【问题标题】:Apache Pulsar: Read / Consume messages from an integer specified message id to an end message Id?Apache Pulsar:从指定的整数消息 ID 读取/使用消息到结束消息 ID?
【发布时间】:2022-02-09 11:53:45
【问题描述】:

使用 Kafka,我可以指定一个整数消息 id 开始消费和结束消息,例如如下:

 kafkacat -b kafka:9092 -t messages -o 11000 -c 11333

但是,在 Apache Pulsar 中似乎没有指定整数开始和停止消息的相同功能!

公平地说,可以指定一个开始消息 id 和结束消息 id,如果它们已经被跟踪并以字节格式保存,使用一个非常复杂的过程,这势必会影响性能和代码复杂性。

如本例所示:

client, err := NewClient(pulsar.ClientOptions{
    URL: lookupURL,
})

if err != nil {
    log.Fatal(err)
}
defer client.Close()

topic := "topic-1"
ctx := context.Background()

// create producer
producer, err := client.CreateProducer(pulsar.ProducerOptions{
    Topic:           topic,
    DisableBatching: true,
})
if err != nil {
    log.Fatal(err)
}
defer producer.Close()

// send 10 messages
msgIDs := [10]MessageID{}
for i := 0; i < 10; i++ {
    msgID, err := producer.Send(ctx, &pulsar.ProducerMessage{
        Payload: []byte(fmt.Sprintf("hello-%d", i)),
    })
    assert.NoError(t, err)
    assert.NotNil(t, msgID)
    msgIDs[i] = msgID
}

// create reader on 5th message (not included)
reader, err := client.CreateReader(pulsar.ReaderOptions{
    Topic:          topic,
    StartMessageID: msgIDs[4],
})

if err != nil {
    log.Fatal(err)
}
defer reader.Close()

// receive the remaining 5 messages
for i := 5; i < 10; i++ {
    msg, err := reader.Next(context.Background())
    if err != nil {
    log.Fatal(err)
}

// create reader on 5th message (included)
readerInclusive, err := client.CreateReader(pulsar.ReaderOptions{
    Topic:                   topic,
    StartMessageID:          msgIDs[4],
    StartMessageIDInclusive: true,
})

if err != nil {
    log.Fatal(err)
}
defer readerInclusive.Close()

但是,这对于多个并发阅读器来说是复杂且不可靠(或复杂)的,并且需要使用外部构造来跟踪已处理的消息,然后才能使用开始/结束语义进行检索。

有什么方法可以实现(最好通过golang)

【问题讨论】:

  • 无论使用何种语言,完成您所要求的唯一方法就是将消费者寻找到那些偏移量并计算已处理消息的数量。不过,您是否有特定原因需要跟踪偏移量/ID?
  • 你是对的。我现在采取了将所有消息流式传输出来的方法,只需在我想要的范围内挑选和保存消息。关于原因,它是在给定的时间窗口中挑选出特定的消息序列并重放它们,以便重现该期间的系统负载模式。

标签: go apache-kafka apache-pulsar


【解决方案1】:

我发现以下简单方法就足够了(概念证明脚本):

package main

import (
    "context"
    "fmt"
    "io/ioutil"
    "log"
    "os"
    "strconv"
    "strings"
    "time"

    "github.com/apache/pulsar-client-go/pulsar"
)

func writeBytesToFile(f string, byteSlice []byte) int {
    // Open a new file for writing only

    f = "./data/" + f

    file, err := os.OpenFile(
        f,
        os.O_WRONLY|os.O_TRUNC|os.O_CREATE,
        0666,
    )
    if err != nil {
        log.Fatal(err)
    }
    defer file.Close()

    // Write bytes to file
    bytesWritten, err := file.Write(byteSlice)

    if err != nil {
        log.Fatal(err)
    }

    log.Printf("Wrote %d bytes.\n", bytesWritten)

    return bytesWritten
}

func readBackByEntryId(msgDir string, msgIndex string) (yourBytes []byte) {

    //We know the file name by convention
    fname := msgDir + "/" + msgIndex + ".dat"

    yourBytes, err := ioutil.ReadFile(fname)

    if err != nil {
        log.Printf("error reading %s", fname)
        return nil
    }

    return yourBytes
}

func getFiles(aDir string) []string {

    var theFiles []string

    files, err := ioutil.ReadDir("./data/")

    if err != nil {
        log.Fatal(err)
    }

    for _, f := range files {

        theFiles = append(theFiles, f.Name())

    }

    return theFiles
}

func streamAll(reader pulsar.Reader, startMsgIndex int64, stopMsgIndex int64) {

    read := false

    for reader.HasNext() {

        msg, err := reader.Next(context.Background())

        if err != nil {
            log.Fatal(err)
        }

        //can I access the details of the message ? yes
        fmt.Printf("%v -> %#v\n", msg.ID().EntryID(), msg.ID())

        //Can i serialize into bytes? Yes
        myBytes := msg.ID().Serialize()

        //Can I store it somewhere? Perhaps a map ? or even on disk in a file ?
        //In other words: Can I write a byte[] slice to a file? Yes!
        msgIndex := msg.ID().EntryID()

        if msgIndex == startMsgIndex {
            fmt.Println("start read: ", msgIndex)
            read = true
        }

        if msgIndex > stopMsgIndex {
            fmt.Println("stop reading: ", msgIndex)
            read = false
        }

        if read == false {

            fmt.Println("skipping ", msgIndex)

        } else {

            fname := strconv.FormatInt(msgIndex, 10) + ".dat"

            fmt.Println("written bytes: ", writeBytesToFile(fname, myBytes))

            fmt.Printf("Received message msgId: %#v -- content: '%s' published at %v\n",
                msg.ID(), string(msg.Payload()), msg.PublishTime())

        }

        /*
            //FYI - to save and reread a msgId from store: https://githubmemory.com/@storm-5
            msgId := msg.ID()
            msgIdBytes := msgId.Serialize()
            idNew, _ := pulsar.DeserializeMessageID(msgIdBytes)

            readerInclusive, err := client.CreateReader(pulsar.ReaderOptions{
                Topic:                   "ragnarok/transactions/requests",
                StartMessageID:          idNew,
                StartMessageIDInclusive: true,
            })
        */
    }

}

func retrieveRange(client pulsar.Client) {

    someFiles := getFiles("./data/")

    for _, f := range someFiles {

        fIndex := strings.Split(f, ".")[0]

        fmt.Println("re-reading message index -> ", fIndex)

        msgIdBytes := readBackByEntryId("./data", fIndex)

        fmt.Printf("boom -> %#v\n", msgIdBytes)

        idNew, err := pulsar.DeserializeMessageID(msgIdBytes)

        if err != nil {
            log.Fatal(err)
        }

        fmt.Println("Got message entry id => ", idNew.EntryID())

        readerInclusive, err := client.CreateReader(pulsar.ReaderOptions{
            Topic:                   "ragnarok/transactions/requests",
            StartMessageID:          idNew,
            StartMessageIDInclusive: true,
        })

        if err != nil {
            log.Fatal(err)
        }

        defer readerInclusive.Close()

        //defer readerInclusive.Close()
        fmt.Println("bleep!")

        msg, err := readerInclusive.Next(context.Background())

        if err != nil {
            log.Fatal(err)
        }

        //fmt.Println("retrieved message -> ", string(msg.Payload()))
        fmt.Printf("Retrieved message ID message msgId: %#v -- content: '%s' published at %v\n",
            msg.ID(), string(msg.Payload()), msg.PublishTime())

    }
}

func main() {

    client, err := pulsar.NewClient(
        pulsar.ClientOptions{
            URL:               "pulsar://localhost:6650",
            OperationTimeout:  30 * time.Second,
            ConnectionTimeout: 30 * time.Second,
        })

    if err != nil {
        log.Fatalf("Could not instantiate Pulsar client: %v", err)
    }

    defer client.Close()

    reader, err := client.CreateReader(pulsar.ReaderOptions{
        Topic:          "ragnarok/transactions/requests",
        StartMessageID: pulsar.EarliestMessageID(),
    })

    if err != nil {
        log.Fatal(err)
    }

    defer reader.Close()

    if err != nil {
        log.Fatal(err)
    }

    var startMsgId int64 = 55
    var stopMsgId int64 = 66

    //stream all the messages from the earliest to latest
    //pick a subset between a start and stop id
    streamAll(reader, startMsgId, stopMsgId)

    //retrieve the picked range
    retrieveRange(client)

}



【讨论】:

  • 您真的需要阅读所有内容然后检查 ID 吗? pulsar客户端没有kafka这样的seek功能吗?
  • 我能找到的唯一文档(关闭匹配)是:pulsar.apache.org/docs/en/client-libraries-go/#reader-example,使用起来非常尴尬,可能意味着更大的性能损失。
  • 见函数Seek(msgID MessageID)。你不需要从最早的偏移量开始,等到你到达起始 id 来处理任何东西
猜你喜欢
  • 1970-01-01
  • 2014-07-20
  • 2022-11-02
  • 1970-01-01
  • 1970-01-01
  • 1970-01-01
  • 2022-08-02
  • 1970-01-01
  • 2020-10-20
相关资源
最近更新 更多