自建ES通过OSS快照迁移至阿里云ES-CSDN博客

在同一个VPC内使用3台ECS部署自建ES集群，可以通过给自建Elasticsearch打快照存储到OSS，在利用OSS中的快照进行数据库恢复，将自建Elasticsearch迁移至阿里云Elasticsearch。
TIM_20180411231841

安装OSS快照仓库插件，下载插件并解压到ES集群各节点Elasticsearch根目录下的plugins目录

# wget https://github.com/zhichen/elasticsearch-repository-oss/releases/download/v5.5.3/elasticsearch-repository-oss-5.5.3.zip
# unzip elasticsearch-repository-oss-5.5.3.zip -d /srv/elasticsearch-5.4.3/plugins/

修改配置文件中的版本号为当前ES集群的版本，完成后需要重启ES各节点。

# cat /srv/elasticsearch-5.4.3/plugins/elasticsearch/plugin-descriptor.properties |grep version |grep -v "^#"
version=5.4.3
java.version=1.8
elasticsearch.version=5.4.3

新购与自建ES集群在同一个region的标准存储类型OSS，（阿里云暂不支持归档存储类型OSS）。

在自建ES上创建仓库，OSS所属阿里云主账号AK信息，若使用子账号需要有管理OSS的权限。

PUT _snapshot/robin_backup
{
    "type": "oss",
    "settings": {
        "endpoint": "http://oss-cn-hangzhou-internal.aliyuncs.com", 
        "access_key_id": "Lxxxxxxxxxa",
        "secret_access_key": "Nxxxxxxxxxxxxxxxxxxpt",
        "bucket": "shasengtest", 
        "compress": true
    }
}

可以根据实际情况在setting中增加设置

        "chunk_size": "500mb", #限制snapshot过程中分块的大小，超过这个大小，数据将会被分块上传到OSS中
        "base_path": "snapshot/" #设置仓库的起始位置，默认为根目录

获取全部仓库信息，设置了robin_backup及robin_backup2两个仓库

# curl -XGET es1:9001/_snapshot/?pretty
{
  "robin_backup" : {
    "type" : "oss",
    "settings" : {
      "bucket" : "shasengtest",
      "compress" : "true",
      "endpoint" : "http://oss-cn-hangzhou-internal.aliyuncs.com"
    }
  },
  "robin_backup2" : {
    "type" : "oss",
    "settings" : {
      "bucket" : "robin7",
      "compress" : "true",
      "endpoint" : "http://oss-cn-hangzhou-internal.aliyuncs.com"
    }
  }
}

删除仓库

DELETE _snapshot/robin_backup1

获取指定仓库信息

# curl -XGET es1:9001/_snapshot/robin_backup?pretty
{
  "robin_backup" : {
    "type" : "oss",
    "settings" : {
      "bucket" : "shasengtest",
      "compress" : "true",
      "endpoint" : "http://oss-cn-hangzhou-internal.aliyuncs.com"
    }
  }
}

备份当前自建ES集群所有打开的索引到指定仓库，创建快照并命名为test

PUT _snapshot/robin_backup/test

获得test快照信息，可以看到备份了当前集群全部的索引test，test1，.kibana

# curl -XGET es1:9001/_snapshot/robin_backup/test?pretty
{
  "snapshots" : [
    {
      "snapshot" : "test",
      "uuid" : "oXtFY174RyeKRgYFi_dXKA",
      "version_id" : 5040399,
      "version" : "5.4.3",
      "indices" : [
        "test",
        "test1",
        ".kibana"
      ],
      "state" : "SUCCESS",
      "start_time" : "2018-04-10T12:44:26.027Z",
      "start_time_in_millis" : 1523364266027,
      "end_time" : "2018-04-10T12:44:27.204Z",
      "end_time_in_millis" : 1523364267204,
      "duration_in_millis" : 1177,
      "failures" : [ ],
      "shards" : {
        "total" : 11,
        "failed" : 0,
        "successful" : 11
      }
    }
  ]
}

备份指定test2索引到robin_backup下并命名快照为test2

PUT _snapshot/robin_backup/test2
{
  "indices": "test2"
}

获得单个快照test2信息，指包含了test2这一个索引

# curl -XGET es1:9001/_snapshot/robin_backup/test2?pretty
{
  "snapshots" : [
    {
      "snapshot" : "test2",
      "uuid" : "L6MknW92RGe55fZS1HIVUw",
      "version_id" : 5040399,
      "version" : "5.4.3",
      "indices" : [
        "test2"
      ],
      "state" : "SUCCESS",
      "start_time" : "2018-04-10T12:47:19.430Z",
      "start_time_in_millis" : 1523364439430,
      "end_time" : "2018-04-10T12:47:20.099Z",
      "end_time_in_millis" : 1523364440099,
      "duration_in_millis" : 669,
      "failures" : [ ],
      "shards" : {
        "total" : 5,
        "failed" : 0,
        "successful" : 5
      }
    }
  ]
}

删除索引test及test1，并关闭索引.kibana

# curl -XDELETE es2:9001/test
{"acknowledged":true}
# curl -XDELETE es2:9001/test1
{"acknowledged":true}
# curl -XPOST es2:9001/.kibana/_close
{"acknowledged":true}
# curl -XGET es2:9001/_cat/indices
green open  test3   YcU56zKqTNaTNH8Vcr6jpw 5 1 168 0 395.8kb 217.5kb
green open  test2   M-kOMZVkRcWwOwQlFRKRWw 5 1 139 0 343.1kb 171.5kb
      close .kibana Lcv2WmUPTxqPjoNJV2C7sQ

elasticsearch的索引快照与ECS快照不同，并不是打快照那一时刻的整个集群状态。这里可以把索引理解成ECS下挂载的数据盘之一，只需要对重要的数据盘打快照，并不是所有的磁盘都需要打快照，恢复时可以选择只恢复某一块盘的数据，而不是对所有的磁盘进行回滚。

从test快照恢复索引

# curl -XPOST es1:9001/_snapshot/robin_backup/test/_restore
{"accepted":true}

查看恢复效果

# curl -XGET es2:9001/_cat/indices
green open test    fEukT7eWQ-OjD3uzxKmXcg 5 1 211 0 507.6kb 253.8kb
green open test1   mw6MvvuZSlys0YLjo-UU-Q 5 1 277 0 611.3kb 305.6kb
green open .kibana Lcv2WmUPTxqPjoNJV2C7sQ 1 1   2 0  15.4kb   7.7kb
green open test3   YcU56zKqTNaTNH8Vcr6jpw 5 1 168 0 395.8kb 217.5kb
green open test2   M-kOMZVkRcWwOwQlFRKRWw 5 1 139 0 343.1kb 171.5kb

也可以安装cerebro插件查看已存在的索引快照，单击restore即可恢复索引。
cerebro

也可以选择只恢复部分索引

# curl -XPOST es1:9001/_snapshot/robin_backup/test/_restore
{
    "indices": "test1", #指定需要恢复的索引名称
}

在阿里云ES上创建相同的快照仓库，与自建的ES集群共享同一个oss仓库，完成后即可查看到快照信息
snapshot

查看当前阿里云ES集群的索引
snapshot2

在阿里云ES上恢复快照，可以看到已经恢复的test和test1
snapshot3

由于阿里云ES存在.kibana索引，恢复快照时需要先关闭.kibana索引，否则会报错

{
  "error": {
    "root_cause": [
      {
        "type": "snapshot_restore_exception",
        "reason": "[robin_backup:test/oXtFY174RyeKRgYFi_dXKA] cannot restore index [.kibana] because it's open"
      }
    ],
    "type": "snapshot_restore_exception",
    "reason": "[robin_backup:test/oXtFY174RyeKRgYFi_dXKA] cannot restore index [.kibana] because it's open"
  },
  "status": 500
}