is, the streaming process might emit an event that modifies a table row before the snapshot captures the chunk that contains the READ event for that row. When the snapshot eventually emits the corresponding READ event for the row, its value is already superseded. To ensure that incremental snapshot events that arrive out of sequence are processed in the correct logical order, Debezium employs a buffering scheme for resolving collisions. Only after collisions between the snapshot events and the streamed events are resolved does Debezium emit an event record to Kafka.

To assist in resolving collisions between late-arriving READ events and streamed events that modify the same table row, Debezium employs a so-called snapshot window. The snapshot window demarcates the interval during which an incremental snapshot captures data for a specified table chunk. Before the snapshot window for a chunk opens, Debezium follows its usual behavior and emits events from the transaction log directly downstream to the target Kafka topic. But from the moment that the snapshot for a particular chunk opens, until it closes, Debezium performs a de-duplication step to resolve collisions between events that have the same primary key..

For each data collection, the Debezium emits two types of events, and stores the records for them both in a single destination Kafka topic. The snapshot records that it captures directly from a table are emitted as READ operations. Meanwhile, as users continue to update records in the data collection, and the transaction log is updated to reflect each commit, Debezium emits UPDATE or DELETE operations for each change.

As the snapshot window opens, and Debezium begins processing a snapshot chunk, it delivers snapshot records to a memory buffer. During the snapshot windows, the primary keys of the READ events in the buffer are compared to the primary keys of the incoming streamed events. If no match is found, the streamed event record is sent directly to Kafka. If Debezium detects a match, it discards the buffered READ event, and writes the streamed record to the destination topic, because the streamed event logically supersede the static snapshot event. After the snapshot window for the chunk closes, the buffer contains only READ events for which no related transaction log events exist. Debezium emits these remaining READ events to the table’s Kafka topic.

The connector repeats the process for each snapshot chunk.

Currently, you can use either of the following methods to initiate an incremental snapshot:

For more advanced uses, you can fine-tune control of the snapshot by implementing one of the following interfaces:

To provide more flexibility in managing snapshots, Debezium includes a supplementary ad hoc snapshot mechanism, known as a blocking snapshot. Blocking snapshots rely on the Debezium mechanism for sending signals to a Debezium connector.

A blocking snapshot behaves just like an initial snapshot, except that you can trigger it at run time.

You might want to run a blocking snapshot rather than use the standard initial snapshot process in the following situations:

When you run a blocking snapshot, Debezium stops streaming, and then initiates a snapshot of the specified table, following the same process that it uses during an initial snapshot. After the snapshot completes, the streaming is resumed.

You can set the following properties in the data component of a signal:

For example:

A delay might exist between the time that you send the signal to trigger the snapshot, and the time when streaming stops and the snapshot starts. As a result of this delay, after the snapshot completes, the connector might emit some event records that duplicate records captured by the snapshot.

By default, the Oracle connector writes change events for all INSERT, UPDATE, and DELETE operations that occur in a table to a single Apache Kafka topic that is specific to that table. The connector uses the following convention to name change event topics:

topicPrefix.schemaName.tableName

The following list provides definitions for the components of the default name:

For example, if fulfillment is the server name, inventory is the schema name, and the database contains tables with the names orders, customers, and products, the Debezium Oracle connector emits events to the following Kafka topics, one for each table in the database:

The connector applies similar naming conventions to label its internal database schema history topics, schema change topics, and transaction metadata topics.

If the default topic name do not meet your requirements, you can configure custom topic names. To configure custom topic names, you specify regular expressions in the logical topic routing SMT. For more information about using the logical topic routing SMT to customize topic naming, see Topic routing.

When a database client queries a database, the client uses the database’s current schema. However, the database schema can be changed at any time, which means that the connector must be able to identify what the schema was at the time each insert, update, or delete operation was recorded. Also, a connector cannot necessarily apply the current schema to every event. If an event is relatively old, it’s possible that it was recorded before the current schema was applied.

To ensure correct processing of events that occur after a schema change, Oracle includes in the redo log not only the row-level changes that affect the data, but also the DDL statements that are applied to the database. As the connector encounters these DDL statements in the redo log, it parses them and updates an in-memory representation of each table’s schema. The connector uses this schema representation to identify the structure of the tables at the time of each insert, update, or delete operation and to produce the appropriate change event. In a separate database schema history Kafka topic, the connector records all DDL statements along with the position in the redo log where each DDL statement appeared.

When the connector restarts after either a crash or a graceful stop, it starts reading the redo log from a specific position, that is, from a specific point in time. The connector rebuilds the table structures that existed at this point in time by reading the database schema history Kafka topic and parsing all DDL statements up to the point in the redo log where the connector is starting.

This database schema history topic is internal for internal connector use only. Optionally, the connector can also emit schema change events to a different topic that is intended for consumer applications.

You can configure a Debezium Oracle connector to produce schema change events that describe structural changes that are applied to tables in the database. The connector writes schema change events to a Kafka topic named <serverName>, where serverName is the namespace that is specified in the topic.prefix configuration property.

Debezium emits a new message to the schema change topic whenever it streams data from a new table, or when the structure of the table is altered.

Messages that the connector sends to the schema change topic contain a payload, and, optionally, also contain the schema of the change event message.

The schema for the schema change event has the following elements:

The following example shows a typical schema in JSON format.

The payload of a schema change event message includes the following elements:

The following example shows a typical schema change message in JSON format. The message contains a logical representation of the table schema.

In messages that the connector sends to the schema change topic, the message key is the name of the database that contains the schema change. In the following example, the payload field contains the databaseName key:

Debezium can generate events that represent transaction metadata boundaries and that enrich data change event messages.

Database transactions are represented by a statement block that is enclosed between the BEGIN and END keywords. Debezium generates transaction boundary events for the BEGIN and END delimiters in every transaction. Transaction boundary events contain the following fields:

The following example shows a typical transaction boundary message:

Unless overridden via the topic.transaction option, the connector emits transaction events to the <topic.prefix>.transaction topic.

Entries in the Oracle redo logs do not store the original SQL statements that users submit to make DML changes. Instead, a redo entry holds a set of change vectors and a set of object identifiers that represent the tablespace, table, and columns related to these vectors. In other words, redo log entries don’t include the names of the schemas, tables, or columns affected by DML changes.

The Debezium Oracle connector uses the log.mining.strategy configuration property to control how Oracle LogMiner handles the lookup of the object identifiers in the change vectors. In certain situations, one log mining strategy might prove more reliable than another with regard to schema changes. However, before you choose a log mining strategy, it’s important to consider the implications it might have on performance and overhead.

The Debezium Oracle connector integrates with Oracle LogMiner by default. This integration requires a specialized set of steps which includes generating a complex JDBC SQL query to ingest the changes recorded in the transaction logs as change events. The V$LOGMNR_CONTENTS view used by the JDBC SQL query does not have any indices to improve the query’s performance, and so there are different query modes that can be used that control how the SQL query is generated as a way to improve the query’s execution.

The log.mining.query.filter.mode connector property can be configured with one of the following to influence how the JDBC SQL query is generated:

Oracle writes all changes to the redo logs in the order in which they occur, including changes that are later discarded by a rollback. As a result, concurrent changes from separate transactions are intertwined. When the connector first reads the stream of changes, because it cannot immediately determine which changes are committed or rolled back, it temporarily stores the change events in an internal buffer. After a change is committed, the connector writes the change event from the buffer to Kafka. The connector drops change events that are discarded by a rollback.

You can configure the buffering mechanism that the connector uses by setting the property log.mining.buffer.type.

When the Debezium Oracle connector is configured to use LogMiner, it collects change events from Oracle by using a start and end range that is based on system change numbers (SCNs). The connector manages this range automatically, increasing or decreasing the range depending on whether the connector is able to stream changes in near real-time, or must process a backlog of changes due to the volume of large or bulk transactions in the database.

Under certain circumstances, the Oracle database advances the SCN by an unusually high amount, rather than increasing the SCN value at a constant rate. Such a jump in the SCN value can occur because of the way that a particular integration interacts with the database, or as a result of events such as hot backups.

The Debezium Oracle connector relies on the following configuration properties to detect the SCN gap and adjust the mining range.

The connector first compares the difference in the number of changes between the current SCN and the highest SCN in the current mining range. If the difference between the current SCN value and the highest SCN value is greater than the minimum gap size, then the connector has potentially detected a SCN gap. To confirm whether a gap exists, the connector next compares the timestamps of the current SCN and the SCN at the end of the previous mining range. If the difference between the timestamps is less than the maximum time interval, then the existence of an SCN gap is confirmed.

When an SCN gap occurs, the Debezium connector automatically uses the current SCN as the end point for the range of the current mining session. This allows the connector to quickly catch up to the real-time events without mining smaller ranges in between that return no changes because the SCN value was increased by an unexpectedly large number. When the connector performs the preceding steps in response to an SCN gap, it ignores the value that is specified by the log.mining.batch.size.max property. After the connector finishes the mining session and catches back up to real-time events, it resumes enforcement of the maximum log mining batch size.

The Debezium Oracle connector tracks system change numbers in the connector offsets so that when the connector is restarted, it can begin where it left off. These offsets are part of each emitted change event; however, when the frequency of database changes are low (every few hours or days), the offsets can become stale and prevent the connector from successfully restarting if the system change number is no longer available in the transaction logs.

For connectors that use non-CDB mode to connect to Oracle, you can enable heartbeat.interval.ms to force the connector to emit a heartbeat event at regular intervals so that offsets remain synchronized.

For connectors that use CDB mode to connect to Oracle, maintaining synchronization is more complicated. Not only must you set heartbeat.interval.ms, but it’s also necessary to set heartbeat.action.query. Specifying both properties is required, because in CDB mode, the connector specifically tracks changes inside the PDB only. A supplementary mechanism is needed to trigger change events from within the pluggable database. At regular intervals, the heartbeat action query causes the connector to insert a new table row, or update an existing row in the pluggable database. Debezium detects the table changes and emits change events for them, ensuring that offsets remain synchronized, even in pluggable databases that process changes infrequently.

For each changed table, the change event key is structured such that a field exists for each column in the primary key (or unique key constraint) of the table at the time when the event is created.

For example, a customers table that is defined in the inventory database schema, might have the following change event key:

If the value of the <topic.prefix>.transaction configuration property is set to server1, the JSON representation for every change event that occurs in the customers table in the database features the following key structure:

The schema portion of the key contains a Kafka Connect schema that describes the content of the key portion. In the preceding example, the payload value is not optional, the structure is defined by a schema named server1.DEBEZIUM.CUSTOMERS.Key, and there is one required field named id of type int32. The value of the key’s payload field indicates that it is indeed a structure (which in JSON is just an object) with a single id field, whose value is 1004.

Therefore, you can interpret this key as describing the row in the inventory.customers table (output from the connector named server1) whose id primary key column had a value of 1004.

The structure of a value in a change event message mirrors the structure of the message key in the change event in the message, and contains both a schema section and a payload section.

An envelope structure in the payload sections of a change event value contains the following fields:

The schema portion of the event message’s value contains a schema that describes the envelope structure of the payload and the nested fields within it.

The following example shows the value of a create event value from the customers table that is described in the change event keys example:

In the preceding example, notice how the event defines the following schema:

The payload portion of this event’s value, provides information about the event. It describes that a row was created (op=c), and shows that the after field value contains the values that were inserted into the ID, FIRST_NAME, LAST_NAME, and EMAIL columns of the row.

The following example shows an update change event that the connector captures from the same table as the preceding create event.

The payload has the same structure as the payload of a create (insert) event, but the following values are different:

The payload section reveals several other useful pieces of information. For example, by comparing the before and after structures, we can determine how a row changed as the result of a commit. The source structure provides information about Oracle’s record of this change, providing traceability. It also gives us insight into when this event occurred in relation to other events in this topic and in other topics. Did it occur before, after, or as part of the same commit as another event?

The following example shows a delete event for the table that is shown in the preceding create and update event examples. The schema portion of the delete event is identical to the schema portion for those events.

The payload portion of the event reveals several differences when compared to the payload of a create or update event:

The delete event provides consumers with the information that they require to process the removal of this row.

The Oracle connector’s events are designed to work with Kafka log compaction, which allows for the removal of some older messages as long as at least the most recent message for every key is kept. This allows Kafka to reclaim storage space while ensuring the topic contains a complete dataset and can be used for reloading key-based state.

When a row is deleted, the delete event value shown in the preceding example still works with log compaction, because Kafka is able to remove all earlier messages that use the same key. The message value must be set to null to instruct Kafka to remove all messages that share the same key. To make this possible, by default, Debezium’s Oracle connector always follows a delete event with a special tombstone event that has the same key but null value. You can change the default behavior by setting the connector property