ApsaraMQ for RocketMQ:Send and receive ordered messages

Ordem de produção

O broker define a ordem das mensagens com base na sequência em que um único produtor ou thread as envia. Para preservar a ordenação:

• Envie mensagens relacionadas a partir de um único produtor em uma única thread.

• Caso múltiplos produtores ou threads enviem mensagens simultaneamente, o broker as ordenará pelo horário de chegada, o que pode divergir da sequência de negócios pretendida.

Ordem de consumo

O consumidor busca mensagens ordenadas por partição em uma ou mais partições e processa as mensagens de cada partição na ordem de envio. A garantia de ordenação depende do reconhecimento adequado:

• O consumidor precisa confirmar todas as mensagens de um lote antes de buscar o próximo lote na mesma partição.

• Se o broker não receber um reconhecimento (ACK) de uma mensagem antes do tempo limite, ele reenviará essa mensagem.

• Cada reentrega gera um novo identificador de recibo (receipt handle) com um timestamp exclusivo.

//#include <iostream>
#include <fstream>
#include <time.h>
#include "mq_http_sdk/mq_client.h"

using namespace std;
using namespace mq::http::sdk;

int main() {

    MQClient mqClient(
            // The HTTP endpoint. You can obtain the endpoint in the HTTP Endpoint section of the Instance Details page in the ApsaraMQ for RocketMQ console.
            "${HTTP_ENDPOINT}",
            // Make sure that the environment variables ALIBABA_CLOUD_ACCESS_KEY_ID and ALIBABA_CLOUD_ACCESS_KEY_SECRET are configured.
	          // The AccessKey ID that is used for authentication.
	          System.getenv("ALIBABA_CLOUD_ACCESS_KEY_ID"),
	          // The AccessKey secret that is used for authentication.
	          System.getenv("ALIBABA_CLOUD_ACCESS_KEY_SECRET")
            );

    // The topic in which the message is produced. You must create the topic in the ApsaraMQ for RocketMQ console.
    string topic = "${TOPIC}";
    // The ID of the instance to which the topic belongs. You must create the instance in the ApsaraMQ for RocketMQ console.
    // If the instance has a namespace, specify the ID of the instance. If the instance does not have a namespace, set the instanceID parameter to null or an empty string. You can obtain the namespace of the instance on the Instance Details page in the ApsaraMQ for RocketMQ console.
    string instanceId = "${INSTANCE_ID}";

    MQProducerPtr producer;
    if (instanceId == "") {
        producer = mqClient.getProducerRef(topic);
    } else {
        producer = mqClient.getProducerRef(instanceId, topic);
    }

    try {
        // Cyclically send four messages.
        for (int i = 0; i < 8; i++)
        {
            PublishMessageResponse pmResp;
            // The message content.
            TopicMessage pubMsg("Hello, mq!order msg!");
            // The sharding key that is used to distribute ordered messages to a specific partition. Sharding keys can be used to identify partitions. A sharding key is different from a message key.
            pubMsg.setShardingKey(std::to_string(i % 2));
            // The custom attributes of the message.
            pubMsg.putProperty("a",std::to_string(i));
            producer->publishMessage(pubMsg, pmResp);
            cout << "Publish mq message success. Topic is: " << topic
                << ", msgId is:" << pmResp.getMessageId()
                << ", bodyMD5 is:" << pmResp.getMessageBodyMD5() << endl;
        }
    } catch (MQServerException& me) {
        cout << "Request Failed: " + me.GetErrorCode() << ", requestId is:" << me.GetRequestId() << endl;
        return -1;
    } catch (MQExceptionBase& mb) {
        cout << "Request Failed: " + mb.ToString() << endl;
        return -2;
    }

    return 0;
}

#include <vector>
#include <fstream>
#include "mq_http_sdk/mq_client.h"

#ifdef _WIN32
#include <windows.h>
#else
#include <unistd.h>
#endif

using namespace std;
using namespace mq::http::sdk;

int main() {

    MQClient mqClient(
            // The HTTP endpoint. You can obtain the endpoint in the HTTP Endpoint section of the Instance Details page in the ApsaraMQ for RocketMQ console.
            "${HTTP_ENDPOINT}",
            // Make sure that the environment variables ALIBABA_CLOUD_ACCESS_KEY_ID and ALIBABA_CLOUD_ACCESS_KEY_SECRET are configured.
	          // The AccessKey ID that is used for authentication.
	          System.getenv("ALIBABA_CLOUD_ACCESS_KEY_ID"),
	          // The AccessKey secret that is used for authentication.
	          System.getenv("ALIBABA_CLOUD_ACCESS_KEY_SECRET")
            );

    // The topic in which the message is produced. You must create the topic in the ApsaraMQ for RocketMQ console.
    string topic = "${TOPIC}";
    // The ID of the consumer group that you created in the ApsaraMQ for RocketMQ console.
    string groupId = "${GROUP_ID}";
    // The ID of the instance to which the topic belongs. You must create the instance in the ApsaraMQ for RocketMQ console.
    // If the instance has a namespace, specify the ID of the instance. If the instance does not have a namespace, set the instanceID parameter to null or an empty string. You can obtain the namespace of the instance on the Instance Details page in the ApsaraMQ for RocketMQ console.
    string instanceId = "${INSTANCE_ID}";

    MQConsumerPtr consumer;
    if (instanceId == "") {
        consumer = mqClient.getConsumerRef(topic, groupId);
    } else {
        consumer = mqClient.getConsumerRef(instanceId, topic, groupId, "");
    }

    do {
        try {
            std::vector<Message> messages;
            // Consume messages in long polling mode. The consumer may pull partitionally ordered messages from multiple partitions. The consumer consumes messages from the same partition in the order in which the messages are sent.
            // Assume that a consumer pulls partitionally ordered messages from one partition. If the broker fails to receive an acknowledgement (ACK) for a message from the consumer, the broker delivers the message in the partition to the consumer again.
            // The consumer can consume the next batch of messages from a partition only after all messages that are pulled from the partition in the previous batch are acknowledged as consumed.
            // In long polling mode, if no message in the topic is available for consumption, the request is suspended on the broker for the specified period of time. If a message becomes available for consumption within the specified period of time, the broker immediately sends a response to the consumer. In this example, the value is specified as 3 seconds.
            consumer->consumeMessageOrderly(
                    3, // The maximum number of messages that can be consumed at a time. In this example, the value is specified as 3. The maximum value that you can specify is 16.
                    3, // The duration of a long polling cycle. Unit: seconds. In this example, the value is specified as 3. The maximum value that you can specify is 30.
                    messages
            );
            cout << "Consume: " << messages.size() << " Messages!" << endl;

            // The message consumption logic.
            std::vector<std::string> receiptHandles;
            for (std::vector<Message>::iterator iter = messages.begin();
                    iter != messages.end(); ++iter)
            {
                cout << "MessageId: " << iter->getMessageId()
                    << " PublishTime: " << iter->getPublishTime()
                    << " Tag: " << iter->getMessageTag()
                    << " Body: " << iter->getMessageBody()
                    << " FirstConsumeTime: " << iter->getFirstConsumeTime()
                    << " NextConsumeTime: " << iter->getNextConsumeTime()
                    << " ConsumedTimes: " << iter->getConsumedTimes()
                    << " Properties: " << iter->getPropertiesAsString()
                    << " ShardingKey: " << iter->getShardingKey() << endl;
                receiptHandles.push_back(iter->getReceiptHandle());
            }

            // Obtain an ACK from the consumer.
            // If the broker fails to receive an ACK for a message from the consumer before the period of time that is specified by the Message.NextConsumeTime parameter elapses, the broker delivers the message for consumption again.
            // A unique timestamp is specified for the handle of a message each time the message is consumed.
            AckMessageResponse bdmResp;
            consumer->ackMessage(receiptHandles, bdmResp);
            if (!bdmResp.isSuccess()) {
                // If the handle of a message times out, the broker cannot receive an ACK for the message from the consumer.
                const std::vector<AckMessageFailedItem>& failedItems =
                    bdmResp.getAckMessageFailedItem();
                for (std::vector<AckMessageFailedItem>::const_iterator iter = failedItems.begin();
                        iter != failedItems.end(); ++iter)
                {
                    cout << "AckFailedItem: " << iter->errorCode
                        << "  " << iter->receiptHandle << endl;
                }
            } else {
                cout << "Ack: " << messages.size() << " messages suc!" << endl;
            }
        } catch (MQServerException& me) {
            if (me.GetErrorCode() == "MessageNotExist") {
                cout << "No message to consume! RequestId: " + me.GetRequestId() << endl;
                continue;
            }
            cout << "Request Failed: " + me.GetErrorCode() + ".RequestId: " + me.GetRequestId() << endl;
#ifdef _WIN32
            Sleep(2000);
#else
            usleep(2000 * 1000);
#endif
        } catch (MQExceptionBase& mb) {
            cout << "Request Failed: " + mb.ToString() << endl;
#ifdef _WIN32
            Sleep(2000);
#else
            usleep(2000 * 1000);
#endif
        }

    } while(true);
}

Parâmetros de consumeMessageOrderly

Comportamento de reconhecimento

Após processar cada lote, chame ackMessage com os identificadores de recibo de todas as mensagens processadas.

• Bloqueio por lote: O consumidor não consegue buscar o próximo lote de uma partição até que todas as mensagens do lote atual sejam reconhecidas.

• Reentrega por timeout: Caso o broker não receba um ACK antes de NextConsumeTime, ele reenvia a mensagem com um novo identificador de recibo contendo um timestamp exclusivo.

• **Tratamento de MessageNotExist:** Este código de erro indica que não há mensagens disponíveis. O exemplo lida com isso continuando o loop de polling sem atraso.

• Recuperação de erros: Em outras exceções, o exemplo aguarda 2 segundos antes de tentar novamente para evitar loops de erro contínuos.