Elasticsearch 查询超过10000 的解决方案 – Python

文章目录

    • Elasticsearch 查询超过10000 的解决方案 - Python
      • 法1:修改 设置 max_result_size (不推荐)
      • 法2: scroll 分页
      • 法3: search_after 分页

Elasticsearch 查询超过10000 的解决方案 - Python

法1:修改 设置 max_result_size (不推荐)

# 调大查询窗口大小,比如100w (不推荐,慎用)
PUT test/_settings
{
  "index.max_result_window": "1000000"
}

# 查看 查询最大数
GET test/_settings
---
{
  "demo_scroll" : {
    "settings" : {
      "index" : {
        "number_of_shards" : "5",
        "provided_name" : "demo_scroll",
        "max_result_window" : "1000000",
        "creation_date" : "1680832840425",
        "number_of_replicas" : "1",
        "uuid" : "OLV5W_D9R-WBUaZ_QbGeWA",
        "version" : {
          "created" : "6082399"
        }
      }
    }
  }
}

法2: scroll 分页

    def getData(self):
        current_time = datetime.datetime.now()
        one_hour_ago = current_time - datetime.timedelta(hours=24)
        current_time_str = current_time.strftime('%Y-%m-%d %H:%M:%S')
        hours_ago_str = one_hour_ago.strftime('%Y-%m-%d %H:%M:%S')

        # 改为从elasticsearch读取数据
        es = Elasticsearch(hosts='http://127.0.0.1/9200',
                           timeout=1200)
        size = 10000
        query_scroll = {
            "size": size,
            "query": {
                "range": {
                    "create_time.keyword": {
                        "gte": hours_ago_str.__str__(),
                        "lte": current_time_str.__str__()
                    }
                }
            },
            "_source": ["ip_address", "OS", "host", "user", "create_time"],
        }
        scroll = "10m" # 该次连接超时时间设置
        result = []
        # first
        init_res = es.search(index="nac-users", body=query_scroll, scroll=scroll)
        scroll_id = init_res["_scroll_id"]
        for item in init_res["hits"]["hits"]:
            result.append({
                'id': item['_id'],
                'ip_address': item['_source']['ip_address'],
                'operating_system': item['_source']['OS'],
                'hostname': item['_source']['host'],
                'username': item['_source']['user'],
                'date_t': item['_source']['create_time'],
            })
        i = 0
        while i < 16:  # 剩下的数据 一天 24 小时数据估计不会超过 160000
            res = es.scroll(scroll_id=scroll_id, scroll=scroll)
            if len(res["hits"]["hits"]) == 0:
                break
            for item in res["hits"]["hits"]:
                result.append({
                    'id': item['_id'],
                    'ip_address': item['_source']['ip_address'],
                    'operating_system': item['_source']['OS'],
                    'hostname': item['_source']['host'],
                    'username': item['_source']['user'],
                    'date_t': item['_source']['create_time'],
                })
            i = i + 1
            
        # 原始的    
        # {"query": {"match_all": {}}, "size": 10000}
        # res = es.search(index="nac-users", body=query_scroll)
        #
        # result = []
        # for item in res['hits']['hits']:
        #     result.append({
        #         'id': item['_id'],
        #         'ip_address': item['_source']['ip_address'],
        #         'operating_system': item['_source']['OS'],
        #         'hostname': item['_source']['host'],
        #         'username': item['_source']['user'],
        #         'date_t': item['_source']['create_time'],
        #     })
        self.data = pd.DataFrame(result)

法3: search_after 分页

def getData(self):
      current_time = datetime.datetime.now()
      one_hour_ago = current_time - datetime.timedelta(hours=24)
      current_time_str = current_time.strftime('%Y-%m-%d %H:%M:%S')
      hours_ago_str = one_hour_ago.strftime('%Y-%m-%d %H:%M:%S')

      # 改为从elasticsearch读取数据
      es = Elasticsearch(hosts='http://127.0.0.1:9200',
                         timeout=1200)
      size = 10000
      query_scroll = {
          "size": size,
          "query": {
              "range": {
                  "create_time.keyword": {
                      "gte": hours_ago_str.__str__(),
                      "lte": current_time_str.__str__()
                  }
              }
          },
          "sort": [
              {
                  "create_time.keyword": {
                      "order": "desc"
                  }
              }
          ],
          "_source": ["ip_address", "OS", "host", "user", "create_time"],
      }
      result = []
      init_res = es.search(index="nac-users", body=query_scroll)
      if len(init_res["hits"]["hits"]) == 0:
          self.data = pd.DataFrame(result)
          return
      sort = init_res["hits"]["hits"][0]["sort"] # 我这里是用时间来排序的,所以取到的是时间字段
      for item in init_res["hits"]["hits"]:
          result.append({
              'id': item['_id'],
              'ip_address': item['_source']['ip_address'],
              'operating_system': item['_source']['OS'],
              'hostname': item['_source']['host'],
              'username': item['_source']['user'],
              'date_t': item['_source']['create_time'],
          })
      i = 0
      while i < 16:
          query_scroll["search_after"] = sort
          res = es.search(index="nac-users", body=query_scroll)
          sort = res["hits"]["hits"][0]["sort"]
          if len(res["hits"]["hits"]) == 0:
              break
          for item in res["hits"]["hits"]:
              result.append({
                  'id': item['_id'],
                  'ip_address': item['_source']['ip_address'],
                  'operating_system': item['_source']['OS'],
                  'hostname': item['_source']['host'],
                  'username': item['_source']['user'],
                  'date_t': item['_source']['create_time'],
              })
          i = i + 1
    self.data = pd.DataFrame(result)        

还有一个方法是在参考文章2里面提到的track_total_hits,但是我测试的时候没起作用,目前还不太清楚原因。。。

我看参考文章里说到search_after 分页要比scroll快,但是在我的数据上是scroll要快很多,不是特别清楚,可能我这里的数据暂时只有2w多一点,感觉用到search_after 分页需要排序,可能是排序的字段的问题,时间字段我存的是字符串格式,,如有可以修改的地方,欢迎大家指正~ 有更多可以参考的方法欢迎贴在评论区供大家参考~

【参考1】https://juejin.cn/post/7224369270141993019
【参考2】https://blog.csdn.net/u011250186/article/details/125483759