Yanikovic commented on a change in pull request #15140: URL: https://github.com/apache/flink/pull/15140#discussion_r593744502
########## File path: flink-connectors/flink-connector-rabbitmq2/src/main/java/org/apache/flink/connector/rabbitmq2/sink/writer/specalized/RabbitMQSinkWriterAtLeastOnce.java ########## @@ -0,0 +1,192 @@ +/* + * 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.connector.rabbitmq2.sink.writer.specalized; + +import org.apache.flink.api.common.serialization.SerializationSchema; +import org.apache.flink.api.connector.sink.SinkWriter; +import org.apache.flink.connector.rabbitmq2.RabbitMQConnectionConfig; +import org.apache.flink.connector.rabbitmq2.sink.RabbitMQSink; +import org.apache.flink.connector.rabbitmq2.sink.RabbitMQSinkPublishOptions; +import org.apache.flink.connector.rabbitmq2.sink.SerializableReturnListener; +import org.apache.flink.connector.rabbitmq2.sink.SinkMessage; +import org.apache.flink.connector.rabbitmq2.sink.state.RabbitMQSinkWriterState; +import org.apache.flink.connector.rabbitmq2.sink.writer.RabbitMQSinkWriterBase; + +import com.rabbitmq.client.Channel; +import com.rabbitmq.client.ConfirmCallback; +import com.rabbitmq.client.Connection; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Collections; +import java.util.HashSet; +import java.util.List; +import java.util.Set; +import java.util.concurrent.ConcurrentNavigableMap; +import java.util.concurrent.ConcurrentSkipListMap; + +/** + * A {@link SinkWriter} implementation for {@link RabbitMQSink} that has at-least-once semantics, + * meaning it guarantees that outgoing message arrive at RabbitMQ at least once. + * + * <p>At-least-once behaviour is implemented by assigning sequence numbers to arriving messages and + * buffering them together in the state of the writer until an ack arrives or a maximum number of + * retries is exceeded. + * + * <p>Checkpointing is required for at-least-once to work because messages are resend only when a + * checkpoint is triggered (to avoid complex time tracking mechanisms for each individual message). + * On a checkpoint, a resend of all messages that are unacknowledged and at least {@code + * resendIntervalMilliseconds} old is issued. + * + * <p>After a failure, a new writer gets initialized with one or more states that contain + * unacknowledged messages. These messages get resend immediately while buffering them in the new + * state of the writer. + * + * @param <T> Type of the elements in this sink + */ +public class RabbitMQSinkWriterAtLeastOnce<T> extends RabbitMQSinkWriterBase<T> { + protected final ConcurrentNavigableMap<Long, SinkMessage<T>> outstandingConfirms; + private Set<Long> lastSeenMessageIds; + private long lastResendTimestampMilliseconds; + private final long resendIntervalMilliseconds; + + public static final long DEFAULT_MINIMAL_RESEND_INTERVAL = 5000L; + + /** + * Create a new RabbitMQSinkWriterExactlyOnce. + * + * @param connectionConfig configuration parameters used to connect to RabbitMQ + * @param queueName name of the queue to publish to + * @param serializationSchema serialization schema to turn elements into byte representation + * @param publishOptions optionally used to compute routing/exchange for messages + * @param maxRetry number of retries for each message + * @param returnListener returnListener + * @param minimalResendIntervalMilliseconds how long to wait until a resend is triggered + * @param states a list of states to initialize this reader with + */ + public RabbitMQSinkWriterAtLeastOnce( + RabbitMQConnectionConfig connectionConfig, + String queueName, + SerializationSchema<T> serializationSchema, + RabbitMQSinkPublishOptions<T> publishOptions, + int maxRetry, + SerializableReturnListener returnListener, + Long minimalResendIntervalMilliseconds, + List<RabbitMQSinkWriterState<T>> states) { + super( + connectionConfig, + queueName, + serializationSchema, + publishOptions, + maxRetry, + returnListener); + this.outstandingConfirms = new ConcurrentSkipListMap<>(); + this.lastSeenMessageIds = new HashSet<>(); + this.lastResendTimestampMilliseconds = System.currentTimeMillis(); + this.resendIntervalMilliseconds = + minimalResendIntervalMilliseconds != null + ? minimalResendIntervalMilliseconds + : DEFAULT_MINIMAL_RESEND_INTERVAL; + initWithState(states); + } + + private void initWithState(List<RabbitMQSinkWriterState<T>> states) { + for (RabbitMQSinkWriterState<T> state : states) { + for (SinkMessage<T> message : state.getOutstandingMessages()) { + send(message); + } + } + } + + @Override + protected void send(SinkMessage<T> msg) { + long sequenceNumber = rmqChannel.getNextPublishSeqNo(); + super.send(msg); + outstandingConfirms.put(sequenceNumber, msg); + } + + private void resendMessages() { + Set<Long> temp = outstandingConfirms.keySet(); + Set<Long> messagesToResend = new HashSet<>(temp); + messagesToResend.retainAll(lastSeenMessageIds); + for (Long id : messagesToResend) { + // remove the old message from the map, since the message was added a second time + // under a new id or is put into the list of messages to resend + SinkMessage<T> msg = outstandingConfirms.remove(id); + if (msg != null) { + send(msg); + } + } + lastSeenMessageIds = temp; + } + + private ConfirmCallback handleAcknowledgements() { + return (sequenceNumber, multiple) -> { + // multiple flag indicates that all messages < sequenceNumber can be safely acknowledged + if (multiple) { + // create a view of the portion of the map that contains keys < sequenceNumber + ConcurrentNavigableMap<Long, SinkMessage<T>> confirmed = + outstandingConfirms.headMap(sequenceNumber, true); + // changes to the view are reflected in the original map + confirmed.clear(); + } else { + outstandingConfirms.remove(sequenceNumber); + } + }; + } + + private ConfirmCallback handleNegativeAcknowledgements() { + return (sequenceNumber, multiple) -> { + SinkMessage<T> message = outstandingConfirms.get(sequenceNumber); + LOG.error( + "Message with body {} has been nack-ed. Sequence number: {}, multiple: {}", + message.getMessage(), + sequenceNumber, + multiple); + }; Review comment: From RabbitMQ docs: _"Reject one or more incoming messages. This method allows a client to reject one or more incoming messages. It can be used to interrupt and cancel large incoming messages, or return untreatable messages to their original queue. This method is also used by the server to inform publishers on channels in confirm mode of unhandled messages. If a publisher receives this method, it probably needs to republish the offending messages."_ The basic.nack method indicates the message somehow cannot be processed by the RabbitMQ server. IMO, for at-most-once it should be okay to keep the job running if individual messages do not arrive, but for at-least and exactly-once it should probably fail. ---------------------------------------------------------------- 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: us...@infra.apache.org
