Unexpected pri.store.size discrepancy for files indexed in ES2x and ES5x


(Divya Malini) #1

Hi,

We are seeing some noticeable store size discrepancy for documents we indexed in 2 ES versions
In ES version 2.4.1, we indexed docs with type "Mapping1". It has just field1 and _source is ON

{
_ "Mapping1": {_
_ "all": {
_ "enabled": false_
_ },_
_ "properties": {_
_ "field1": {_
_ "type": "text",_
_ "norms": false,_
_ "index_options": "offsets",_
_ "analyzer": "customanalyzer1"_
_ }_
_ }_
_ }_
}

In ES version 5.4.1, we indexed the same docs with type "Mapping2". It has field1, field2, field2.reverse and _source is ON.

{
_ "Mapping2": {_
_ "all": {
_ "enabled": false_
_ },_
_ "properties": {_
_ "field1": {_
_ "type": "text",_
_ "norms": false,_
_ "index_options": "offsets",_
_ "analyzer": "customanalyzer1"_
_ },_
_ "field2": {_
_ "type": "text",_
_ "norms": false,_
_ "index_options": "offsets",_
_ "fields": {_
_ "reverse": {_
_ "type": "text",_
_ "norms": false,_
_ "index_options": "offsets",_
_ "analyzer": "reverseanalyzer"_
_ }_
_ },_
_ "analyzer": "regexanalyzer"_
_ }_
_ }_
_ }_
}

the store size for both the indices is more or less the same. Given that in the new mapping we have introduced additional fields my expectation was our store size should by minimum have been doubled{the terms in field1 and field2 is mostly the same and there is field2.reverse additionally}.

I ran exists query against all 3 fields in the new index. The doc count shows all the fields are present in all the docs as expected.

POST /_count
{
"query": {
"exists": {
"field": "FIELD_NAME"
}
}
}

But I see the pri.store.size difference is very negligible in between these versions ? Could anyone please help understand this case better or give some pointers into debugging the same ? Is there something I missed to understand?

Thanks, Divya


(Divya Malini) #2

Anyone faced such scenario in the past ? Any input to get to the root cause would be very helpful.


(system) #3

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.