用相同的GroupId消费已经消费过的信息，可以进行回滚。

import time
import redis
import json
import math
import Queue
import os

def cont(t):
    time_local = time.localtime(t)
    dt = time.strftime("%Y-%m-%d %H:%M:%S",time_local)
    return dt


def seek2ts(ts,topic,consumer):

    x= [0,0,0,0,0,0 ]
    ts*=1000
    m=[]
    for y in range(6):
       i=TopicPartition(topic, y)
       x[y] =  consumer.offsets_for_times({i:ts})[i][0]
       m.append(i)
    consumer.assign(m)
    for y in range(6):
            i=TopicPartition(topic, y)
            consumer.seek(i,x[y])
    return x


def main():
        consumer = KafkaConsumer(bootstrap_servers=[#集群IP] , group_id ="")
        seek2ts(1530748800#回滚到的时间戳,"#TOPIC",consumer)
        consumer.poll()
        consumer.commit()

        print "done"

main()

在linux查询kafka中GroupId的消费情况：

/root/kafka_2.11-0.11.0.2/bin/kafka-run-class.sh kafka.admin.ConsumerGroupCommand --bootstrap-server k3:9092  --describe --group online_dist2ali 

k3:9092   是kafka地址

online_dist2ali  是GroupId