在Elasticsearch中更改类型和重新索引

Question

I recently upgraded my ELK stack (logstash 2.3.4 using redis 3.2.3, Elasticsearch 2.3.5 and Kibana 4.5.4) from (logstash 1.4.1/1.4.2 using redis 2.8.24, Elasticsearch 1.2.2 and Kibana 3.1.1). The upgrade went well but after the upgrade I had some fields that had conflicting types. This specific fields were dynamically created by logstash so there was no overall mapping in Elasticsearch. I spent a fair amount of time searching on how to change this. Every online article stated I couldn't simply change the field type of the existing data. Many articles referenced I needed to reindex but failed to explain how. Below are the exact steps I did to change the type and reindex.

Get the mapping from the current index needing the field type changed:

curl -XGET http://localhost:9200/logstash-2016.05.30/_mapping?pretty=1 > logstash-2016.05.30

Edit the logstash-2016.05.30 file removing the 2nd line (index name) and 2nd last line (curly bracket) in the file. Failing to do this will NOT update the mappings. I suppose if you edit the index name to the new name it would work but I didn't try that (should have tried I guess).

Edit the logstash-2016.05.30 file and change the type (ie long to string or string to long). You can use the exactly definition used by a similar field.

"http_status" : {
  "type" : "string",
  "norms" : {
    "enabled" : false
  },
  "fields" : {
    "raw" : {
      "type" : "string",
      "index" : "not_analyzed",
      "ignore_above" : 256
    }
  }
},

Change to:

"http_status" : {
  "type" : "long"
},

Next create the new index (append _new or whatever you want)

curl -XPUT  http://localhost:9200/logstash-2016.05.30_new -d @logstash-2016.05.30

Double check the mapping was created correctly

curl -XGET http://localhost:9200/logstash-2016.05.30_new/?pretty

Reindex using the following:

curl -XPOST  http://localhost:9200/_reindex -d '{
"source": {
"index" : "logstash-2016.05.30"
},
"dest" : {
"index" : "logstash-2016.05.30_new"
}
}'

Count the entries in both indexes (should be the same)

curl -XGET http://localhost:9200/logstash-2016.05.30/_count
curl -XGET http://localhost:9200/logstash-2016.05.30_new/_count

Once satisfied the reindexing was successfully delete the old index

curl -XDELETE http://localhost:9200/logstash-2016.05.30

Create an alias so the old index name can still be used

curl -XPOST  http://localhost:9200/_aliases -d '{
"actions": [
{ "add": {
"alias": "logstash-2016.05.30",
"index": "logstash-2016.05.30_new"
}}
]
}'

Lastly, Navigate to Kibana and select Settings and the Index Pattern. Click the reload icon to refresh the field list. All conflicts should be removed.

Obliviously, this isn't really a question unless you feel this could be done another way or these is a problem with doing this.

Answer 1

For Elasticsearch 6, a few small changes are required. Otherwise follow all the instructions closely.

To obtain the mapping use pretty=true instead of pretty=1 :

curl -XGET http://localhost:9200/logstash-2016.05.30/_mapping?pretty=true > logstash-2016.05.30

For all XPUT/XPOST requests, the content type must now be set to application/json.

curl -XPUT  http://localhost:9200/logstash-2016.05.30_new \
  -H 'Content-Type: application/json' -d @logstash-2016.05.30