Fixing IllegalArgumentException in Elasticsearch

In the following post, we are going to see how to solve an error that I found in an application that is using Elasticsearch as storage.

What is Elasticsearch?

Elasticsearch is a non-relational database that is powerful for doing searches by text. The definition from wikipedia:

Elasticsearch is a search engine based on the Lucene library. It provides a distributed, multitenant-capable full-text search engine with an HTTP web interface and schema-free JSON documents

I have used it to store millions of documents for things like storing prices searched.

Error found in the logs

Some days ago I found the following error in the logs of the application that I was working on:

Caused by: ElasticsearchException[Elasticsearch exception [type=illegal_argument_exception, reason=[half_float] supports only finite values, but got [94433.46]]]

According to the elasticsearch documentation the maximum value that can have a half_float is 65500, so if we try to insert a document with a value higher than the maximum allowed we will receive the previous error.

Index with the wrong mapping

Looking at the mappings of the existing index I could see that there were fields with this half_float

"opnl_flights_ow_v2": {
    "mappings": {
      "_doc": {
        "properties": {
          "prices": {
            "properties": {
              "PRICE": {
                "properties": {
                  "amount": {
                    "type": "half_float",
                    "index": false
                  },
                  "currency": {
                    "type": "keyword",
                    "index": false
                  }
                }
              }

How to fix the error

We can update an index template of an existing index. With the execution of the following command, we can update the property amount setting his type to float instead of the half_float that was generating the error.

curl -XPUT 'http://localhost:9200/{name_of_the_index}/_doc/_mapping' -H 'Content-Type: application/json'  -d '
{
    "_doc" : {
        "properties" : {
            "amount" : {"type" : "float", "store" : true}
        }
    }
}

In our case, we don’t need to reindex the documents into a new index because we are only setting a type that is larger than the previous one. It would be different that we change i.e: from String to date, that would need reindexing. of the data from the old index to the new index.

Conclusion

In this post, we have seen how to fix a half_float problem that we had in our Elasticsearch index where we were storing prices that were higher than the half_float maximum value.