experiment-debezium-engine

An experiment with embedding Debezium Engine (the library) directly in an app.

Alternate ways to run Debezium include Debezium Server, Kafka Connect, Pulsar IO, Flink, and more.

A Hands-on Example

Start Postgres

docker-compose up

Build the project

./gradlew build

Run a simple example using the configuration from the example-simple-table-cdc.properties file

java -jar build/libs/experiment-debezium-engine-1.0-SNAPSHOT.jar example-simple-table-cdc.properties

Exec into Postgres

docker-compose exec db psql experiment experiment

Insert some data into the database

-- in db container

INSERT INTO schema1.users (full_name) VALUES ('kate smith');

The Debezium process will print something similar to the following

[
    EmbeddedEngineChangeEvent [
        key=null, 
        value=
            SourceRecord{
                sourcePartition={server=experiment}, 
                sourceOffset={transaction_id=null, lsn_proc=23594904, lsn_commit=23584544, lsn=23594904, txId=570, ts_usec=1652366692822782}
            }
            ConnectRecord{
                topic='experiment.schema1.users', 
                kafkaPartition=null, 
                key=Struct{id=3},
                keySchema=Schema{experiment.schema1.users.Key:STRUCT},
                value=Struct{
                    after=Struct{
                        id=3,
                        full_name=kate smith
                    },
                    source=Struct{
                        version=1.9.2.Final,
                        connector=postgresql,
                        name=experiment,
                        ts_ms=1652366692822,
                        db=experiment,
                        sequence=["23584544","23594904"],
                        schema=schema1,
                        table=users,
                        txId=570,
                        lsn=23594904
                    },
                    op=c,
                    ts_ms=1652366694164
                },
                valueSchema=Schema{experiment.schema1.users.Envelope:STRUCT}, 
                timestamp=null,
                headers=ConnectHeaders(headers=)
            },
        sourceRecord=
            SourceRecord{
                sourcePartition={server=experiment},
                sourceOffset={transaction_id=null, lsn_proc=23594904, lsn_commit=23584544, lsn=23594904, txId=570, ts_usec=1652366692822782}
            } 
            ConnectRecord{
                topic='experiment.schema1.users', 
                kafkaPartition=null,
                key=Struct{id=3}, 
                keySchema=Schema{experiment.schema1.users.Key:STRUCT}, 
                value=Struct{
                    after=Struct{
                        id=3,
                        full_name=kate smith
                    },
                    source=Struct{
                        version=1.9.2.Final,
                        connector=postgresql,
                        name=experiment,
                        ts_ms=1652366692822,
                        db=experiment,
                        sequence=["23584544","23594904"],
                        schema=schema1,
                        table=users,
                        txId=570,
                        lsn=23594904
                    },
                    op=c,
                    ts_ms=1652366694164
                }, 
                valueSchema=Schema{experiment.schema1.users.Envelope:STRUCT},
                timestamp=null, 
                headers=ConnectHeaders(headers=)
            }
    ]
]

The Debezium process should have created a replication slot on Postgres

-- in db container

SELECT * from pg_replication_slots;

--             slot_name             |  plugin  | slot_type | datoid |  database  | temporary | active | active_pid | xmin | catalog_xmin | restart_lsn | confirmed_flush_lsn 
-- ----------------------------------+----------+-----------+--------+------------+-----------+--------+------------+------+--------------+-------------+---------------------
--  experimentdebeziumenginetablecdc | pgoutput | logical   |  16384 | experiment | f         | t      |        103 |      |          570 | 0/16806F8   | 0/16806F8
-- (1 row)

Data Backfills & Snapshots

What if we also want to deliver pre-existing data for backfills?

No problem, we can trigger ad-hoc delivery of backfill snapshot data ad-hoc.

-- In db container
INSERT INTO schema1.debezium_signal (id, type, data) 
VALUES('ad-hoc-1', 'execute-snapshot', '{"data-collections": ["schema1.users", "schema2.users"],"type":"incremental"}');

Now you should see the preexisting rows print out.

Alternatively, backfill snapshots can be configured to be delivered in different scenarios

Transformations (SMT's)

For plugging in simple transformations, Debezium Engine supports Single Message Transformations (SMT) and provides several out of the box.

Transformation Example: Logical Table Router SMT

For example, the Logical Table Router might be useful for merging multiple physical shards into a single logical table.

e.g. merging db1.schema1.users and db1.schema2.users tables into a single all_dbs.all_shards.users logical table.

Restart the Debezium process using the configuration in the example-transform-merge-shards-into-logical-table.properties, file which enables the io.debezium.transforms.ByLogicalTableRouter transformation.

java -jar build/libs/experiment-debezium-engine-1.0-SNAPSHOT.jar example-transform-merge-shards-into-logical-table.properties

Insert some data into the database

-- in db container

INSERT INTO schema1.users (full_name) VALUES ('kate smith');

The Debezium process will print something similar to the following. Note how this affects topic, key.

[
    EmbeddedEngineChangeEvent [
        key=null, 
        value=
            SourceRecord{
                sourcePartition={server=experiment}, 
                sourceOffset={transaction_id=null, lsn_proc=23599552, lsn_commit=23599392, lsn=23599552, txId=584, ts_usec=1652367418051690}
            } 
            ConnectRecord{
                topic='persistent://public/default/all_dbs.all_shards.users', 
                kafkaPartition=null, 
                key=Struct{id=4,__dbz__physicalTableIdentifier=experiment.schema1.users}, 
                keySchema=Schema{persistent___public_default_all_dbs.all_shards.users.Key:STRUCT}, 
                value=Struct{
                    after=Struct{
                        id=4,
                        full_name=kate smith
                    },
                    source=Struct{
                        version=1.9.2.Final,
                        connector=postgresql,
                        name=experiment,
                        ts_ms=1652367418051,
                        db=experiment,
                        sequence=["23599392","23599552"],
                        schema=schema1,
                        table=users,
                        txId=584,
                        lsn=23599552
                    },
                    op=c,
                    ts_ms=1652367418131
                }, 
                valueSchema=Schema{persistent___public_default_all_dbs.all_shards.users.Envelope:STRUCT}, 
                timestamp=null, 
                headers=ConnectHeaders(headers=)
            }, 
        sourceRecord=
            SourceRecord{
                sourcePartition={server=experiment}, 
                sourceOffset={transaction_id=null, lsn_proc=23599552, lsn_commit=23599392, lsn=23599552, txId=584, ts_usec=1652367418051690}} 
            ConnectRecord{
                topic='persistent://public/default/all_dbs.all_shards.users', 
                kafkaPartition=null, 
                key=Struct{id=4,__dbz__physicalTableIdentifier=experiment.schema1.users}, 
                keySchema=Schema{persistent___public_default_all_dbs.all_shards.users.Key:STRUCT}, 
                value=Struct{
                    after=Struct{
                        id=4,
                        full_name=kate smith
                    },
                    source=Struct{
                        version=1.9.2.Final,
                        connector=postgresql,
                        name=experiment,
                        ts_ms=1652367418051,
                        db=experiment,
                        sequence=["23599392","23599552"],
                        schema=schema1,
                        table=users,
                        txId=584,
                        lsn=23599552
                    },
                    op=c,
                    ts_ms=1652367418131
                },
                valueSchema=Schema{persistent___public_default_all_dbs.all_shards.users.Envelope:STRUCT},
                timestamp=null, 
                headers=ConnectHeaders(headers=)
            }
    ]
]

Transformation Example: Outbox SMT

Now let's try a different configuration of Debezium Server, this time configured as a transactional outbox router. It will route payloads written to the outbox table in the database directly to the specified topic.

Restart the Debezium process using the configuration in the example-transform-outbox.properties file, which enables the io.debezium.transforms.outbox.EventRouter transformation.

java -jar build/libs/experiment-debezium-engine-1.0-SNAPSHOT.jar example-transform-outbox.properties

Write to the outbox table

-- in db container

INSERT INTO schema1.outbox (aggregatetype, aggregateid, type, payload) 
VALUES (
  'persistent://public/default/aggregatetype', 
  'aggregateid', 
  'type', 
  '{"hello":"world"}'::jsonb
 );

The Debezium process will print something similar to the following. Note how this affects topic, key, value.