elasticsearch 索引 red 状态恢复 green

方案一

找到状态为 red 的索引

curl -X GET "http://172.xxx.xxx.174:9288/_cat/indices?v="

red    open   index                          5   1    3058268        97588      2.6gb          1.3gb

状态为 red 是无法对外提供服务的，说明有主节点没有分配到对应的机子上。

找到 UNASSIGNED 节点

_cat/shards 能够看到节点的分配情况

curl -X GET "http://172.xxx.xxx.174:9288/_cat/shards"

index                            shard prirep state        docs   store   ip             node         
index                      1    p     STARTED     764505 338.6mb 172.xxx.xxx.174 Calypso      
index                      1    r     STARTED     764505 338.6mb 172.xxx.xxx.89  Savage Steel
index                      2    p     STARTED     763750 336.6mb 172.xxx.xxx.174 Calypso      
index                      2    r     STARTED     763750 336.6mb 172.xxx.xxx.88  Temugin      
index                      3    p     STARTED     764537 340.2mb 172.xxx.xxx.89  Savage Steel
index                      3    r     STARTED     764537 340.2mb 172.xxx.xxx.88  Temugin      
index                      4    p     STARTED     765476 339.3mb 172.xxx.xxx.89  Savage Steel
index                      4    r     STARTED     765476 339.3mb 172.xxx.xxx.88  Temugin      
index                      0    p     UNASSIGNED                                             
index                      0    r     UNASSIGNED    

index 有一个主节点 0 和一个副本 0 处于 UNASSIGNED 状态，也就是没有分配到机子上，因为主节点没有分配到机子上，所以状态为 red。
从 ip 列可以看出一共有三台机子，尾数分别为 174，89 以及 88。一共有 10 个 index 所以对应的 elasticsearch 的 index.number_of_shards: 5，index.number_of_replicas: 1。一共有 10 个分片，可以按照 3，3，4 这样分配到三台不同的机子上。88 和 89 机子都分配多个节点，所以可以将另外一个主节点分配到 174 机子上。

找出机子的 id

找到 174 机子对应的 id，后续重新分配主节点得要用到

curl -X GET "http://172.xxx.xxx.174:9288/_nodes/process?v="
{
  "cluster_name": "es2.3.2-titan-cl","nodes": {
    "Leivp0laTYSqvMVm49SulQ": {
      "name": "Calypso","transport_address": "172.xxx.xxx.174:9388","host": "172.xxx.xxx.174","ip": "172.xxx.xxx.174","version": "2.3.2","build": "b9e4a6a","http_address": "172.xxx.xxx.174:9288","process": {
        "refresh_interval_in_millis": 1000,"id": 32130,"mlockall": false
      }
    },"EafIS3ByRrm4g-14KmY_wg": {
      "name": "Savage Steel","transport_address": "172.xxx.xxx.89:9388","host": "172.xxx.xxx.89","ip": "172.xxx.xxx.89","http_address": "172.xxx.xxx.89:9288","id": 7560,"tojQ9EiXS0m6ZP16N7Ug3A": {
      "name": "Temugin","transport_address": "172.xxx.xxx.88:9388","host": "172.xxx.xxx.88","ip": "172.xxx.xxx.88","http_address": "172.xxx.xxx.88:9288","id": 47701,"mlockall": false
      }
    }
  }
}

174 机子对应的 id 为 Leivp0laTYSqvMVm49SulQ。

为了简单也可以直接将该主分片放到 master 机子上，但是如果节点过于集中肯定会影响性能，同时会影响宕机后数据丢失的可能性，所以建议根据机子目前节点的分布情况重新分配。

curl -X GET "http://172.xxx.xxx.174:9288/_cat/master?v="
id                     host          ip            node         
EafIS3ByRrm4g-14KmY_wg 172.xxx.xxx.89 172.xxx.xxx.89 Savage Steel

分配 UNASSIGNED 节点到机子

得要找到 UNASSIGNED 状态的主分片才能够重新分配，如果重新分配不是 UNASSIGNED 状态的主分片，例如我视图重新分配 shard 1 会出现如下的错误。

curl -X POST -d '{
    "commands" : [ {
      "allocate" : {
          "index" : "index","shard" : 1,"node" : "EafIS3ByRrm4g-14KmY_wg","allow_primary" : true
      }
    }]
}' "http://172.xxx.xxx.174:9288/_cluster/reroute"

{
  "error": {
    "root_cause": [
      {
        "type": "remote_transport_exception","reason": "[Savage Steel][172.xxx.xxx.89:9388][cluster:admin/reroute]"
      }
    ],"type": "illegal_argument_exception","reason": "[allocate] failed to find [index][1] on the list of unassigned shards"
  },"status": 400
}

重新分配 index shard 0 到某一台机子。_cluster/reroute 的参数 allow_primary 得要小心，有概率会导致数据丢失。具体的看看官方文档该接口的说明吧。

curl -X POST -d '{
    "commands" : [ {
      "allocate" : {
          "index" : "index","shard" : 0,"node" : "Leivp0laTYSqvMVm49SulQ","allow_primary" : true
      }
    }]
}' "http://172.xxx.xxx.174:9288/_cluster/reroute"

{
  "acknowledged": true,.........
  "index": {
    "shards": {
      "0": [
        {
          "state": "INITIALIZING","primary": true,"node": "Leivp0laTYSqvMVm49SulQ","relocating_node": null,"shard": 0,"index": "index","version": 1,"allocation_id": {
            "id": "wk5q0CryQpmworGFalfWQQ"
          },"unassigned_info": {
            "reason": "INDEX_CREATED","at": "2017-03-23T12:27:33.405Z","details": "force allocation from previous reason INDEX_REOPENED,null"
          }
        },{
          "state": "UNASSIGNED","primary": false,"node": null,"unassigned_info": {
            "reason": "INDEX_REOPENED","at": "2017-03-23T11:56:25.568Z"
          }
        }
      ]
      }
    }
    .............
}

输出结果只罗列出了关键部分，主节点处于 INITIALIZING 状态，在看看索引的状态

curl -X GET "http://172.xxx.xxx.174:9288/_cat/indices?v="

green  open   index                          5   1    3058268        97588      2.6gb          1.3gb

索引状态已经为 green，恢复正常使用。

以上参考 ELASTICSEARCH几个问题的解决

方案二

导致集群变red，很可能是因为集群中有机子宕机了，其中一部分数据没有同步完成，因此将之前宕机的机子起来，和现有集群同步完成，集群也就恢复了。
另外也可以找一台空的机子，与现有的机子组成集群，索引会自动平衡，如果集群没有数据丢失，也是可以将集群恢复正常。

欢迎转载，但请注明本文链接，谢谢你。
2017.3.24 12:15