1
0

[HUDI-1290] Add Debezium Source for deltastreamer (#4063)

* add source for postgres debezium

* Add tests for debezium payload

* Fix test

* Fix test

* Add tests for debezium source

* Add tests for debezium source

* Fix schema for debezium

* Fix checkstyle issues

* Fix config issue for schema registry

* Add mysql source for debezium

* Fix checkstyle issues an tests

* Improve code for merging toasted values

* Improve code for merging toasted values

Co-authored-by: Rajesh Mahindra <rmahindra@Rajeshs-MacBook-Pro.local>
This commit is contained in:
rmahindra123
2021-11-24 17:57:02 -08:00
committed by GitHub
parent abc0175cf7
commit 83f8ed2ae3
13 changed files with 1520 additions and 1 deletions

View File

@@ -0,0 +1,89 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.common.model.debezium;
import org.apache.hudi.common.model.OverwriteWithLatestAvroPayload;
import org.apache.hudi.common.util.Option;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericRecord;
import org.apache.avro.generic.IndexedRecord;
import org.apache.log4j.LogManager;
import org.apache.log4j.Logger;
import java.io.IOException;
/**
* Base class that provides support for seamlessly applying changes captured via Debezium.
* <p>
* Debezium change event types are determined for the op field in the payload
* <p>
* - For inserts, op=i
* - For deletes, op=d
* - For updates, op=u
* - For snapshort inserts, op=r
* <p>
* This payload implementation will issue matching insert, delete, updates against the hudi table
*/
public abstract class AbstractDebeziumAvroPayload extends OverwriteWithLatestAvroPayload {
private static final Logger LOG = LogManager.getLogger(AbstractDebeziumAvroPayload.class);
public AbstractDebeziumAvroPayload(GenericRecord record, Comparable orderingVal) {
super(record, orderingVal);
}
public AbstractDebeziumAvroPayload(Option<GenericRecord> record) {
super(record);
}
@Override
public Option<IndexedRecord> getInsertValue(Schema schema) throws IOException {
IndexedRecord insertRecord = getInsertRecord(schema);
return handleDeleteOperation(insertRecord);
}
@Override
public Option<IndexedRecord> combineAndGetUpdateValue(IndexedRecord currentValue, Schema schema) throws IOException {
// Step 1: If the time occurrence of the current record in storage is higher than the time occurrence of the
// insert record (including a delete record), pick the current record.
if (shouldPickCurrentRecord(currentValue, getInsertRecord(schema), schema)) {
return Option.of(currentValue);
}
// Step 2: Pick the insert record (as a delete record if its a deleted event)
return getInsertValue(schema);
}
protected abstract boolean shouldPickCurrentRecord(IndexedRecord currentRecord, IndexedRecord insertRecord, Schema schema) throws IOException;
private Option<IndexedRecord> handleDeleteOperation(IndexedRecord insertRecord) {
boolean delete = false;
if (insertRecord instanceof GenericRecord) {
GenericRecord record = (GenericRecord) insertRecord;
Object value = record.get(DebeziumConstants.FLATTENED_OP_COL_NAME);
delete = value != null && value.toString().equalsIgnoreCase(DebeziumConstants.DELETE_OP);
}
return delete ? Option.empty() : Option.of(insertRecord);
}
private IndexedRecord getInsertRecord(Schema schema) throws IOException {
return super.getInsertValue(schema).get();
}
}

View File

@@ -0,0 +1,81 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.common.model.debezium;
import java.util.Arrays;
import java.util.Collections;
import java.util.List;
/**
* Constants used by {@link DebeziumSource} and {@link DebeziumAvroPayload}.
*/
public class DebeziumConstants {
// INPUT COLUMNS
public static final String INCOMING_BEFORE_FIELD = "before";
public static final String INCOMING_AFTER_FIELD = "after";
public static final String INCOMING_SOURCE_FIELD = "source";
public static final String INCOMING_OP_FIELD = "op";
public static final String INCOMING_TS_MS_FIELD = "ts_ms";
public static final String INCOMING_SOURCE_NAME_FIELD = "source.name";
public static final String INCOMING_SOURCE_TS_MS_FIELD = "source.ts_ms";
public static final String INCOMING_SOURCE_TXID_FIELD = "source.txId";
// INPUT COLUMNS SPECIFIC TO MYSQL
public static final String INCOMING_SOURCE_FILE_FIELD = "source.file";
public static final String INCOMING_SOURCE_POS_FIELD = "source.pos";
public static final String INCOMING_SOURCE_ROW_FIELD = "source.row";
// INPUT COLUMNS SPECIFIC TO POSTGRES
public static final String INCOMING_SOURCE_LSN_FIELD = "source.lsn";
public static final String INCOMING_SOURCE_XMIN_FIELD = "source.xmin";
// OUTPUT COLUMNS
public static final String FLATTENED_OP_COL_NAME = "_change_operation_type";
public static final String UPSTREAM_PROCESSING_TS_COL_NAME = "_upstream_event_processed_ts_ms";
public static final String FLATTENED_SHARD_NAME = "db_shard_source_partition";
public static final String FLATTENED_TS_COL_NAME = "_event_origin_ts_ms";
public static final String FLATTENED_TX_ID_COL_NAME = "_event_tx_id";
// OUTPUT COLUMNS SPECIFIC TO MYSQL
public static final String FLATTENED_FILE_COL_NAME = "_event_bin_file";
public static final String FLATTENED_POS_COL_NAME = "_event_pos";
public static final String FLATTENED_ROW_COL_NAME = "_event_row";
public static final String ADDED_SEQ_COL_NAME = "_event_seq";
// OUTPUT COLUMNS SPECIFIC TO POSTGRES
public static final String FLATTENED_LSN_COL_NAME = "_event_lsn";
public static final String FLATTENED_XMIN_COL_NAME = "_event_xmin";
// Other Constants
public static final String DELETE_OP = "d";
// List of meta data columns
public static List<String> META_COLUMNS = Collections.unmodifiableList(Arrays.asList(
FLATTENED_OP_COL_NAME,
UPSTREAM_PROCESSING_TS_COL_NAME,
FLATTENED_TS_COL_NAME,
FLATTENED_TX_ID_COL_NAME,
FLATTENED_LSN_COL_NAME,
FLATTENED_XMIN_COL_NAME,
FLATTENED_SHARD_NAME
));
}

View File

@@ -0,0 +1,67 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.common.model.debezium;
import org.apache.hudi.common.util.Option;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericRecord;
import org.apache.avro.generic.IndexedRecord;
import org.apache.log4j.LogManager;
import org.apache.log4j.Logger;
import java.io.IOException;
/**
* Provides support for seamlessly applying changes captured via Debezium for MysqlDB.
* <p>
* Debezium change event types are determined for the op field in the payload
* <p>
* - For inserts, op=i
* - For deletes, op=d
* - For updates, op=u
* - For snapshort inserts, op=r
* <p>
* This payload implementation will issue matching insert, delete, updates against the hudi table
*/
public class MySqlDebeziumAvroPayload extends AbstractDebeziumAvroPayload {
private static final Logger LOG = LogManager.getLogger(MySqlDebeziumAvroPayload.class);
public MySqlDebeziumAvroPayload(GenericRecord record, Comparable orderingVal) {
super(record, orderingVal);
}
public MySqlDebeziumAvroPayload(Option<GenericRecord> record) {
super(record);
}
private String extractSeq(IndexedRecord record) {
return ((CharSequence) ((GenericRecord) record).get(DebeziumConstants.ADDED_SEQ_COL_NAME)).toString();
}
@Override
protected boolean shouldPickCurrentRecord(IndexedRecord currentRecord, IndexedRecord insertRecord, Schema schema) throws IOException {
String currentSourceSeq = extractSeq(currentRecord);
String insertSourceSeq = extractSeq(insertRecord);
// Pick the current value in storage only if its Seq (file+pos) is latest
// compared to the Seq (file+pos) of the insert value
return insertSourceSeq.compareTo(currentSourceSeq) < 0;
}
}

View File

@@ -0,0 +1,130 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.common.model.debezium;
import org.apache.hudi.common.util.Option;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericRecord;
import org.apache.avro.generic.IndexedRecord;
import org.apache.log4j.LogManager;
import org.apache.log4j.Logger;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.nio.charset.StandardCharsets;
import java.util.List;
/**
* Provides support for seamlessly applying changes captured via Debezium for PostgresDB.
* <p>
* Debezium change event types are determined for the op field in the payload
* <p>
* - For inserts, op=i
* - For deletes, op=d
* - For updates, op=u
* - For snapshort inserts, op=r
* <p>
* This payload implementation will issue matching insert, delete, updates against the hudi table
*/
public class PostgresDebeziumAvroPayload extends AbstractDebeziumAvroPayload {
private static final Logger LOG = LogManager.getLogger(PostgresDebeziumAvroPayload.class);
public static final String DEBEZIUM_TOASTED_VALUE = "__debezium_unavailable_value";
public PostgresDebeziumAvroPayload(GenericRecord record, Comparable orderingVal) {
super(record, orderingVal);
}
public PostgresDebeziumAvroPayload(Option<GenericRecord> record) {
super(record);
}
private Long extractLSN(IndexedRecord record) {
GenericRecord genericRecord = (GenericRecord) record;
return (Long) genericRecord.get(DebeziumConstants.FLATTENED_LSN_COL_NAME);
}
@Override
protected boolean shouldPickCurrentRecord(IndexedRecord currentRecord, IndexedRecord insertRecord, Schema schema) throws IOException {
Long currentSourceLSN = extractLSN(currentRecord);
Long insertSourceLSN = extractLSN(insertRecord);
// Pick the current value in storage only if its LSN is latest compared to the LSN of the insert value
return insertSourceLSN < currentSourceLSN;
}
@Override
public Option<IndexedRecord> combineAndGetUpdateValue(IndexedRecord currentValue, Schema schema) throws IOException {
// Specific to Postgres: If the updated record has TOASTED columns,
// we will need to keep the previous value for those columns
// see https://debezium.io/documentation/reference/connectors/postgresql.html#postgresql-toasted-values
Option<IndexedRecord> insertOrDeleteRecord = super.combineAndGetUpdateValue(currentValue, schema);
if (insertOrDeleteRecord.isPresent()) {
mergeToastedValuesIfPresent(insertOrDeleteRecord.get(), currentValue);
}
return insertOrDeleteRecord;
}
private void mergeToastedValuesIfPresent(IndexedRecord incomingRecord, IndexedRecord currentRecord) {
List<Schema.Field> fields = incomingRecord.getSchema().getFields();
fields.forEach(field -> {
// There are only four avro data types that have unconstrained sizes, which are
// NON-NULLABLE STRING, NULLABLE STRING, NON-NULLABLE BYTES, NULLABLE BYTES
if (((GenericData.Record) incomingRecord).get(field.name()) != null
&& (containsStringToastedValues(incomingRecord, field) || containsBytesToastedValues(incomingRecord, field))) {
((GenericData.Record) incomingRecord).put(field.name(), ((GenericData.Record) currentRecord).get(field.name()));
}
});
}
/**
* Returns true if a column is either of type string or a union of one or more strings that contain a debezium toasted value.
*
* @param incomingRecord The incoming avro record
* @param field the column of interest
* @return
*/
private boolean containsStringToastedValues(IndexedRecord incomingRecord, Schema.Field field) {
return ((field.schema().getType() == Schema.Type.STRING
|| (field.schema().getType() == Schema.Type.UNION && field.schema().getTypes().stream().anyMatch(s -> s.getType() == Schema.Type.STRING)))
// Check length first as an optimization
&& ((CharSequence) ((GenericData.Record) incomingRecord).get(field.name())).length() == DEBEZIUM_TOASTED_VALUE.length()
&& DEBEZIUM_TOASTED_VALUE.equals(((CharSequence) ((GenericData.Record) incomingRecord).get(field.name())).toString()));
}
/**
* Returns true if a column is either of type bytes or a union of one or more bytes that contain a debezium toasted value.
*
* @param incomingRecord The incoming avro record
* @param field the column of interest
* @return
*/
private boolean containsBytesToastedValues(IndexedRecord incomingRecord, Schema.Field field) {
return ((field.schema().getType() == Schema.Type.BYTES
|| (field.schema().getType() == Schema.Type.UNION && field.schema().getTypes().stream().anyMatch(s -> s.getType() == Schema.Type.BYTES)))
// Check length first as an optimization
&& ((ByteBuffer) ((GenericData.Record) incomingRecord).get(field.name())).array().length == DEBEZIUM_TOASTED_VALUE.length()
&& DEBEZIUM_TOASTED_VALUE.equals(new String(((ByteBuffer) ((GenericData.Record) incomingRecord).get(field.name())).array(), StandardCharsets.UTF_8)));
}
}

View File

@@ -0,0 +1,131 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or mo contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.common.model.debezium;
import org.apache.hudi.common.util.Option;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericRecord;
import org.apache.avro.generic.IndexedRecord;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Test;
import java.io.IOException;
import java.util.Arrays;
import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertFalse;
public class TestMySqlDebeziumAvroPayload {
private static final String KEY_FIELD_NAME = "Key";
private Schema avroSchema;
@BeforeEach
void setUp() {
this.avroSchema = Schema.createRecord(Arrays.asList(
new Schema.Field(KEY_FIELD_NAME, Schema.create(Schema.Type.INT), "", 0),
new Schema.Field(DebeziumConstants.FLATTENED_OP_COL_NAME, Schema.create(Schema.Type.STRING), "", null),
new Schema.Field(DebeziumConstants.ADDED_SEQ_COL_NAME, Schema.create(Schema.Type.STRING), "", null)
));
}
@Test
public void testInsert() throws IOException {
GenericRecord insertRecord = createRecord(0, Operation.INSERT, "00001.111");
MySqlDebeziumAvroPayload payload = new MySqlDebeziumAvroPayload(insertRecord, "00001.111");
validateRecord(payload.getInsertValue(avroSchema), 0, Operation.INSERT, "00001.111");
}
@Test
public void testPreCombine() {
GenericRecord insertRecord = createRecord(0, Operation.INSERT, "00002.111");
MySqlDebeziumAvroPayload insertPayload = new MySqlDebeziumAvroPayload(insertRecord, "00002.111");
GenericRecord updateRecord = createRecord(0, Operation.UPDATE, "00001.111");
MySqlDebeziumAvroPayload updatePayload = new MySqlDebeziumAvroPayload(updateRecord, "00001.111");
GenericRecord deleteRecord = createRecord(0, Operation.DELETE, "00002.11");
MySqlDebeziumAvroPayload deletePayload = new MySqlDebeziumAvroPayload(deleteRecord, "00002.11");
assertEquals(insertPayload, insertPayload.preCombine(updatePayload));
assertEquals(deletePayload, deletePayload.preCombine(updatePayload));
assertEquals(insertPayload, deletePayload.preCombine(insertPayload));
}
@Test
public void testMergeWithUpdate() throws IOException {
GenericRecord updateRecord = createRecord(1, Operation.UPDATE, "00002.11");
MySqlDebeziumAvroPayload payload = new MySqlDebeziumAvroPayload(updateRecord, "00002.11");
GenericRecord existingRecord = createRecord(1, Operation.INSERT, "00001.111");
Option<IndexedRecord> mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
validateRecord(mergedRecord, 1, Operation.UPDATE, "00002.11");
GenericRecord lateRecord = createRecord(1, Operation.UPDATE, "00000.222");
payload = new MySqlDebeziumAvroPayload(lateRecord, "00000.222");
mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
validateRecord(mergedRecord, 1, Operation.INSERT, "00001.111");
}
@Test
public void testMergeWithDelete() throws IOException {
GenericRecord deleteRecord = createRecord(2, Operation.DELETE, "00002.11");
MySqlDebeziumAvroPayload payload = new MySqlDebeziumAvroPayload(deleteRecord, "00002.11");
GenericRecord existingRecord = createRecord(2, Operation.UPDATE, "00001.111");
Option<IndexedRecord> mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
// expect nothing to be committed to table
assertFalse(mergedRecord.isPresent());
GenericRecord lateRecord = createRecord(2, Operation.DELETE, "00000.222");
payload = new MySqlDebeziumAvroPayload(lateRecord, "00000.222");
mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
validateRecord(mergedRecord, 2, Operation.UPDATE, "00001.111");
}
private GenericRecord createRecord(int primaryKeyValue, Operation op, String seqValue) {
GenericRecord record = new GenericData.Record(avroSchema);
record.put(KEY_FIELD_NAME, primaryKeyValue);
record.put(DebeziumConstants.FLATTENED_OP_COL_NAME, op.op);
record.put(DebeziumConstants.ADDED_SEQ_COL_NAME, seqValue);
return record;
}
private void validateRecord(Option<IndexedRecord> iRecord, int primaryKeyValue, Operation op, String seqValue) {
IndexedRecord record = iRecord.get();
assertEquals(primaryKeyValue, (int) record.get(0));
assertEquals(op.op, record.get(1).toString());
assertEquals(seqValue, record.get(2).toString());
}
private enum Operation {
INSERT("c"),
UPDATE("u"),
DELETE("d");
public final String op;
Operation(String op) {
this.op = op;
}
}
}

View File

@@ -0,0 +1,181 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or mo contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.common.model.debezium;
import org.apache.hudi.common.util.Option;
import org.apache.avro.Schema;
import org.apache.avro.SchemaBuilder;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericRecord;
import org.apache.avro.generic.IndexedRecord;
import org.apache.avro.util.Utf8;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Test;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.nio.charset.StandardCharsets;
import java.util.Arrays;
import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertFalse;
import static org.junit.jupiter.api.Assertions.assertNull;
public class TestPostgresDebeziumAvroPayload {
private static final String KEY_FIELD_NAME = "Key";
private Schema avroSchema;
@BeforeEach
void setUp() {
this.avroSchema = Schema.createRecord(Arrays.asList(
new Schema.Field(KEY_FIELD_NAME, Schema.create(Schema.Type.INT), "", 0),
new Schema.Field(DebeziumConstants.FLATTENED_OP_COL_NAME, Schema.create(Schema.Type.STRING), "", null),
new Schema.Field(DebeziumConstants.FLATTENED_LSN_COL_NAME, Schema.create(Schema.Type.LONG), "", null)
));
}
@Test
public void testInsert() throws IOException {
GenericRecord insertRecord = createRecord(0, Operation.INSERT, 100L);
PostgresDebeziumAvroPayload payload = new PostgresDebeziumAvroPayload(insertRecord, 100L);
validateRecord(payload.getInsertValue(avroSchema), 0, Operation.INSERT, 100L);
}
@Test
public void testPreCombine() {
GenericRecord insertRecord = createRecord(0, Operation.INSERT, 120L);
PostgresDebeziumAvroPayload insertPayload = new PostgresDebeziumAvroPayload(insertRecord, 120L);
GenericRecord updateRecord = createRecord(0, Operation.UPDATE, 99L);
PostgresDebeziumAvroPayload updatePayload = new PostgresDebeziumAvroPayload(updateRecord, 99L);
GenericRecord deleteRecord = createRecord(0, Operation.DELETE, 111L);
PostgresDebeziumAvroPayload deletePayload = new PostgresDebeziumAvroPayload(deleteRecord, 111L);
assertEquals(insertPayload, insertPayload.preCombine(updatePayload));
assertEquals(deletePayload, deletePayload.preCombine(updatePayload));
assertEquals(insertPayload, deletePayload.preCombine(insertPayload));
}
@Test
public void testMergeWithUpdate() throws IOException {
GenericRecord updateRecord = createRecord(1, Operation.UPDATE, 100L);
PostgresDebeziumAvroPayload payload = new PostgresDebeziumAvroPayload(updateRecord, 100L);
GenericRecord existingRecord = createRecord(1, Operation.INSERT, 99L);
Option<IndexedRecord> mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
validateRecord(mergedRecord, 1, Operation.UPDATE, 100L);
GenericRecord lateRecord = createRecord(1, Operation.UPDATE, 98L);
payload = new PostgresDebeziumAvroPayload(lateRecord, 98L);
mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
validateRecord(mergedRecord, 1, Operation.INSERT, 99L);
}
@Test
public void testMergeWithDelete() throws IOException {
GenericRecord deleteRecord = createRecord(2, Operation.DELETE, 100L);
PostgresDebeziumAvroPayload payload = new PostgresDebeziumAvroPayload(deleteRecord, 100L);
GenericRecord existingRecord = createRecord(2, Operation.UPDATE, 99L);
Option<IndexedRecord> mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
// expect nothing to be committed to table
assertFalse(mergedRecord.isPresent());
GenericRecord lateRecord = createRecord(2, Operation.DELETE, 98L);
payload = new PostgresDebeziumAvroPayload(lateRecord, 98L);
mergedRecord = payload.combineAndGetUpdateValue(existingRecord, avroSchema);
validateRecord(mergedRecord, 2, Operation.UPDATE, 99L);
}
@Test
public void testMergeWithToastedValues() throws IOException {
Schema avroSchema = SchemaBuilder.builder()
.record("test_schema")
.namespace("test_namespace")
.fields()
.name(DebeziumConstants.FLATTENED_LSN_COL_NAME).type().longType().noDefault()
.name("string_col").type().stringType().noDefault()
.name("byte_col").type().bytesType().noDefault()
.name("string_null_col_1").type().nullable().stringType().noDefault()
.name("byte_null_col_1").type().nullable().bytesType().noDefault()
.name("string_null_col_2").type().nullable().stringType().noDefault()
.name("byte_null_col_2").type().nullable().bytesType().noDefault()
.endRecord();
GenericRecord oldVal = new GenericData.Record(avroSchema);
oldVal.put(DebeziumConstants.FLATTENED_LSN_COL_NAME, 100L);
oldVal.put("string_col", "valid string value");
oldVal.put("byte_col", ByteBuffer.wrap("valid byte value".getBytes()));
oldVal.put("string_null_col_1", "valid string value");
oldVal.put("byte_null_col_1", ByteBuffer.wrap("valid byte value".getBytes()));
oldVal.put("string_null_col_2", null);
oldVal.put("byte_null_col_2", null);
GenericRecord newVal = new GenericData.Record(avroSchema);
newVal.put(DebeziumConstants.FLATTENED_LSN_COL_NAME, 105L);
newVal.put("string_col", PostgresDebeziumAvroPayload.DEBEZIUM_TOASTED_VALUE);
newVal.put("byte_col", ByteBuffer.wrap(PostgresDebeziumAvroPayload.DEBEZIUM_TOASTED_VALUE.getBytes()));
newVal.put("string_null_col_1", null);
newVal.put("byte_null_col_1", null);
newVal.put("string_null_col_2", "valid string value");
newVal.put("byte_null_col_2", ByteBuffer.wrap("valid byte value".getBytes()));
PostgresDebeziumAvroPayload payload = new PostgresDebeziumAvroPayload(Option.of(newVal));
GenericRecord outputRecord = (GenericRecord) payload
.combineAndGetUpdateValue(oldVal, avroSchema).get();
assertEquals("valid string value", outputRecord.get("string_col"));
assertEquals("valid byte value", new String(((ByteBuffer) outputRecord.get("byte_col")).array(), StandardCharsets.UTF_8));
assertNull(outputRecord.get("string_null_col_1"));
assertNull(outputRecord.get("byte_null_col_1"));
assertEquals("valid string value", ((Utf8) outputRecord.get("string_null_col_2")).toString());
assertEquals("valid byte value", new String(((ByteBuffer) outputRecord.get("byte_null_col_2")).array(), StandardCharsets.UTF_8));
}
private GenericRecord createRecord(int primaryKeyValue, Operation op, long lsnValue) {
GenericRecord record = new GenericData.Record(avroSchema);
record.put(KEY_FIELD_NAME, primaryKeyValue);
record.put(DebeziumConstants.FLATTENED_OP_COL_NAME, op.op);
record.put(DebeziumConstants.FLATTENED_LSN_COL_NAME, lsnValue);
return record;
}
private void validateRecord(Option<IndexedRecord> iRecord, int primaryKeyValue, Operation op, long lsnValue) {
IndexedRecord record = iRecord.get();
assertEquals(primaryKeyValue, (int) record.get(0));
assertEquals(op.op, record.get(1).toString());
assertEquals(lsnValue, (long) record.get(2));
}
private enum Operation {
INSERT("c"),
UPDATE("u"),
DELETE("d");
public final String op;
Operation(String op) {
this.op = op;
}
}
}

View File

@@ -49,7 +49,7 @@ public class SchemaRegistryProvider extends SchemaProvider {
*/
public static class Config {
private static final String SRC_SCHEMA_REGISTRY_URL_PROP = "hoodie.deltastreamer.schemaprovider.registry.url";
public static final String SRC_SCHEMA_REGISTRY_URL_PROP = "hoodie.deltastreamer.schemaprovider.registry.url";
private static final String TARGET_SCHEMA_REGISTRY_URL_PROP =
"hoodie.deltastreamer.schemaprovider.registry.targetUrl";
}

View File

@@ -0,0 +1,243 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.utilities.sources.debezium;
import java.io.IOException;
import java.util.Arrays;
import java.util.List;
import java.util.stream.Collectors;
import org.apache.avro.Schema;
import org.apache.avro.Schema.Field;
import org.apache.avro.Schema.Type;
import org.apache.avro.generic.GenericRecord;
import org.apache.hudi.AvroConversionUtils;
import org.apache.hudi.DataSourceWriteOptions;
import org.apache.hudi.common.config.TypedProperties;
import org.apache.hudi.common.util.Option;
import org.apache.hudi.common.util.collection.Pair;
import org.apache.hudi.exception.HoodieException;
import org.apache.hudi.utilities.deltastreamer.HoodieDeltaStreamerMetrics;
import org.apache.hudi.utilities.schema.SchemaProvider;
import org.apache.hudi.utilities.schema.SchemaRegistryProvider;
import org.apache.hudi.utilities.sources.RowSource;
import org.apache.hudi.utilities.sources.helpers.AvroConvertor;
import org.apache.hudi.utilities.sources.helpers.KafkaOffsetGen;
import org.apache.hudi.utilities.sources.helpers.KafkaOffsetGen.CheckpointUtils;
import org.apache.kafka.common.serialization.StringDeserializer;
import org.apache.log4j.LogManager;
import org.apache.log4j.Logger;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SparkSession;
import org.apache.spark.sql.functions;
import org.apache.spark.sql.types.DataTypes;
import org.apache.spark.sql.types.StructField;
import org.apache.spark.sql.types.StructType;
import org.apache.spark.streaming.kafka010.KafkaUtils;
import org.apache.spark.streaming.kafka010.LocationStrategies;
import org.apache.spark.streaming.kafka010.OffsetRange;
/**
* Base class for Debezium streaming source which expects change events as Kafka Avro records.
* Obtains the schema from the confluent schema-registry.
*/
public abstract class DebeziumSource extends RowSource {
private static final Logger LOG = LogManager.getLogger(DebeziumSource.class);
// these are native kafka's config. do not change the config names.
private static final String NATIVE_KAFKA_KEY_DESERIALIZER_PROP = "key.deserializer";
private static final String NATIVE_KAFKA_VALUE_DESERIALIZER_PROP = "value.deserializer";
private static final String OVERRIDE_CHECKPOINT_STRING = "hoodie.debezium.override.initial.checkpoint.key";
private static final String CONNECT_NAME_KEY = "connect.name";
private static final String DATE_CONNECT_NAME = "custom.debezium.DateString";
private final KafkaOffsetGen offsetGen;
private final HoodieDeltaStreamerMetrics metrics;
private final SchemaRegistryProvider schemaRegistryProvider;
private final String deserializerClassName;
public DebeziumSource(TypedProperties props, JavaSparkContext sparkContext,
SparkSession sparkSession,
SchemaProvider schemaProvider,
HoodieDeltaStreamerMetrics metrics) {
super(props, sparkContext, sparkSession, schemaProvider);
props.put(NATIVE_KAFKA_KEY_DESERIALIZER_PROP, StringDeserializer.class);
deserializerClassName = props.getString(DataSourceWriteOptions.KAFKA_AVRO_VALUE_DESERIALIZER_CLASS().key(),
DataSourceWriteOptions.KAFKA_AVRO_VALUE_DESERIALIZER_CLASS().defaultValue());
try {
props.put(NATIVE_KAFKA_VALUE_DESERIALIZER_PROP, Class.forName(deserializerClassName));
} catch (ClassNotFoundException e) {
String error = "Could not load custom avro kafka deserializer: " + deserializerClassName;
LOG.error(error);
throw new HoodieException(error, e);
}
// Currently, debezium source requires Confluent/Kafka schema-registry to fetch the latest schema.
if (schemaProvider == null || !(schemaProvider instanceof SchemaRegistryProvider)) {
schemaRegistryProvider = new SchemaRegistryProvider(props, sparkContext);
} else {
schemaRegistryProvider = (SchemaRegistryProvider) schemaProvider;
}
offsetGen = new KafkaOffsetGen(props);
this.metrics = metrics;
}
@Override
protected Pair<Option<Dataset<Row>>, String> fetchNextBatch(Option<String> lastCkptStr, long sourceLimit) {
String overrideCheckpointStr = props.getString(OVERRIDE_CHECKPOINT_STRING, "");
OffsetRange[] offsetRanges = offsetGen.getNextOffsetRanges(lastCkptStr, sourceLimit, metrics);
long totalNewMsgs = CheckpointUtils.totalNewMessages(offsetRanges);
LOG.info("About to read " + totalNewMsgs + " from Kafka for topic :" + offsetGen.getTopicName());
if (totalNewMsgs == 0) {
// If there are no new messages, use empty dataframe with no schema. This is because the schema from schema registry can only be considered
// up to date if a change event has occurred.
return Pair.of(Option.of(sparkSession.emptyDataFrame()), overrideCheckpointStr.isEmpty() ? CheckpointUtils.offsetsToStr(offsetRanges) : overrideCheckpointStr);
} else {
try {
String schemaStr = schemaRegistryProvider.fetchSchemaFromRegistry(props.getString(SchemaRegistryProvider.Config.SRC_SCHEMA_REGISTRY_URL_PROP));
Dataset<Row> dataset = toDataset(offsetRanges, offsetGen, schemaStr);
LOG.info(String.format("Spark schema of Kafka Payload for topic %s:\n%s", offsetGen.getTopicName(), dataset.schema().treeString()));
LOG.info(String.format("New checkpoint string: %s", CheckpointUtils.offsetsToStr(offsetRanges)));
return Pair.of(Option.of(dataset), overrideCheckpointStr.isEmpty() ? CheckpointUtils.offsetsToStr(offsetRanges) : overrideCheckpointStr);
} catch (IOException exc) {
LOG.error("Fatal error reading and parsing incoming debezium event", exc);
throw new HoodieException("Fatal error reading and parsing incoming debezium event", exc);
}
}
}
/**
* Debezium Kafka Payload has a nested structure, flatten it specific to the Database type.
* @param rawKafkaData Dataset of the Debezium CDC event from the kafka
* @return A flattened dataset.
*/
protected abstract Dataset<Row> processDataset(Dataset<Row> rawKafkaData);
/**
* Converts a Kafka Topic offset into a Spark dataset.
*
* @param offsetRanges Offset ranges
* @param offsetGen KafkaOffsetGen
* @return Spark dataset
*/
private Dataset<Row> toDataset(OffsetRange[] offsetRanges, KafkaOffsetGen offsetGen, String schemaStr) {
AvroConvertor convertor = new AvroConvertor(schemaStr);
Dataset<Row> kafkaData;
if (deserializerClassName.equals(StringDeserializer.class.getName())) {
kafkaData = AvroConversionUtils.createDataFrame(
KafkaUtils.<String, String>createRDD(sparkContext, offsetGen.getKafkaParams(), offsetRanges, LocationStrategies.PreferConsistent())
.map(obj -> convertor.fromJson(obj.value()))
.rdd(), schemaStr, sparkSession);
} else {
kafkaData = AvroConversionUtils.createDataFrame(
KafkaUtils.createRDD(sparkContext, offsetGen.getKafkaParams(), offsetRanges, LocationStrategies.PreferConsistent())
.map(obj -> (GenericRecord) obj.value())
.rdd(), schemaStr, sparkSession);
}
// Flatten debezium payload, specific to each DB type (postgres/ mysql/ etc..)
Dataset<Row> debeziumDataset = processDataset(kafkaData);
// Some required transformations to ensure debezium data types are converted to spark supported types.
return convertArrayColumnsToString(convertColumnToNullable(sparkSession,
convertDateColumns(debeziumDataset, new Schema.Parser().parse(schemaStr))));
}
/**
* Converts string formatted date columns into Spark date columns.
*
* @param dataset Spark dataset
* @param schema Avro schema from Debezium
* @return Converted dataset
*/
public static Dataset<Row> convertDateColumns(Dataset<Row> dataset, Schema schema) {
if (schema.getField("before") != null) {
List<String> dateFields = schema.getField("before")
.schema()
.getTypes()
.get(1)
.getFields()
.stream()
.filter(field -> {
if (field.schema().getType() == Type.UNION) {
return field.schema().getTypes().stream().anyMatch(
schemaInUnion -> DATE_CONNECT_NAME.equals(schemaInUnion.getProp(CONNECT_NAME_KEY))
);
} else {
return DATE_CONNECT_NAME.equals(field.schema().getProp(CONNECT_NAME_KEY));
}
}).map(Field::name).collect(Collectors.toList());
LOG.info("Date fields: " + dateFields.toString());
for (String dateCol : dateFields) {
dataset = dataset.withColumn(dateCol, functions.col(dateCol).cast(DataTypes.DateType));
}
}
return dataset;
}
/**
* Utility function for converting columns to nullable. This is useful when required to make a column nullable to match a nullable column from Debezium change
* events.
*
* @param sparkSession SparkSession object
* @param dataset Dataframe to modify
* @return Modified dataframe
*/
private static Dataset<Row> convertColumnToNullable(SparkSession sparkSession, Dataset<Row> dataset) {
List<String> columns = Arrays.asList(dataset.columns());
StructField[] modifiedStructFields = Arrays.stream(dataset.schema().fields()).map(field -> columns
.contains(field.name()) ? new StructField(field.name(), field.dataType(), true, field.metadata()) : field)
.toArray(StructField[]::new);
return sparkSession.createDataFrame(dataset.rdd(), new StructType(modifiedStructFields));
}
/**
* Converts Array types to String types because not all Debezium array columns are supported to be converted
* to Spark array columns.
*
* @param dataset Dataframe to modify
* @return Modified dataframe
*/
private static Dataset<Row> convertArrayColumnsToString(Dataset<Row> dataset) {
List<String> arrayColumns = Arrays.stream(dataset.schema().fields())
.filter(field -> field.dataType().typeName().toLowerCase().startsWith("array"))
.map(StructField::name)
.collect(Collectors.toList());
for (String colName : arrayColumns) {
dataset = dataset.withColumn(colName, functions.col(colName).cast(DataTypes.StringType));
}
return dataset;
}
}

View File

@@ -0,0 +1,104 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.utilities.sources.debezium;
import org.apache.hudi.common.config.TypedProperties;
import org.apache.hudi.common.model.debezium.DebeziumConstants;
import org.apache.hudi.utilities.deltastreamer.HoodieDeltaStreamerMetrics;
import org.apache.hudi.utilities.schema.SchemaProvider;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SQLContext;
import org.apache.spark.sql.SparkSession;
import org.apache.spark.sql.api.java.UDF2;
import org.apache.spark.sql.types.DataTypes;
import static org.apache.spark.sql.functions.callUDF;
/**
* Source for incrementally ingesting debezium generated change logs for Mysql DB.
*/
public class MysqlDebeziumSource extends DebeziumSource {
private final SQLContext sqlContext;
private final String generateUniqueSeqUdfFn = "mysql_generate_order_key";
public MysqlDebeziumSource(TypedProperties props, JavaSparkContext sparkContext,
SparkSession sparkSession,
SchemaProvider schemaProvider,
HoodieDeltaStreamerMetrics metrics) {
super(props, sparkContext, sparkSession, schemaProvider, metrics);
this.sqlContext = sparkSession.sqlContext();
sqlContext.udf().register(generateUniqueSeqUdfFn, (UDF2<String, Long, String>) MysqlDebeziumSource::generateUniqueSequence, DataTypes.StringType);
}
/**
* Debezium Kafka Payload has a nested structure (see https://debezium.io/documentation/reference/1.4/connectors/mysql.html).
* This function flattens this nested structure for the Mysql data, and also extracts a subset of Debezium metadata fields.
*
* @param rowDataset Dataset containing Debezium Payloads
* @return New dataset with flattened columns
*/
@Override
protected Dataset<Row> processDataset(Dataset<Row> rowDataset) {
Dataset<Row> flattenedDataset = rowDataset;
if (rowDataset.columns().length > 0) {
// Only flatten for non-empty schemas
Dataset<Row> insertedOrUpdatedData = rowDataset
.selectExpr(
String.format("%s as %s", DebeziumConstants.INCOMING_OP_FIELD, DebeziumConstants.FLATTENED_OP_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_TS_MS_FIELD, DebeziumConstants.UPSTREAM_PROCESSING_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_NAME_FIELD, DebeziumConstants.FLATTENED_SHARD_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TS_MS_FIELD, DebeziumConstants.FLATTENED_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TXID_FIELD, DebeziumConstants.FLATTENED_TX_ID_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_FILE_FIELD, DebeziumConstants.FLATTENED_FILE_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_POS_FIELD, DebeziumConstants.FLATTENED_POS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_ROW_FIELD, DebeziumConstants.FLATTENED_ROW_COL_NAME),
String.format("%s.*", DebeziumConstants.INCOMING_AFTER_FIELD)
)
.filter(rowDataset.col(DebeziumConstants.INCOMING_OP_FIELD).notEqual(DebeziumConstants.DELETE_OP));
Dataset<Row> deletedData = rowDataset
.selectExpr(
String.format("%s as %s", DebeziumConstants.INCOMING_OP_FIELD, DebeziumConstants.FLATTENED_OP_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_TS_MS_FIELD, DebeziumConstants.UPSTREAM_PROCESSING_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_NAME_FIELD, DebeziumConstants.FLATTENED_SHARD_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TS_MS_FIELD, DebeziumConstants.FLATTENED_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TXID_FIELD, DebeziumConstants.FLATTENED_TX_ID_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_FILE_FIELD, DebeziumConstants.FLATTENED_FILE_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_POS_FIELD, DebeziumConstants.FLATTENED_POS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_ROW_FIELD, DebeziumConstants.FLATTENED_ROW_COL_NAME),
String.format("%s.*", DebeziumConstants.INCOMING_BEFORE_FIELD)
)
.filter(rowDataset.col(DebeziumConstants.INCOMING_OP_FIELD).equalTo(DebeziumConstants.DELETE_OP));
flattenedDataset = insertedOrUpdatedData.union(deletedData);
}
return flattenedDataset.withColumn(DebeziumConstants.ADDED_SEQ_COL_NAME,
callUDF(generateUniqueSeqUdfFn, flattenedDataset.col(DebeziumConstants.FLATTENED_FILE_COL_NAME),
flattenedDataset.col(DebeziumConstants.FLATTENED_POS_COL_NAME)));
}
private static String generateUniqueSequence(String fileId, Long pos) {
return fileId.substring(fileId.lastIndexOf('.') + 1).concat("." + pos);
}
}

View File

@@ -0,0 +1,87 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.utilities.sources.debezium;
import org.apache.hudi.common.config.TypedProperties;
import org.apache.hudi.common.model.debezium.DebeziumConstants;
import org.apache.hudi.utilities.deltastreamer.HoodieDeltaStreamerMetrics;
import org.apache.hudi.utilities.schema.SchemaProvider;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SparkSession;
/**
* Source for incrementally ingesting debezium generated change logs for PostgresDB.
*/
public class PostgresDebeziumSource extends DebeziumSource {
public PostgresDebeziumSource(TypedProperties props, JavaSparkContext sparkContext,
SparkSession sparkSession,
SchemaProvider schemaProvider,
HoodieDeltaStreamerMetrics metrics) {
super(props, sparkContext, sparkSession, schemaProvider, metrics);
}
/**
* Debezium Kafka Payload has a nested structure (see https://debezium.io/documentation/reference/1.4/connectors/postgresql.html#postgresql-create-events).
* This function flattens this nested structure for the Postgres data, and also extracts a subset of Debezium metadata fields.
*
* @param rowDataset Dataset containing Debezium Payloads
* @return New dataset with flattened columns
*/
@Override
protected Dataset<Row> processDataset(Dataset<Row> rowDataset) {
if (rowDataset.columns().length > 0) {
// Pick selective debezium and postgres meta fields: pick the row values from before field for delete record
// and row values from after field for insert or update records.
Dataset<Row> insertedOrUpdatedData = rowDataset
.selectExpr(
String.format("%s as %s", DebeziumConstants.INCOMING_OP_FIELD, DebeziumConstants.FLATTENED_OP_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_TS_MS_FIELD, DebeziumConstants.UPSTREAM_PROCESSING_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_NAME_FIELD, DebeziumConstants.FLATTENED_SHARD_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TS_MS_FIELD, DebeziumConstants.FLATTENED_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TXID_FIELD, DebeziumConstants.FLATTENED_TX_ID_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_LSN_FIELD, DebeziumConstants.FLATTENED_LSN_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_XMIN_FIELD, DebeziumConstants.FLATTENED_XMIN_COL_NAME),
String.format("%s.*", DebeziumConstants.INCOMING_AFTER_FIELD)
)
.filter(rowDataset.col(DebeziumConstants.INCOMING_OP_FIELD).notEqual(DebeziumConstants.DELETE_OP));
Dataset<Row> deletedData = rowDataset
.selectExpr(
String.format("%s as %s", DebeziumConstants.INCOMING_OP_FIELD, DebeziumConstants.FLATTENED_OP_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_TS_MS_FIELD, DebeziumConstants.UPSTREAM_PROCESSING_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_NAME_FIELD, DebeziumConstants.FLATTENED_SHARD_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TS_MS_FIELD, DebeziumConstants.FLATTENED_TS_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_TXID_FIELD, DebeziumConstants.FLATTENED_TX_ID_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_LSN_FIELD, DebeziumConstants.FLATTENED_LSN_COL_NAME),
String.format("%s as %s", DebeziumConstants.INCOMING_SOURCE_XMIN_FIELD, DebeziumConstants.FLATTENED_XMIN_COL_NAME),
String.format("%s.*", DebeziumConstants.INCOMING_BEFORE_FIELD)
)
.filter(rowDataset.col(DebeziumConstants.INCOMING_OP_FIELD).equalTo(DebeziumConstants.DELETE_OP));
return insertedOrUpdatedData.union(deletedData);
} else {
return rowDataset;
}
}
}

View File

@@ -0,0 +1,209 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.utilities.sources.debezium;
import org.apache.hudi.common.config.TypedProperties;
import org.apache.hudi.common.model.debezium.DebeziumConstants;
import org.apache.hudi.common.util.Option;
import org.apache.hudi.utilities.UtilHelpers;
import org.apache.hudi.utilities.deltastreamer.HoodieDeltaStreamerMetrics;
import org.apache.hudi.utilities.deltastreamer.SourceFormatAdapter;
import org.apache.hudi.utilities.schema.SchemaProvider;
import org.apache.hudi.utilities.schema.SchemaRegistryProvider;
import org.apache.hudi.utilities.sources.InputBatch;
import org.apache.hudi.utilities.testutils.UtilitiesTestBase;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericRecord;
import org.apache.kafka.clients.consumer.ConsumerConfig;
import org.apache.kafka.common.serialization.StringDeserializer;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import org.apache.spark.streaming.kafka010.KafkaTestUtils;
import org.junit.jupiter.api.AfterAll;
import org.junit.jupiter.api.AfterEach;
import org.junit.jupiter.api.BeforeAll;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.params.ParameterizedTest;
import org.junit.jupiter.params.provider.Arguments;
import org.junit.jupiter.params.provider.MethodSource;
import java.io.IOException;
import java.util.UUID;
import java.util.stream.Stream;
import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertTrue;
import static org.junit.jupiter.params.provider.Arguments.arguments;
import static org.mockito.Mockito.mock;
public abstract class TestAbstractDebeziumSource extends UtilitiesTestBase {
private static final String TEST_TOPIC_NAME = "hoodie_test";
private final HoodieDeltaStreamerMetrics metrics = mock(HoodieDeltaStreamerMetrics.class);
private KafkaTestUtils testUtils;
@BeforeAll
public static void initClass() throws Exception {
UtilitiesTestBase.initClass(false);
}
@AfterAll
public static void cleanupClass() {
UtilitiesTestBase.cleanupClass();
}
@BeforeEach
public void setup() throws Exception {
super.setup();
testUtils = new KafkaTestUtils();
testUtils.setup();
}
@AfterEach
public void teardown() throws Exception {
super.teardown();
testUtils.teardown();
}
private TypedProperties createPropsForJsonSource() {
TypedProperties props = new TypedProperties();
props.setProperty("hoodie.deltastreamer.source.kafka.topic", TEST_TOPIC_NAME);
props.setProperty("bootstrap.servers", testUtils.brokerAddress());
props.setProperty("auto.offset.reset", "earliest");
props.setProperty(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, "false");
props.setProperty("hoodie.deltastreamer.schemaprovider.registry.url", "localhost");
props.setProperty("hoodie.deltastreamer.source.kafka.value.deserializer.class", StringDeserializer.class.getName());
props.setProperty(ConsumerConfig.GROUP_ID_CONFIG, UUID.randomUUID().toString());
return props;
}
protected abstract String getIndexName();
protected abstract String getSourceClass();
protected abstract String getSchema();
protected abstract GenericRecord generateMetaFields(GenericRecord record);
protected abstract void validateMetaFields(Dataset<Row> records);
@ParameterizedTest
@MethodSource("testArguments")
public void testDebeziumEvents(Operation operation) throws Exception {
String sourceClass = getSourceClass();
// topic setup.
testUtils.createTopic(TEST_TOPIC_NAME, 2);
TypedProperties props = createPropsForJsonSource();
SchemaProvider schemaProvider = new MockSchemaRegistryProvider(props, jsc, this);
SourceFormatAdapter debeziumSource = new SourceFormatAdapter(UtilHelpers.createSource(sourceClass, props, jsc, sparkSession, schemaProvider, metrics));
testUtils.sendMessages(TEST_TOPIC_NAME, new String[] {generateDebeziumEvent(operation).toString()});
InputBatch<Dataset<Row>> fetch = debeziumSource.fetchNewDataInRowFormat(Option.empty(), 10);
assertEquals(1, fetch.getBatch().get().count());
// Ensure the before fields are picked for DELETE CDC Events,
// and after fields are picked for INSERT and UPDATE CDC Events.
final String fieldPrefix = (operation.equals(Operation.DELETE)) ? "before_" : "after_";
assertTrue(fetch.getBatch().get().select("type").collectAsList().stream()
.allMatch(r -> r.getString(0).startsWith(fieldPrefix)));
assertTrue(fetch.getBatch().get().select("type").collectAsList().stream()
.allMatch(r -> r.getString(0).startsWith(fieldPrefix)));
// Validate DB specific meta fields
validateMetaFields(fetch.getBatch().get());
}
private GenericRecord generateDebeziumEvent(Operation op) {
Schema schema = new Schema.Parser().parse(getSchema());
String indexName = getIndexName().concat(".ghschema.gharchive.Value");
GenericRecord rec = new GenericData.Record(schema);
rec.put(DebeziumConstants.INCOMING_OP_FIELD, op.op);
rec.put(DebeziumConstants.INCOMING_TS_MS_FIELD, 100L);
// Before
Schema.Field beforeField = schema.getField(DebeziumConstants.INCOMING_BEFORE_FIELD);
Schema beforeSchema = beforeField.schema().getTypes().get(beforeField.schema().getIndexNamed(indexName));
GenericRecord beforeRecord = new GenericData.Record(beforeSchema);
beforeRecord.put("id", 1);
beforeRecord.put("date", "1/1/2020");
beforeRecord.put("type", "before_type");
beforeRecord.put("payload", "before_payload");
beforeRecord.put("timestamp", 1000L);
rec.put(DebeziumConstants.INCOMING_BEFORE_FIELD, beforeRecord);
// After
Schema.Field afterField = schema.getField(DebeziumConstants.INCOMING_AFTER_FIELD);
Schema afterSchema = afterField.schema().getTypes().get(afterField.schema().getIndexNamed(indexName));
GenericRecord afterRecord = new GenericData.Record(afterSchema);
afterRecord.put("id", 1);
afterRecord.put("date", "1/1/2021");
afterRecord.put("type", "after_type");
afterRecord.put("payload", "after_payload");
afterRecord.put("timestamp", 3000L);
rec.put(DebeziumConstants.INCOMING_AFTER_FIELD, afterRecord);
return generateMetaFields(rec);
}
private static class MockSchemaRegistryProvider extends SchemaRegistryProvider {
private final String schema;
public MockSchemaRegistryProvider(TypedProperties props, JavaSparkContext jssc, TestAbstractDebeziumSource source) {
super(props, jssc);
schema = source.getSchema();
}
@Override
public String fetchSchemaFromRegistry(String registryUrl) throws IOException {
return schema;
}
}
private static Stream<Arguments> testArguments() {
return Stream.of(
arguments(Operation.INSERT),
arguments(Operation.UPDATE),
arguments(Operation.DELETE)
);
}
private enum Operation {
INSERT("c"),
UPDATE("u"),
DELETE("d");
public final String op;
Operation(String op) {
this.op = op;
}
}
}

View File

@@ -0,0 +1,100 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.utilities.sources.debezium;
import org.apache.hudi.common.model.debezium.DebeziumConstants;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericRecord;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import static org.junit.jupiter.api.Assertions.assertTrue;
public class TestMysqlDebeziumSource extends TestAbstractDebeziumSource {
private static final String MYSQL_GITHUB_SCHEMA = "{\"connect.name\": \"mysql.ghschema.gharchive.Envelope\",\n"
+ " \"fields\": [{\"default\": null,\"name\": \"before\",\"type\": [\"null\",{\"connect.name\": \"mysql.ghschema.gharchive.Value\",\n"
+ " \"fields\": [{\"name\": \"id\",\"type\": \"string\"},{\"name\": \"date\",\"type\": \"string\"},{\"default\": null,\"name\": \"timestamp\",\n"
+ " \"type\": [\"null\",\"long\"]},{\"default\": null,\"name\": \"type\",\"type\": [\"null\",\"string\"]},{\"default\": null,\"name\": \"payload\",\n"
+ " \"type\": [\"null\",\"string\"]},{\"default\": null,\"name\": \"org\",\"type\": [\"null\",\"string\"]},{\"default\": null,\"name\": \"created_at\",\n"
+ " \"type\": [\"null\",\"long\"]},{\"default\": null,\"name\": \"public\",\"type\": [\"null\",\"boolean\"]}],\"name\": \"Value\",\"type\": \"record\"\n"
+ " }]},{\"default\": null,\"name\": \"after\",\"type\": [\"null\",\"Value\"]},{\"name\": \"source\",\"type\": {\"connect.name\": \"io.debezium.connector.mysql.Source\",\n"
+ " \"fields\": [{\"name\": \"connector\",\"type\": \"string\"},{\"name\": \"name\",\"type\": \"string\"},{\"name\": \"ts_ms\",\"type\": \"long\"},\n"
+ " {\"name\": \"db\",\"type\": \"string\"},{\"name\": \"table\",\"type\": \"string\"},{\"default\": null,\n"
+ " \"name\": \"txId\",\"type\": [\"null\",\"long\"]},{\"name\": \"file\",\"type\": \"string\"},{\"default\": null,\"name\": \"pos\",\"type\": [\"null\",\"long\"]},{\"default\": null,\n"
+ " \"name\": \"row\",\"type\": [\"null\",\"long\"]}],\"name\": \"Source\",\"namespace\": \"io.debezium.connector.mysql\",\"type\": \"record\"\n"
+ " }},{\"name\": \"op\",\"type\": \"string\"},{\"default\": null,\"name\": \"ts_ms\",\"type\": [\"null\",\"long\"]},{\"default\": null,\"name\": \"transaction\",\n"
+ " \"type\": [\"null\",{\"fields\": [{\"name\": \"id\",\"type\": \"string\"},{\"name\": \"total_order\",\"type\": \"long\"},{\"name\": \"data_collection_order\",\n"
+ " \"type\": \"long\"}],\"name\": \"ConnectDefault\",\"namespace\": \"io.confluent.connect.avro\",\"type\": \"record\"}]}],\"name\": \"Envelope\",\n"
+ " \"namespace\": \"mysql.ghschema.gharchive\",\"type\": \"record\"}";
private static final String TEST_DB = "ghschema";
private static final String TEST_TABLE = "gharchive";
private static final long TEST_TS_MS = 12345L;
private static final long TEST_TXID = 543L;
private static final String TEST_FILE = "mysql-bin.00007";
private static final long TEST_POS = 98765L;
private static final String EXPECTED_TEST_SEQ = "00007.98765";
@Override
protected String getIndexName() {
return "mysql";
}
@Override
protected String getSourceClass() {
return MysqlDebeziumSource.class.getName();
}
@Override
protected String getSchema() {
return MYSQL_GITHUB_SCHEMA;
}
@Override
protected GenericRecord generateMetaFields(GenericRecord rec) {
Schema schema = new Schema.Parser().parse(getSchema());
// Source fields specific to Mysql DB
GenericRecord sourceRecord = new GenericData.Record(schema.getField(DebeziumConstants.INCOMING_SOURCE_FIELD).schema());
sourceRecord.put("name", getIndexName());
sourceRecord.put("connector", getIndexName());
sourceRecord.put("db", TEST_DB);
sourceRecord.put("table", TEST_TABLE);
sourceRecord.put("ts_ms", TEST_TS_MS);
sourceRecord.put("txId", TEST_TXID);
sourceRecord.put("file", TEST_FILE);
sourceRecord.put("pos", TEST_POS);
rec.put(DebeziumConstants.INCOMING_SOURCE_FIELD, sourceRecord);
return rec;
}
@Override
protected void validateMetaFields(Dataset<Row> records) {
assertTrue(records.select(DebeziumConstants.FLATTENED_SHARD_NAME).collectAsList().stream()
.allMatch(r -> r.getString(0).equals(getIndexName())));
assertTrue(records.select(DebeziumConstants.FLATTENED_TS_COL_NAME).collectAsList().stream()
.allMatch(r -> r.getLong(0) == TEST_TS_MS));
assertTrue(records.select(DebeziumConstants.FLATTENED_TX_ID_COL_NAME).collectAsList().stream()
.allMatch(r -> r.getLong(0) == TEST_TXID));
assertTrue(records.select(DebeziumConstants.ADDED_SEQ_COL_NAME).collectAsList().stream()
.allMatch(r -> r.getString(0).equals(EXPECTED_TEST_SEQ)));
}
}

View File

@@ -0,0 +1,97 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hudi.utilities.sources.debezium;
import org.apache.hudi.common.model.debezium.DebeziumConstants;
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericRecord;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import static org.junit.jupiter.api.Assertions.assertTrue;
public class TestPostgresDebeziumSource extends TestAbstractDebeziumSource {
private static final String POSTGRES_GITHUB_SCHEMA = "{\"connect.name\": \"postgres.ghschema.gharchive.Envelope\",\n"
+ " \"fields\": [{\"default\": null,\"name\": \"before\",\"type\": [\"null\",{\"connect.name\": \"postgres.ghschema.gharchive.Value\",\n"
+ " \"fields\": [{\"name\": \"id\",\"type\": \"string\"},{\"name\": \"date\",\"type\": \"string\"},{\"default\": null,\"name\": \"timestamp\",\n"
+ " \"type\": [\"null\",\"long\"]},{\"default\": null,\"name\": \"type\",\"type\": [\"null\",\"string\"]},{\"default\": null,\"name\": \"payload\",\n"
+ " \"type\": [\"null\",\"string\"]},{\"default\": null,\"name\": \"org\",\"type\": [\"null\",\"string\"]},{\"default\": null,\"name\": \"created_at\",\n"
+ " \"type\": [\"null\",\"long\"]},{\"default\": null,\"name\": \"public\",\"type\": [\"null\",\"boolean\"]}],\"name\": \"Value\",\"type\": \"record\"\n"
+ " }]},{\"default\": null,\"name\": \"after\",\"type\": [\"null\",\"Value\"]},{\"name\": \"source\",\"type\": {\"connect.name\": \"io.debezium.connector.postgresql.Source\",\n"
+ " \"fields\": [{\"name\": \"connector\",\"type\": \"string\"},{\"name\": \"name\",\"type\": \"string\"},{\"name\": \"ts_ms\",\"type\": \"long\"},\n"
+ " {\"name\": \"db\",\"type\": \"string\"},{\"name\": \"schema\",\"type\": \"string\"},{\"name\": \"table\",\"type\": \"string\"},{\"default\": null,\n"
+ " \"name\": \"txId\",\"type\": [\"null\",\"long\"]},{\"default\": null,\"name\": \"lsn\",\"type\": [\"null\",\"long\"]},{\"default\": null,\n"
+ " \"name\": \"xmin\",\"type\": [\"null\",\"long\"]}],\"name\": \"Source\",\"namespace\": \"io.debezium.connector.postgresql\",\"type\": \"record\"\n"
+ " }},{\"name\": \"op\",\"type\": \"string\"},{\"default\": null,\"name\": \"ts_ms\",\"type\": [\"null\",\"long\"]},{\"default\": null,\"name\": \"transaction\",\n"
+ " \"type\": [\"null\",{\"fields\": [{\"name\": \"id\",\"type\": \"string\"},{\"name\": \"total_order\",\"type\": \"long\"},{\"name\": \"data_collection_order\",\n"
+ " \"type\": \"long\"}],\"name\": \"ConnectDefault\",\"namespace\": \"io.confluent.connect.avro\",\"type\": \"record\"}]}],\"name\": \"Envelope\",\n"
+ " \"namespace\": \"postgres.ghschema.gharchive\",\"type\": \"record\"}";
private static final String TEST_DB = "postgres";
private static final String TEST_SCHEMA = "ghschema";
private static final String TEST_TABLE = "gharchive";
private static final long TEST_TS_MS = 12345L;
private static final long TEST_TXID = 543L;
private static final long TEST_LSN = 98765L;
@Override
protected String getIndexName() {
return "postgres";
}
@Override
protected String getSourceClass() {
return PostgresDebeziumSource.class.getName();
}
@Override
protected String getSchema() {
return POSTGRES_GITHUB_SCHEMA;
}
@Override
protected GenericRecord generateMetaFields(GenericRecord rec) {
Schema schema = new Schema.Parser().parse(getSchema());
// Source fields specific to Postgres DB
GenericRecord sourceRecord = new GenericData.Record(schema.getField(DebeziumConstants.INCOMING_SOURCE_FIELD).schema());
sourceRecord.put("name", getIndexName());
sourceRecord.put("connector", getIndexName());
sourceRecord.put("db", TEST_DB);
sourceRecord.put("schema", TEST_SCHEMA);
sourceRecord.put("table", TEST_TABLE);
sourceRecord.put("ts_ms", TEST_TS_MS);
sourceRecord.put("txId", TEST_TXID);
sourceRecord.put("lsn", TEST_LSN);
rec.put(DebeziumConstants.INCOMING_SOURCE_FIELD, sourceRecord);
return rec;
}
@Override
protected void validateMetaFields(Dataset<Row> records) {
assertTrue(records.select(DebeziumConstants.FLATTENED_TS_COL_NAME).collectAsList().stream()
.allMatch(r -> r.getLong(0) == TEST_TS_MS));
assertTrue(records.select(DebeziumConstants.FLATTENED_TX_ID_COL_NAME).collectAsList().stream()
.allMatch(r -> r.getLong(0) == TEST_TXID));
assertTrue(records.select(DebeziumConstants.FLATTENED_LSN_COL_NAME).collectAsList().stream()
.allMatch(r -> r.getLong(0) == TEST_LSN));
}
}