Sizing indexes is a tedious task, there are a few stats that are missing that would be extremely useful to expedite the index sizing process:
- avg_key_size: There is a need to know the average doc_id size. Currently there is a stat for just this is a total of the doc_id+sec_bytes.
- avg_sec_bytes: There is a need to know the average total secondary bytes size. Currently there is a stat for just this is a total of the doc_id+sec_bytes.
- avg_array_size: Sizing arrays is almost impossible based on currently available stats. For array indexes, we need to know the average length of the array.
- avg_array_bytes: This goes with the above, we need to know the average size of each array element.
- percent_of_docs: This is the projection selectivity or the percentage of documents in the index vs that of documents in the bucket. This can be calculated by dividing the index item_count / curr_items from the bucket stats but that only works for non-array indexes.