From 66d0f0b6515c3352b2aaecc1005bf86d4baa180c Mon Sep 17 00:00:00 2001 From: wudi Date: Fri, 22 May 2026 10:10:14 +0800 Subject: [PATCH] [fix](streaming-job) Avoid NPE on cross-table DML during snapshot chunk read (#63435) ### What problem does this PR solve? During the snapshot phase, the chunk reader (`IncrementalSourceScanFetcher`) consumes from a change-event queue that may also contain DML records from **other tables** being captured concurrently. When such a foreign-table record reached `isChangeRecordInChunkRange`, the code compared it against the **current chunk's** PK range via `isRecordBetween(...)`. Two problems: 1. The foreign table's schema may not yet be loaded for this fetcher, so extracting its PK throws NPE. 2. Even if the schema were loaded, the foreign table's PK columns do not necessarily align with this chunk's bounds, so any range comparison is meaningless and the record would be incorrectly merged into the wrong chunk's output buffer. This patch adds an explicit `TableId` check at the very start of `isChangeRecordInChunkRange`: records whose `TableId` does not match `currentSnapshotSplit.getTableId()` are skipped before any PK-based comparison runs. --- .../IncrementalSourceScanFetcher.java | 284 ++++++++++++++++++ 1 file changed, 284 insertions(+) create mode 100644 fs_brokers/cdc_client/src/main/java/org/apache/flink/cdc/connectors/base/source/reader/external/IncrementalSourceScanFetcher.java diff --git a/fs_brokers/cdc_client/src/main/java/org/apache/flink/cdc/connectors/base/source/reader/external/IncrementalSourceScanFetcher.java b/fs_brokers/cdc_client/src/main/java/org/apache/flink/cdc/connectors/base/source/reader/external/IncrementalSourceScanFetcher.java new file mode 100644 index 00000000000000..a2c51e4e0cf2e8 --- /dev/null +++ b/fs_brokers/cdc_client/src/main/java/org/apache/flink/cdc/connectors/base/source/reader/external/IncrementalSourceScanFetcher.java @@ -0,0 +1,284 @@ +/* + * 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.flink.cdc.connectors.base.source.reader.external; + +import org.apache.flink.cdc.common.annotation.VisibleForTesting; +import org.apache.flink.cdc.connectors.base.source.meta.split.SnapshotSplit; +import org.apache.flink.cdc.connectors.base.source.meta.split.SourceRecords; +import org.apache.flink.cdc.connectors.base.source.meta.split.SourceSplitBase; +import org.apache.flink.util.FlinkRuntimeException; + +import org.apache.flink.shaded.guava31.com.google.common.util.concurrent.ThreadFactoryBuilder; + +import io.debezium.connector.base.ChangeEventQueue; +import io.debezium.pipeline.DataChangeEvent; +import io.debezium.relational.TableId; +import org.apache.kafka.connect.data.Struct; +import org.apache.kafka.connect.source.SourceRecord; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import javax.annotation.Nullable; + +import java.util.ArrayList; +import java.util.Collections; +import java.util.HashMap; +import java.util.Iterator; +import java.util.List; +import java.util.Map; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; +import java.util.concurrent.ThreadFactory; +import java.util.concurrent.TimeUnit; +import java.util.concurrent.atomic.AtomicBoolean; + +import static org.apache.flink.cdc.connectors.base.source.meta.wartermark.WatermarkEvent.isEndWatermarkEvent; +import static org.apache.flink.cdc.connectors.base.source.meta.wartermark.WatermarkEvent.isHighWatermarkEvent; +import static org.apache.flink.cdc.connectors.base.source.meta.wartermark.WatermarkEvent.isLowWatermarkEvent; +import static org.apache.flink.util.Preconditions.checkState; + +/** + * Copy from https://github.com/apache/flink-cdc/blob/release-3.6.0/flink-cdc-connect/flink-cdc-source-connectors/flink-cdc-base/src/main/java/org/apache/flink/cdc/connectors/base/source/reader/external/IncrementalSourceScanFetcher.java + * Fix with FLINK-39633, change to method isChangeRecordInChunkRange + */ +public class IncrementalSourceScanFetcher implements Fetcher { + + private static final Logger LOG = LoggerFactory.getLogger(IncrementalSourceScanFetcher.class); + + public AtomicBoolean hasNextElement; + public AtomicBoolean reachEnd; + + private final FetchTask.Context taskContext; + private final ExecutorService executorService; + private volatile ChangeEventQueue queue; + private volatile Throwable readException; + + // task to read snapshot for current split + private FetchTask snapshotSplitReadTask; + private SnapshotSplit currentSnapshotSplit; + + private static final long READER_CLOSE_TIMEOUT_SECONDS = 30L; + + public IncrementalSourceScanFetcher(FetchTask.Context taskContext, int subtaskId) { + this.taskContext = taskContext; + ThreadFactory threadFactory = + new ThreadFactoryBuilder() + .setNameFormat("debezium-snapshot-reader-" + subtaskId) + .setUncaughtExceptionHandler( + (thread, throwable) -> setReadException(throwable)) + .build(); + this.executorService = Executors.newSingleThreadExecutor(threadFactory); + this.hasNextElement = new AtomicBoolean(false); + this.reachEnd = new AtomicBoolean(false); + } + + @Override + public void submitTask(FetchTask fetchTask) { + this.snapshotSplitReadTask = fetchTask; + this.currentSnapshotSplit = fetchTask.getSplit().asSnapshotSplit(); + taskContext.configure(currentSnapshotSplit); + this.queue = taskContext.getQueue(); + this.hasNextElement.set(true); + this.reachEnd.set(false); + + executorService.execute( + () -> { + try { + snapshotSplitReadTask.execute(taskContext); + } catch (Exception e) { + setReadException(e); + } + }); + } + + @Override + public boolean isFinished() { + return currentSnapshotSplit == null + || (!snapshotSplitReadTask.isRunning() && !hasNextElement.get() && reachEnd.get()); + } + + @Nullable + @Override + public Iterator pollSplitRecords() throws InterruptedException { + checkReadException(); + + if (hasNextElement.get()) { + if (taskContext.getSourceConfig().isSkipSnapshotBackfill()) { + return pollWithoutBuffer(); + } else { + return pollWithBuffer(); + } + } + // the data has been polled, no more data + reachEnd.compareAndSet(false, true); + return null; + } + + public Iterator pollWithoutBuffer() throws InterruptedException { + checkReadException(); + List batch = queue.poll(); + final List records = new ArrayList<>(); + for (DataChangeEvent event : batch) { + if (isEndWatermarkEvent(event.getRecord())) { + hasNextElement.set(false); + break; + } + records.add(event.getRecord()); + } + + return Collections.singletonList(new SourceRecords(records)).iterator(); + } + + public Iterator pollWithBuffer() throws InterruptedException { + // eg: + // data input: [low watermark event][snapshot events][high watermark event][change + // events][end watermark event] + // data output: [low watermark event][normalized events][high watermark event] + boolean reachChangeLogStart = false; + boolean reachChangeLogEnd = false; + SourceRecord lowWatermark = null; + SourceRecord highWatermark = null; + Map outputBuffer = new HashMap<>(); + while (!reachChangeLogEnd) { + checkReadException(); + List batch = queue.poll(); + for (DataChangeEvent event : batch) { + SourceRecord record = event.getRecord(); + if (lowWatermark == null) { + lowWatermark = record; + assertLowWatermark(lowWatermark); + continue; + } + + if (highWatermark == null && isHighWatermarkEvent(record)) { + highWatermark = record; + // snapshot events capture end and begin to capture stream events + reachChangeLogStart = true; + continue; + } + + if (reachChangeLogStart && isEndWatermarkEvent(record)) { + // capture to end watermark events, stop the loop + reachChangeLogEnd = true; + break; + } + + if (!reachChangeLogStart) { + outputBuffer.put((Struct) record.key(), record); + } else { + if (isChangeRecordInChunkRange(record)) { + // rewrite overlapping snapshot records through the record key + taskContext.rewriteOutputBuffer(outputBuffer, record); + } + } + } + } + // snapshot split return its data once + hasNextElement.set(false); + + final List normalizedRecords = new ArrayList<>(); + normalizedRecords.add(lowWatermark); + normalizedRecords.addAll(taskContext.formatMessageTimestamp(outputBuffer.values())); + normalizedRecords.add(highWatermark); + + final List sourceRecordsSet = new ArrayList<>(); + sourceRecordsSet.add(new SourceRecords(normalizedRecords)); + return sourceRecordsSet.iterator(); + } + + private void checkReadException() { + if (readException != null) { + throw new FlinkRuntimeException( + String.format( + "Read split %s error due to %s.", + currentSnapshotSplit, readException.getMessage()), + readException); + } + } + + private void setReadException(Throwable throwable) { + LOG.error( + String.format( + "Execute snapshot read task for snapshot split %s fail", + currentSnapshotSplit), + throwable); + if (readException == null) { + readException = throwable; + } else { + readException.addSuppressed(throwable); + } + } + + @Override + public void close() { + try { + if (taskContext != null) { + taskContext.close(); + } + + if (snapshotSplitReadTask != null) { + snapshotSplitReadTask.close(); + } + + if (executorService != null) { + executorService.shutdown(); + if (!executorService.awaitTermination( + READER_CLOSE_TIMEOUT_SECONDS, TimeUnit.SECONDS)) { + LOG.warn( + "Failed to close the scan fetcher in {} seconds.", + READER_CLOSE_TIMEOUT_SECONDS); + } + } + } catch (Exception e) { + LOG.error("Close scan fetcher error", e); + } + } + + @VisibleForTesting + public ExecutorService getExecutorService() { + return executorService; + } + + private void assertLowWatermark(SourceRecord lowWatermark) { + checkState( + isLowWatermarkEvent(lowWatermark), + String.format( + "The first record should be low watermark signal event, but actual is %s", + lowWatermark)); + } + + @VisibleForTesting + boolean isChangeRecordInChunkRange(SourceRecord record) { + if (!taskContext.isDataChangeRecord(record)) { + return false; + } + // Skip records of other captured tables; their schema may not be loaded yet + // and their PKs do not align with this chunk's bounds. + TableId recordTableId = taskContext.getTableId(record); + if (recordTableId == null || !recordTableId.equals(currentSnapshotSplit.getTableId())) { + return false; + } + return taskContext.isRecordBetween( + record, currentSnapshotSplit.getSplitStart(), currentSnapshotSplit.getSplitEnd()); + } + + @VisibleForTesting + void setCurrentSnapshotSplit(SnapshotSplit currentSnapshotSplit) { + this.currentSnapshotSplit = currentSnapshotSplit; + } +} \ No newline at end of file