Too_many_buckets_exception

[Pieter_Agenbag]

Hi... I have a too_many_buckets_exception problem. I have read some of the other related questions and although the composite aggregation did help to a certain degree ... it does not seem to support auto_date_histogram

My query ..seen below , does a simple terms aggregation (and in this case its limited to a single "attribute" term) and then sub-aggregated on auto_date_histogram and finally a min metric aggregation.
Sure I'm querying 180 indexes , but seeing how I only have 1 "Attribute" term and specifying only 20 buckets to the auto_date_historgram , i didnt expect it to be a problem ?

From what I understand of how the auto_date_histogram works (and its not a lot) ...it merges buckets at collection time , from smaller into larger buckets to maintain the "target amount" of buckets ... which in turn suggested to me that the number of indexes /shards should not play a major role in the equation. Or would it still roll up the buckets per shard and then collate them at the end ?

Regardless... any suggestions then how to get auto_date_histogram working across a large number of indices ? (a composite aggregation with fixed calendar_interval does work ..but adds additional challenges and will end up requiring additional queries)

P.S. I'm on ES 7.5

GET performance.graph-*/_search
{
  "size": 0,
  "aggs": {
    "attribute": {
      "terms": {
        "field": "attribute"
      },
      "aggs": {
        "datehisto": {
          "auto_date_histogram": {
            "field": "timestamp",
            "buckets": 20
          },
          "aggs": {
            "min": {
              "min": {
                "field": "value"
              }
            }
          }
        }
      }
    }
  },
  "query": {
    "term": {
      "attribute": 68964466
    }
  }
}