您可以使用Kafka的时间管理器（TimeBasedUUID）函数将消息的延迟时间与新消息的主题一起存储在Kafka中。 在此之后，将启动一个Kafka Consumer Group，该组将，根据您设置的时间，在指定时间后消耗并发送该消息。以下是Kafka消息延迟发送的代码示例。

在这个示例中，使用UUID来生成具有唯一ID的随机字符串。消息格式将包含（topic，message，delay）。在示例中，我们设置了一个延迟时间为3秒钟。

producer.py:

import time
import json    
from kafka import KafkaProducer    
from uuid import uuid4

producer = KafkaProducer(bootstrap_servers='localhost:9092')

def send_to_kafka(topic, data, delay):
    obj = {"data": data, "topic": topic, "delay": delay}
    future = producer.send("delayed-messages", json.dumps(obj).encode('utf-8'), key=str(uuid4()).encode('utf-8'))
    result = future.get(timeout=60)
    print(result)

if __name__ == "__main__":
    send_to_kafka("my-topic", {"id": 1, "message": "Hello World!"}, 3)

consumer.py:

import time    
import json    
from kafka import KafkaConsumer, KafkaProducer    
from datetime import datetime, timedelta    
from uuid import UUID, uuid4  
  
KAFKA_TOPIC = "delayed-messages"  
KAFKA_CONSUMER_GROUP = "delayed-consumer-group"  
KAFKA_SERVER = "localhost:9092"

consumer = KafkaConsumer(KAFKA_TOPIC, bootstrap_servers=KAFKA_SERVER, groupId=KAFKA_CONSUMER_GROUP)    
producer = KafkaProducer(bootstrap_servers='localhost:9092')

for message in consumer:  
    value = message.value  
    
    data = json.loads(value.decode('utf-8'))  
    currtime = datetime.now()  
    timestamp = (currtime + timedelta(seconds=data["delay"])).strftime('%Y-%m-%d %H:%M:%S')

    if currtime >= datetime.strptime(timestamp, '%Y-%m-%d %H:%M:%S'):  
        print("Message sent: "  + str(value))  
        future = producer.send(data["topic"], json.dumps(data["data"]).encode('utf-8'))  
        result = future.get(timeout=60)  
        print(result)