Change Data Capture | Ignite Documentation

Ignite Summit Europe — Recordings Available — Watch Now


Change Data Capture


Change Data Capture (CDC) is a data processing pattern used to asynchronously receive entries that have been changed on the local node so that action can be taken using the changed entry.

CDC is an experimental feature. API or design architecture might be changed.

Below are some CDC use cases:

  • Streaming changes in Warehouse;

  • Updating search index;

  • Calculating statistics (streaming queries);

  • Auditing logs;

  • Async interaction with extenal system: Moderation, business process invocation, etc.

Ignite implements CDC with the application and Java API.

Below are the CDC application and the Ignite node integrated via WAL archive segments:

CDC design

When CDC is enabled, the Ignite server node creates a hard link to each WAL archive segment in the special db/cdc/{consistency_id} directory. The application runs on a different JVM and processes newly archived WAL segments. When the segment is fully processed by, it is removed. The actual disk space is free when both links (archive and CDC) are removed.

State of consumption is a pointer to the last processed event. Consumer can tell to to save the consumption state. On startup event processing will be continued from the last saved state.


Ignite Node

Name Description Default value


Flag to enable CDC on the server node.



Path to the CDC directory



Timeout to forcefully archive the WAL segment even it is not complete.

-1 (disabled)

CDC Application

CDC is configured in the same way as the Ignite node - via the spring XML file:

  • requires both Ignite and CDC configurations to start;

  • IgniteConfiguration is used to determine common options like a path to the CDC directory, node consistent id, and other parameters;

  • CdcConfiguration contains options.

Name Description Default value


Timeout to wait for lock acquiring. CDC locks directory on a startup to ensure there is no concurrent processing the same directory.

1000 milliseconds.


Amount of time application sleeps between subsequent checks when no new files available.

1000 milliseconds.


Flag to specify if key and value of changed entries should be provided in binary format.



Implementation of org.apache.ignite.cdc.CdcConsumer that consumes entries changes.



Array of SPI’s to export CDC metrics. See metrics documentation, also.




Below is a single change of the data reflected by CdcEvent.

Name Description


Key for the changed entry.


Value for the changed entry. This method will return null if the event reflects removal.


ID of the cache where the change happens. The value is equal to the CACHE_ID from SYS.CACHES.


Partition of the changed entry.


Flag to distinguish if operation happens on the primary or a backup node.


Comparable version of the changed entry. Internally, Ignite maintains ordered versions of each entry so any changes of the same entry can be sorted.


The consumer of change events. It should be implemented by the user.

Name Description

void start(MetricRegistry)

Invoked one-time at the start of the CDC application. MetricRegistry should be used to export the consumer-specific metrics.

boolean onEvents(Iterator<CdcEvent> events)

The main method that processes changes. When this method returns true, the state is saved on the disk. State points to the event next to the last read event. In case of any failure, consumption will continue from the last saved state.

void stop()

Invokes one-time at the stop of the CDC application.

Metrics uses the same SPI to export metrics as Ignite does. The following metrics are provided by the application (additional metrics can be provided by the consumer):




Index of the currently processing WAL segment.


Index of the WAL segment that contains the last committed state.


Committed offset in bytes inside the WAL segment.


Timestamp (in milliseconds) indicating the last segment processing start.


Binary meta-directory the application reads data from.


Marshaller directory the application reads data from.


The CDC directory the application reads data from.

Logging uses the same logging configuration as the Ignite node does. The only difference is that the log is written in the"ignite-cdc.log" file.


Important implements the fail-fast approach. It just fails in case of any error. The restart procedure should be configured with the OS tools.
  1. Find the required shared directories. Take the values from the provided IgniteConfiguration.

  2. Lock the CDC directory.

  3. Load the saved state.

  4. Start the consumer.

  5. Infinitely wait for the newly available segment and process it.

  6. Stop the consumer in case of a failure or a received stop signal.


Ignite extensions project has cdc-ext module which provides two way to setup cross cluster replication based on CDC. Detailed documentation can be found on page.