danny0405 commented on a change in pull request #2430: URL: https://github.com/apache/hudi/pull/2430#discussion_r560731527
########## File path: hudi-flink/src/main/java/org/apache/hudi/operator/StreamWriteOperatorCoordinator.java ########## @@ -0,0 +1,419 @@ +/* + * 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.operator; + +import org.apache.hudi.client.FlinkTaskContextSupplier; +import org.apache.hudi.client.HoodieFlinkWriteClient; +import org.apache.hudi.client.WriteStatus; +import org.apache.hudi.client.common.HoodieFlinkEngineContext; +import org.apache.hudi.common.fs.FSUtils; +import org.apache.hudi.common.model.HoodieTableType; +import org.apache.hudi.common.table.HoodieTableMetaClient; +import org.apache.hudi.common.util.Option; +import org.apache.hudi.exception.HoodieException; +import org.apache.hudi.operator.event.BatchWriteSuccessEvent; +import org.apache.hudi.util.StreamerUtil; + +import org.apache.flink.annotation.VisibleForTesting; +import org.apache.flink.configuration.Configuration; +import org.apache.flink.core.memory.DataInputViewStreamWrapper; +import org.apache.flink.core.memory.DataOutputViewStreamWrapper; +import org.apache.flink.runtime.jobgraph.OperatorID; +import org.apache.flink.runtime.operators.coordination.OperatorCoordinator; +import org.apache.flink.runtime.operators.coordination.OperatorEvent; +import org.apache.flink.util.Preconditions; +import org.apache.hadoop.fs.FileSystem; +import org.apache.hadoop.fs.Path; +import org.jetbrains.annotations.Nullable; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.ByteArrayInputStream; +import java.io.ByteArrayOutputStream; +import java.io.DataInputStream; +import java.io.DataOutputStream; +import java.io.IOException; +import java.util.Arrays; +import java.util.Collection; +import java.util.HashMap; +import java.util.List; +import java.util.Objects; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.CompletionException; +import java.util.concurrent.TimeUnit; +import java.util.stream.Collectors; + +/** + * {@link OperatorCoordinator} for {@link StreamWriteFunction}. + * + * <p>This coordinator starts a new instant when a new checkpoint starts. It commits the instant when all the + * operator tasks write the buffer successfully for a round of checkpoint. + * + * <p>If there is no data for a round of checkpointing, it rolls back the metadata. + * + * @see StreamWriteFunction for the work flow and semantics + */ +public class StreamWriteOperatorCoordinator + implements OperatorCoordinator { + private static final Logger LOG = LoggerFactory.getLogger(StreamWriteOperatorCoordinator.class); + + /** + * Config options. + */ + private final Configuration conf; + + /** + * Write client. + */ + private transient HoodieFlinkWriteClient writeClient; + + private long inFlightCheckpoint = -1; + + /** + * Current REQUESTED instant, for validation. + */ + private String inFlightInstant = ""; + + /** + * Event buffer for one round of checkpointing. When all the elements are non-null and have the same + * write instant, then the instant succeed and we can commit it. + */ + private transient BatchWriteSuccessEvent[] eventBuffer; + + /** + * Task number of the operator. + */ + private final int parallelism; + + /** + * Constructs a StreamingSinkOperatorCoordinator. + * + * @param conf The config options + * @param parallelism The operator task number + */ + public StreamWriteOperatorCoordinator( + Configuration conf, + int parallelism) { + this.conf = conf; + this.parallelism = parallelism; + } + + @Override + public void start() throws Exception { + // initialize event buffer + reset(); + // writeClient + initWriteClient(); + // init table, create it if not exists. + initTable(); + } + + @Override + public void close() { + if (writeClient != null) { + writeClient.close(); + } + this.eventBuffer = null; + } + + @Override + public void checkpointCoordinator(long checkpointId, CompletableFuture<byte[]> result) { + try { + final String errMsg = "A new checkpoint starts while the last checkpoint buffer" + + " data has not finish writing, roll back the last write and throw"; + checkAndForceCommit(errMsg); + this.inFlightInstant = this.writeClient.startCommit(); + this.inFlightCheckpoint = checkpointId; + LOG.info("Create instant [{}], at checkpoint [{}]", this.inFlightInstant, checkpointId); + result.complete(writeCheckpointBytes()); + } catch (Throwable throwable) { + // when a checkpoint fails, throws directly. + result.completeExceptionally( + new CompletionException( + String.format("Failed to checkpoint Instant %s for source %s", + this.inFlightInstant, this.getClass().getSimpleName()), throwable)); + } + } + + @Override + public void notifyCheckpointComplete(long checkpointId) { + // start to commit the instant. + checkAndCommitWithRetry(); + } + + @Override + public void notifyCheckpointAborted(long checkpointId) { + Preconditions.checkState(inFlightCheckpoint == checkpointId, + "The aborted checkpoint should always be the last checkpoint"); + checkAndForceCommit("The last checkpoint was aborted, roll back the last write and throw"); + } + + @Override + public void resetToCheckpoint(long l, @Nullable byte[] checkpointData) throws Exception { + if (checkpointData != null) { + // restore when any checkpoint completed + deserializeCheckpointAndRestore(checkpointData); + } + } + + @Override + public void handleEventFromOperator(int i, OperatorEvent operatorEvent) { + // no event to handle + Preconditions.checkState(operatorEvent instanceof BatchWriteSuccessEvent, + "The coordinator can only handle BatchWriteSuccessEvent"); + BatchWriteSuccessEvent event = (BatchWriteSuccessEvent) operatorEvent; + Preconditions.checkState(event.getInstantTime().equals(this.inFlightInstant), + String.format("Receive an unexpected event for instant %s from task %d", + event.getInstantTime(), event.getTaskID())); + this.eventBuffer[event.getTaskID()] = event; + } + + @Override + public void subtaskFailed(int i, @Nullable Throwable throwable) { + // no operation + } + + @Override + public void subtaskReset(int i, long l) { + // no operation + } + + // ------------------------------------------------------------------------- + // Utilities + // ------------------------------------------------------------------------- + + @SuppressWarnings("rawtypes") + private void initWriteClient() { + writeClient = new HoodieFlinkWriteClient( + new HoodieFlinkEngineContext(new FlinkTaskContextSupplier(null)), + StreamerUtil.getHoodieClientConfig(this.conf), + true); + } + + private void initTable() throws IOException { + final String basePath = this.conf.getString(FlinkOptions.PATH); + final org.apache.hadoop.conf.Configuration hadoopConf = StreamerUtil.getHadoopConf(); + // Hadoop FileSystem + try (FileSystem fs = FSUtils.getFs(basePath, hadoopConf)) { + if (!fs.exists(new Path(basePath, HoodieTableMetaClient.METAFOLDER_NAME))) { + HoodieTableMetaClient.initTableType( + hadoopConf, + basePath, + HoodieTableType.valueOf(this.conf.getString(FlinkOptions.TABLE_TYPE)), + this.conf.getString(FlinkOptions.TABLE_NAME), + "archived", + this.conf.getString(FlinkOptions.PAYLOAD_CLASS), + 1); + LOG.info("Table initialized"); + } else { + LOG.info("Table [{}/{}] already exists, no need to initialize the table", + basePath, this.conf.getString(FlinkOptions.TABLE_NAME)); + } + } + } + + static byte[] readBytes(DataInputStream in, int size) throws IOException { + byte[] bytes = new byte[size]; + in.readFully(bytes); + return bytes; + } + + /** + * Serialize the coordinator state. The current implementation may not be super efficient, + * but it should not matter that much because most of the state should be rather small. + * Large states themselves may already be a problem regardless of how the serialization + * is implemented. + * + * @return A byte array containing the serialized state of the source coordinator. + * @throws IOException When something goes wrong in serialization. + */ + private byte[] writeCheckpointBytes() throws IOException { + try (ByteArrayOutputStream baos = new ByteArrayOutputStream(); + DataOutputStream out = new DataOutputViewStreamWrapper(baos)) { + + out.writeLong(this.inFlightCheckpoint); + byte[] serializedInstant = this.inFlightInstant.getBytes(); + out.writeInt(serializedInstant.length); + out.write(serializedInstant); + out.flush(); + return baos.toByteArray(); + } + } + + /** + * Restore the state of this source coordinator from the state bytes. + * + * @param bytes The checkpoint bytes that was returned from {@link #writeCheckpointBytes()} + * @throws Exception When the deserialization failed. + */ + private void deserializeCheckpointAndRestore(byte[] bytes) throws Exception { + try (ByteArrayInputStream bais = new ByteArrayInputStream(bytes); + DataInputStream in = new DataInputViewStreamWrapper(bais)) { + long checkpointID = in.readLong(); + int serializedInstantSize = in.readInt(); + byte[] serializedInstant = readBytes(in, serializedInstantSize); + this.inFlightCheckpoint = checkpointID; + this.inFlightInstant = new String(serializedInstant); + } + } + + private void reset() { + this.inFlightInstant = ""; + this.eventBuffer = new BatchWriteSuccessEvent[this.parallelism]; + } + + private void checkAndForceCommit(String errMsg) { + if (!checkReady()) { + // forced but still has inflight instant + String inflightInstant = writeClient.getInflightAndRequestedInstant(this.conf.getString(FlinkOptions.TABLE_TYPE)); + if (inflightInstant != null) { + assert inflightInstant.equals(this.inFlightInstant); + writeClient.rollback(this.inFlightInstant); + throw new HoodieException(errMsg); + } + if (Arrays.stream(eventBuffer).allMatch(Objects::isNull)) { + // The last checkpoint finished successfully. + return; + } + } + doCommit(); + } + + private void checkAndCommitWithRetry() { + int retryTimes = this.conf.getInteger(FlinkOptions.RETRY_TIMES); + long retryIntervalMillis = this.conf.getLong(FlinkOptions.RETRY_INTERVAL_MS); + int tryTimes = 0; + while (tryTimes++ < retryTimes) { + try { + if (!checkReady()) { + // Do not throw if the try times expires but the event buffer are still not ready, + // because we have a force check when next checkpoint starts. + waitFor(retryIntervalMillis); + continue; + } + doCommit(); + return; + } catch (Throwable throwable) { + String cause = throwable.getCause() == null ? "" : throwable.getCause().toString(); + LOG.warn("Try to commit the instant {} failed, with times {} and cause {}", this.inFlightInstant, tryTimes, cause); + if (tryTimes == retryTimes) { + throw new HoodieException(throwable); + } + waitFor(retryIntervalMillis); + } + } + } + + private void waitFor(long intervalMillis) { + try { + TimeUnit.MILLISECONDS.sleep(intervalMillis); + } catch (InterruptedException e) { + LOG.error("Thread interrupted while waiting to retry the instant commits"); + throw new HoodieException(e); + } + } + + /** Checks the buffer is ready to commit. */ + private boolean checkReady() { + return Arrays.stream(eventBuffer).allMatch(event -> + event != null && event.getInstantTime().equals(this.inFlightInstant)); + } + + /** Performs the actual commit action. */ + private void doCommit() { + List<WriteStatus> writeResults = Arrays.stream(eventBuffer) + .map(BatchWriteSuccessEvent::getWriteStatuses) + .flatMap(Collection::stream) + .collect(Collectors.toList()); + + if (writeResults.size() == 0) { + // No data has written, clear the metadata file + this.writeClient.deletePendingInstant(this.conf.getString(FlinkOptions.TABLE_TYPE), this.inFlightInstant); + reset(); + return; + } + + // commit or rollback + long totalErrorRecords = writeResults.stream().map(WriteStatus::getTotalErrorRecords).reduce(Long::sum).orElse(0L); + long totalRecords = writeResults.stream().map(WriteStatus::getTotalRecords).reduce(Long::sum).orElse(0L); + boolean hasErrors = totalErrorRecords > 0; + + if (!hasErrors || this.conf.getBoolean(FlinkOptions.IGNORE_FAILED_BATCH)) { + HashMap<String, String> checkpointCommitMetadata = new HashMap<>(); + if (hasErrors) { + LOG.warn("Some records failed to merge but forcing commit since commitOnErrors set to true. Errors/Total=" + + totalErrorRecords + "/" + totalRecords); + } + + boolean success = writeClient.commit(this.inFlightInstant, writeResults, Option.of(checkpointCommitMetadata)); + if (success) { + reset(); + LOG.info("Commit instant [{}] success!", this.inFlightInstant); + } else { + throw new HoodieException(String.format("Commit instant [%s] failed!", this.inFlightInstant)); + } + } else { + LOG.error("Error when writing. Errors/Total=" + totalErrorRecords + "/" + totalRecords); + LOG.error("The first 100 error messages"); + writeResults.stream().filter(WriteStatus::hasErrors).limit(100).forEach(ws -> { + LOG.error("Global error for partition path {} and fileID {}: {}", + ws.getGlobalError(), ws.getPartitionPath(), ws.getFileId()); + if (ws.getErrors().size() > 0) { + ws.getErrors().forEach((key, value) -> LOG.trace("Error for key:" + key + " and value " + value)); + } + }); + // Rolls back instant + writeClient.rollback(this.inFlightInstant); + throw new HoodieException(String.format("Commit instant [%s] failed and rolled back !", this.inFlightInstant)); + } + } + + @VisibleForTesting + public BatchWriteSuccessEvent[] getEventBuffer() { + return eventBuffer; + } + + @VisibleForTesting + public String getInFlightInstant() { Review comment: > How did the writer know the current instant Fetch through the `HoodieFlinkWriteClient`, actually the `HoodieTableMetaClient`. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
