学习用Node.js和Elasticsearch构建搜索引擎（7）：零停机时间更新索引配置或迁移索引

上一篇说到如果一个索引的mapping设置过了，想要修改type或analyzer，通常的做法是新建一个索引，重新设置mapping，再把数据同步过来。

那么如何实现零停机时间更新索引配置或迁移索引？这就需要用到索引的别名设置。

思路：

1、假设我们的索引是demo_v1，我们定义了一个别名demo，以后所有的操作都用别名demo操作。

2、现在索引demo_v1的mapping设置或者其他一些设置不满足我们的需求了，我们需要修改。新建一个索引demo_v2，同时设置好最新配置。

3、同步索引demo_v1的数据到索引demo_v2。直到同步完。

4、移除索引demo_v1的别名demo，同时设置索引demo_v2的别名为demo。

5、删除索引demo_v1。

6、迁移完成。以后如果还有设置变更，可以按照这个思路继续设置索引demo_v3、demo_v4……

接下来用一个例子说明实现过程，实际项目中我也是按这个思路做的。如果有一些命令操作看不懂，可参看上一篇文章。

1、创建索引demo_v1

> curl -XPUT 'localhost:9200/demo_v1'
{"acknowledged":true,"shards_acknowledged":true}%

2、给索引demo_v1添加几条数据

#给索引demo_v1添加了type=fruit的3条数据，每条数据用name和tag两个字段
> curl -XPOST 'localhost:9200/_bulk?pretty' -d'
{ "index" : { "_index" : "demo_v1", "_type" : "fruit", "_id" : "" }}
{ "name" : "苹果","tag":"苹果,水果,红富士"}
{ "create" : { "_index" : "demo_v1", "_type" : "fruit", "_id" : "" }}
{ "name" : "香蕉","tag":"香蕉,水果,海南,弯弯,小黄人"}
{ "index" : { "_index" : "demo_v1", "_type" : "fruit", "_id" : "" }}
{ "name" : "西瓜","tag":"西瓜,水果,圆形,绿,闰土"}
'
#返回
{
  "took" : ,
  "errors" : false,
  "items" : [
    {
      "index" : {
        "_index" : "demo_v1",
        "_type" : "fruit",
        "_id" : "",
        "_version" : ,
        "result" : "created",
        "_shards" : {
          "total" : ,
          "successful" : ,
          "failed" :
        },
        "created" : true,
        "status" :
      }
    },
    {
      "create" : {
        "_index" : "demo_v1",
        "_type" : "fruit",
        "_id" : "",
        "_version" : ,
        "result" : "created",
        "_shards" : {
          "total" : ,
          "successful" : ,
          "failed" :
        },
        "created" : true,
        "status" :
      }
    },
    {
      "index" : {
        "_index" : "demo_v1",
        "_type" : "fruit",
        "_id" : "",
        "_version" : ,
        "result" : "created",
        "_shards" : {
          "total" : ,
          "successful" : ,
          "failed" :
        },
        "created" : true,
        "status" :
      }
    }
  ]
}

3、给索引demo_v1设置别名demo

#设置别名
> curl -XPUT 'localhost:9200/demo_v1/_alias/demo'
{"acknowledged":true}%

4、使用别名查看信息

#使用别名查看一下数据，是可以查询到的
> curl -XGET 'localhost:9200/demo/fruit/_search?pretty'

#查看mapping
> curl -XGET 'localhost:9200/demo/fruit/_mapping?pretty'
#返回
{
  "demo_v1" : {
    "mappings" : {
      "fruit" : {
        "properties" : {
          "name" : {
            "type" : "text",
            "fields" : {
              "keyword" : {
                "type" : "keyword",
                "ignore_above" :
              }
            }
          },
          "tag" : {
            "type" : "text",
            "fields" : {
              "keyword" : {
                "type" : "keyword",
                "ignore_above" :
              }
            }
          }
        }
      }
    }
  }
}

#检索数据
> curl -XGET 'http://localhost:9200/demo/fruit/_search?pretty' -d '{
    "query" : {
        "term" : { "tag" : "水" }
   }
}'
#返回
{
  "took" : ,
  "timed_out" : false,
  "_shards" : {
    "total" : ,
    "successful" : ,
    "failed" :
  },
  "hits" : {
    "total" : ,
    "max_score" : 0.28582606,
    "hits" : [
      {
        "_index" : "demo_v1",
        "_type" : "fruit",
        "_id" : "",
        "_score" : 0.28582606,
        "_source" : {
          "name" : "苹果",
          "tag" : "苹果,水果,红富士"
        }
      },
      {
        "_index" : "demo_v1",
        "_type" : "fruit",
        "_id" : "",
        "_score" : 0.27233246,
        "_source" : {
          "name" : "西瓜",
          "tag" : "西瓜,水果,圆形,绿,闰土"
        }
      },
      {
        "_index" : "demo_v1",
        "_type" : "fruit",
        "_id" : "",
        "_score" : 0.24257512,
        "_source" : {
          "name" : "香蕉",
          "tag" : "香蕉,水果,海南,弯弯,小黄人"
        }
      }
    ]
  }
}

数据因为先前创建索引时没有设置mapping，所以这些设置都是默认设置，分词器也默认标准分词器。

上面检索标签tag中带有“水”的数据，都查询出来了，说明默认分词器把“水果”这个词拆分了。

如果我们需要tag字段按照逗号分词，“水果”作为一个完整的词不拆分该怎么弄呢？

5、新建一个索引demo_v2，同时自定义逗号分词器，并把逗号分词器应用到tag字段上

#新建索引demo_v2
> curl -XPUT 'http://localhost:9200/demo_v2/' -d'{
  "settings": {
    "index": {
      "analysis": {
        "analyzer": {
          "douhao_analyzer": {
            "pattern": ",",
            "type": "pattern"
          }
        }
      },
      "number_of_shards": ,
      "number_of_replicas":
    }
  },
  "mappings": {
    "fruit": {
      "properties": {
        "name": {
          "type": "text",
          "index": "not_analyzed"
        },
        "tag": {
          "type": "string",
          "analyzer": "douhao_analyzer",
          "search_analyzer": "douhao_analyzer"
        }
      }
    }
  }
}'
#返回
{"acknowledged":true,"shards_acknowledged":true}%

关于mapping设置及分词器设置可参见官方文档：

https://www.elastic.co/guide/en/elasticsearch/reference/5.3/mapping.html#mapping-type

https://www.elastic.co/guide/en/elasticsearch/reference/5.3/analysis-analyzers.html

6、同步索引demo_v1中的数据到demo_v2

我使用工具elasticdump同步数据，ElasticDump是一个ElasticSearch的数据导入导出开源工具包。

官方地址：https://github.com/taskrabbit/elasticsearch-dump

同步命令如下：

> elasticdump --input='http://localhost:9200/demo_v1'  --output='http://localhost:9200/demo_v2' --type=data
Wed,  Jun  :: GMT | starting dump
Wed,  Jun  :: GMT | got  objects from source elasticsearch (offset: )
Wed,  Jun  :: GMT | sent  objects to destination elasticsearch, wrote
Wed,  Jun  :: GMT | got  objects from source elasticsearch (offset: )
Wed,  Jun  :: GMT | Total Writes:
Wed,  Jun  :: GMT | dump complete

7、验证一下demo_v2中的数据

#检索tag中包含“水”的数据，检索不到就是正常的
curl -XGET 'http://localhost:9200/demo_v2/fruit/_search?pretty' -d '{
    "query" : {
        "term" : { "tag" : "水" }
   }
}'

#检索tag中包含“水果”的数据，可以全部检索到
curl -XGET 'http://localhost:9200/demo_v2/fruit/_search?pretty' -d '{
    "query" : {
        "term" : { "tag" : "水果" }
   }
}'

8、移除索引demo_v1的别名demo，同时设置索引demo_v2的别名为demo

curl -XPOST 'localhost:9200/_aliases?pretty' -d'{
  "actions" : [
  { "remove" : { "index" : "demo_v1", "alias" : "demo" } },
  { "add" : { "index" : "demo_v2", "alias" : "demo" } }
]}'

9、删除索引demo_v1

curl -XDELETE 'localhost:9200/demo_v1'

至此整个迁移完成

ok!