Get all distincts values of a field ( more than 10k values)

[humartinez]

Hi there,
Perhaps this question was asked many times but I haven't seen an answer that fits me. Im looking for the best way to get all the distincts values of a field in a group of indices and I manage to create a script in order to do this but It doesn't feel right to me so Im asking help form the experts.

partitions=25
rm run*
for (( i=0;i<$partitions;i++))
do
  curl -s -u user:password 'http://10.x.x.x:9200/index-*'/_search?pretty -H 'Content-Type: application/json' -d"
  {
     \"size\": 0,
     \"aggs\": {
        \"expired_sessions\": {
           \"terms\": {
              \"field\": \"data.device.deviceid\",
              \"include\": {
                 \"partition\": $i,
                 \"num_partitions\": $partitions
              },
              \"size\": 10000
           }
        }
     }
  }
  " > run.$i
done
cat run*|jq .aggregations.expired_sessions.buckets[].key

In fact the number of devices is differs when I run a cardinality query

GET index-*/_search
{
  "size": 0,
  "aggs": {
    "count": {
      "cardinality": {
        "field": "data.device.deviceid"
      }
    }
  }
}

The cardinality query retuns 62102 vs the term agg that returns 61990

[Mark_Harwood]

A couple of points:

1. the cardinality aggs is, by design, approximate - see the docs
2. if you don’t need to sort the terms by a child agg the ‘composite’ agg is probably simpler than using the ‘terms’ aggregation with partitioning

[humartinez]

Colud you please give me an example of composite agg to do this, I can't figure It out

[Mark_Harwood]

The after param is what allows you to page the composite agg.