elasticsearch5.0.1集群一次误删除kibana索引引发的血案

elasticsearch集群中一次删除kibana索引引发的血案


1.问题发生的过程：
早上的时候有某个索引无法看到报表数据，于是就点该报表多次，估计集群被点挂了，报错：Elasticsearch is still initializing the kibana index

当时有点慌估计是昏了头，直接根据baidu某篇博文的提示进行了操作
curl -XDELETE http://localhost:9200/.kibana

这下悲剧发生了，kibana控制台没有东西了，业务部门无法查询报表。。。

后面在http://stackoverflow.com/上搜索到了老外的解释，会丢失所有的配置、索引、图形和报表：

the curl -XDELETE http://localhost:9200/.kibana command works fine, however you lost all your kibana' settings (indexes, graphs, dashboards)


2.解决思路
之前有5个节点部署在阿里云其中有两个节点因为磁盘空间不足扩容，扩容前做了镜像，抱着试试的态度新买服务器挂接上之前的镜像，
找到之前的操作记录：


发现 .kibana索引的uuid为E9kS4THKREKR36IuPICPIA，大小大约有179k，如下

# curl '10.26.241.237:9200/_cat/indices?v'health status index               uuid                   pri rep docs.count docs.deleted store.size pri.store.sizegreen  open   .kibana             E9kS4THKREKR36IuPICPIA   1   1         41            5    179.3kb         89.6kbgreen  open   voice:user:login    VMpfcIiFS9OixB-J7-ZLWw   5   1  695465821            0    323.4gb        162.3gbgreen  open   voice:user:logout   W8MAAbp7RO6ZYYx7FcfcVA   5   1  686515590            0      279gb        138.3gbgreen  open   push:task:result    qWs38E_eQbCicgB312PE8w   5   1   91340303            0     22.1gb         10.9gbyellow open   push:user:req       wqJi6jTFT-a0ZN-57Z63Yw   5   1 8476340396            0      5.4tb          3.4tbgreen  open   voice:mic:used      pdZZr8mdSwirBGSAYTvIwg   5   1    8925004            0      3.5gb          1.7gbgreen  open   user:register       LZ_DwUpDRAyc0gfsFlCdGA   5   1   64626999            0     22.4gb         11.1gbgreen  open   push:user:app       0ivr0VubTCG5mFM0yoq34w   5   1  138403810            0    305.4gb        152.2gbgreen  open   voice:send:text     rrS8Kd4nRlim7wEOMqJ1wA   5   1      27960            0     21.6mb         10.8mbgreen  open   voice:mic:lose      iUhw676hTTSsyJnv3K0s6Q   5   1    2847952            0      1.7gb          902mbgreen  open   user:login          C-qSmB0ST2CrMkb7snrbKQ   5   1 2623263965            0    606.4gb        304.1gbgreen  open   push:user:task      7XPBJeBWRbas5t1XwiKt2Q   5   1   53518658            0       21gb         10.5gbgreen  open   speech:voice:result 7wISjRCeQZSY4SsCNunh8w   5   1  650234559            0      578gb        289.6gbgreen  open   script:user:info    qTZEpjkmRiSyyL1WBpMS5g   5   1          0            0      1.4kb           766bgreen  open   speech:voice:upload VBrFZq8QScOFYN1jsCkN4A   5   1 4469453923            0      1.9tb        973.5gb

发现虽然有kibana的索引但是只有在es的数据目录/data/es/data/nodes/0/indices下查找，发现只有不到30k明显不是我们需要的数据

将这台服务器退掉，重新再次购买了一台服务器，挂载上另外一台之前保存的镜像，发现此时的大小为212k左右，和没有出问题之前差不多大小，估计这个就是我们需要的东西了，有点运气成分在吧，5台服务器每个索引有两份，直接将这个数据拷贝到目标服务器发现不行(直接就被es删除掉了)
于是在网上搜索到了elasticdump工具，可以通过这个工具将索引导出然后倒入到目标服务器

原服务器：10.30.138.62
目标服务器：10.26.241.237

3.安装elasticdump工具

yum install epel-release -y
 
yum install nodejs npm -y
  
npm install elasticdump -y


4.导出具体配置信息

# node_modules/elasticdump/bin/elasticdump --ignore-errors=true  --scrollTime=120m  --bulk=true --input=http://10.30.138.62:9200/.kibana   --output=data.json  --type=data
Fri, 10 Mar 2017 08:17:45 GMT | starting dump
Fri, 10 Mar 2017 08:17:45 GMT | got 40 objects from source elasticsearch (offset: 0)
Fri, 10 Mar 2017 08:17:45 GMT | sent 40 objects to destination file, wrote 40
Fri, 10 Mar 2017 08:17:45 GMT | got 0 objects from source elasticsearch (offset: 40)
Fri, 10 Mar 2017 08:17:45 GMT | Total Writes: 40
Fri, 10 Mar 2017 08:17:45 GMT | dump complete


 
导出mapping信息
# /root/node_modules/elasticdump/bin/elasticdump --ignore-errors=true  --scrollTime=120m  --bulk=true --input=http://10.30.138.62:9200/.kibana   --output=mapping.json  --type=mapping
Fri, 10 Mar 2017 08:19:11 GMT | starting dump
Fri, 10 Mar 2017 08:19:11 GMT | got 1 objects from source elasticsearch (offset: 0)
Fri, 10 Mar 2017 08:19:11 GMT | sent 1 objects to destination file, wrote 1
Fri, 10 Mar 2017 08:19:11 GMT | got 0 objects from source elasticsearch (offset: 1)
Fri, 10 Mar 2017 08:19:11 GMT | Total Writes: 1
Fri, 10 Mar 2017 08:19:11 GMT | dump complete


5.恢复到目标服务器中


直接在安装了elasticdump工具的服务器上操作，节省在目标服务器上安装工具的时间


导入mapping
# node_modules/elasticdump/bin/elasticdump --input=mapping.json  --output=http://10.26.241.237:9200/.kibana --type=mapping
Fri, 10 Mar 2017 08:23:08 GMT | starting dump
Fri, 10 Mar 2017 08:23:08 GMT | got 1 objects from source file (offset: 0)
Fri, 10 Mar 2017 08:23:08 GMT | sent 1 objects to destination elasticsearch, wrote 7
Fri, 10 Mar 2017 08:23:08 GMT | got 0 objects from source file (offset: 1)
Fri, 10 Mar 2017 08:23:08 GMT | Total Writes: 7
Fri, 10 Mar 2017 08:23:08 GMT | dump complete


导入具体的kibana配置信息


# node_modules/elasticdump/bin/elasticdump --input=data.json  --output=http://10.26.241.237:9200/.kibana --type=data
Fri, 10 Mar 2017 08:23:25 GMT | starting dump
Fri, 10 Mar 2017 08:23:25 GMT | got 40 objects from source file (offset: 0)
Fri, 10 Mar 2017 08:23:25 GMT | sent 40 objects to destination elasticsearch, wrote 40
Fri, 10 Mar 2017 08:23:25 GMT | got 0 objects from source file (offset: 40)
Fri, 10 Mar 2017 08:23:25 GMT | Total Writes: 40
Fri, 10 Mar 2017 08:23:25 GMT | dump complete


总结，通过elasicdump可以方便的备份kibana配置，或是迁移


操作之前一定要测试或者明白操作的结果，不能乱操作，特别是生产环境和自己慌乱的情况下
记录是个好习惯
阿里云的镜像功能不错，可以使用它做镜像策略，例如每天备份一次，数据量大可以2天备份一次，以防万一