Control your Elasticsearch document updates with extra speed.

• So, Elasticsearch provides update API's like _update and _bulk which helps us in updating elasticsearch document. But elasticsearch updates the document by merging the given [NEW] document with the [OLD] document, which leads to a case where

  • The older documents fields which are not present in the new given document will be retained, so if you want delete a field you have to explicitly send a update query with a script. And everyone knows that elasticsearch scripts are slower (eg: to delete a field from 13 million documents it will take 50 min in 5 node cluster)

  eg : Old Document

       {
           "a": 23,
           "b": 40
       }
  _New Document_ : /_update api
  
       {
           "a": 19,
           "c": 45
       }
  
  Now If I send this to elasticsearch the result output document will be :
  
       {
           "a": 19,
           "b": 40,
           "c": 45
       }
  
  So the field "b"is retained, but my use case is to remove the "b", as I am not sending it in the request.
  

  • The Advance Plugin can do this for you.

  eg : Old Document

       {
           "a": 23,
           "b": 40
       }
  _New Document_ : /_advanceupdate api
  
       {
           "a": 19,
           "c": 45
       }
  
  Now If I send this to elasticsearch the result output document will be :
  
       {
           "a": 19,
           "c": 45
       }
  

  • Elasticsearch Updates first update documents in the primary shard and then it updates the replica shards, now that increased the total time required for update of document.
  • You can reduce the total time by setting the replica to -1, but if you have a lot of documents which takes 1 hour to get updated and in meanwhile your nodes goes down you dont have a protection of the replicas to recover i.e, you loose your data.
  • Advance-Update gives you the functionality where your old data is safe and speed for updating documents is much higher than the normal updates.

• POST _advanceupdate

  usage :

   /_update
  
   {
       "a": 19,
       "c": 45
   }
  

• POST/PUT _advancebulk

   /_advancebulk
  
    { "update" : {"_id" : "2", "_type" : "type1", "_index" : "test"} }
    { "doc" : {"b": 12,"f": 14,"m":15}, "doc_as_upsert" : true}
    { "update" : {"_id" : "2", "_type" : "type1", "_index" : "test"} }
    { "doc" : {"s": 15,"l": 14,"k":12}, "doc_as_upsert" : true}
  

• Elasticsearch 5.6.0
• Change the thread_pool.bulk.queue_size to high enough so that your documents don't get skipped or else you can keep it to -1 in your elasticsearch.yml file.

Download and install elasticsearch 5.6.0 from

• Elasticsearch

Download the latest Advance Update plugin from

• Releases

Install the plugin :

    <ES directory>/elasticsearch-plugin install file:///Downloads/advance-update.zip

Test the installation by visiting https://localhost:9200/_cat/plugins

Works with only elasticsearch 5.6.0

• Configure copying elasticsearch documents from primary shard to replicas
• Updating a key in all the documents.

• Viraj Parab - Initial work - CrazyCompiler