Amazon OpenSearch Serverless - Index Configuration, Stemmer Override Issue

0

Hi, We are considering migration to the OpenSearch Serverless from the managed domains. However, our index configuration is quite big and consists of multiple mappings, and analyzers, including the stemmer overriding.

Sadly, after thorough investigation we have found out that OpenSearch Serverless doesn't work with any overriden stemmer. I have prepared simple MVP showcasing the issue. Index settings configuration:

{
    "settings": {
        "analysis": {
            "analyzer": {
                "my_analyzer": {
                    "tokenizer": "standard",
                    "filter": [
                        "lowercase",
                        "custom_stems",
                        "porter_stem"
                    ]
                }
            },
            "filter": {
                "custom_stems": {
                    "type": "stemmer_override",
                    "rules": [
                        "running, runs => run",
                        "stemmer => stemmer"
                    ]
                }
            }
        }
    },
    "mappings": {
        "properties": {
            "textfield": {
                "type": "text",
                "analyzer": "my_analyzer"
            }
        }
    }
}

After such index is created, none of the documents can be added. After adding even the simplest

{
    "textfield": "any value"
}

The following error is returned after around 11s:

{
    "error": {
        "root_cause": [
            {
                "type": "exception",
                "reason": "Internal error occurred while processing request"
            }
        ],
        "type": "exception",
        "reason": "Internal error occurred while processing request"
    },
    "status": 500
}

Can someone please clarify whether the stemmer override analyzer works on the amazon opensearch serverless? As far as I know we cannot have access to any logs created by the serverless, so we cannot investigate it more. It is actually quite a blocker for us, as we would have to adjust the configuration of our current index working perfectly fine on the managed domains. Thanks in advance! Hubert

已提問 8 個月前檢視次數 166 次
沒有答案

您尚未登入。 登入 去張貼答案。

一個好的回答可以清楚地回答問題並提供建設性的意見回饋,同時有助於提問者的專業成長。

回答問題指南