那麼可以用es的scroll方式，實際是一種深度分頁機制

直接上程式碼：

#-*- coding:utf8 -*-

from elasticsearch import Elasticsearch, helpers
import json
import pdb

class ElasticsearchService:

    def __init__(self, hosts):
        self.__elasticsearch = Elasticsearch(hosts, sniff_on_start=True, sniff_on_connection_fail=True, sniffer_timeout=60, timeout=30, retry_on_timeout=True, max_retries=5)

    def search_scroll(self, index, doc_type, query):
        try:
            return self.__elasticsearch.search(index=index, doc_type=doc_type, body=query, search_type="query_then_fetch", scroll="1m")
        except BaseException as e:
            print str(e)
            pass

        return {}

    def scroll_scan(self, query):
        try:
            resJson = self.__elasticsearch.scroll(body=query)
            return resJson 
        except BaseException as e:
            print str(e)
            pass

        return []

    def delete_by_query(self, index, doc_type, query):
        try:
            resJson = self.__elasticsearch.delete_by_query(index=index, doc_type=doc_type, body=query)
            return resJson.get('deleted')
        except BaseException as e:
            print 'delete fail'
            pass

        return 0

    def update_by_query(self, index, doc_type, query):
        try:
            resJson = self.__elasticsearch.update_by_query(index=index, doc_type=doc_type, body=query)
            return resJson.get('updated')
        except BaseException as e:
            pass

        return 0 

    def insert_bulk(self, data_lst):
        try:
            return helpers.bulk(self.__elasticsearch, data_lst, request_timeout=60) 
        except BaseException as e:
            return [0, []]


if __name__ == '__main__':
    es_hosts = ["10.234.22.22:9200"]
    baike_all_index = 'baike_index'
    baike_all_type = 'baike_all'
    elastic_service = ElasticsearchService(es_hosts)
    #這裡是進行第一次查詢，query中size指定每個批次的大小，返回的結果中不僅有查詢到的資料，還有一個scroll_id， 這個scrool_id可以認為是下一次查詢的起始位置
    res = elastic_service.search_scroll(baike_all_index, baike_all_type, {"query": {"match_all": {}},"_source": ["url"], "size": 10000})
    hits = res.get('hits')
    if hits.get('total') > 0:
        for hit in hits.get('hits'):
            print hit['_source']['url']  #這裡是取我自己的資料
    while res.get('_scroll_id') and hits.get('total') > 0:
        #後續的每次查詢都需要帶上上一次查詢結果中得到的scroll_id引數
        res = elastic_service.scroll_scan({'scroll': '1m', 'scroll_id': res.get('_scroll_id')})
        hits = res.get('hits')
        if hits.get('total') > 0:
            for hit in hits.get('hits'):
                print hit['_source']['url']