After reindex, new index has same documents but data size is much less

[Rosita_Andrade]

Hi,

This seems to be a similar case as After reindex, new index has same documents but data size is much less.

I used the Reindex API and my new index is significantly smaller in size. I have a feeling, that it might be because of the mapping and deletion requests.

The source ES index is version 5.6.11. The destination index is 7.11.2

The only changes I have done in the mapping are the following:

1. Changed the mapping of one of the fields from

"mapping": {
                "analyzer": "autocomplete",
                "type": "keyword"
              }

to

"mapping": {
                "analyzer": "autocomplete",
                "type": "text"
              }

because the new version would complain about deprecation.

2. In the index settings I had

     "analyzer": {
            "reverse": {
              "filter": [
                "reverse"
              ],
              "type": "custom",
              "tokenizer": "standard"
            },
            "autocomplete": {
              "filter": [
                "standard",
                "lowercase",
                "stop",
                "kstem",
                "edge_ngram"
              ],
              "type": "custom",
              "tokenizer": "standard"
            },

I removed the "standard" filter cause the new version does not allow me to create mapping with this one.

My question is can this small change in mapping, can have significant difference in the size of the index despite the same number of documents being indexed?

Thanks!

[Rosita_Andrade]

Ok, following up with a finding. Even if I use the deprecated version for the analyser on keyword, the index size is still significantly lower.

[warkolm (Mark Walkom)]

Welcome to our community!

@Rosita_Andrade

My question is can this small change in mapping, can have significant difference in the size of the index despite the same number of documents being indexed?

It's possible, you are coming from a pretty old version to one that has had substantial work done on things like this.

[Rosita_Andrade]

Alright, Thank you very much for your response!