-
Notifications
You must be signed in to change notification settings - Fork 25.6k
Description
We recently added a new field type for exponential histograms to elasticsearch. The primary goal of this is to allow the direct usage of histograms ingested via OTLP.
In the EDOT collector, we currently convert all histograms to T-Digest and store them with histogram field type.
In order to allow a transition to the new exponential histogram field type without a breaking change, we need to support the new type in all queryDSL aggregations previously supported by histogram.
- Add support for value_count Add support for value_count for exponential_histogram field #136163
- Add support for sum Add queryDSL sum and average aggregations for exponential histograms #136544
- Add support for average Add queryDSL sum and average aggregations for exponential histograms #136544
- Add support for min
- Add support for max
- Add support for percentiles
- Add support for percentile rank
- Add support for range
- Add support for histogram Add queryDSL histogram aggregation for exponential_histogram fields #136888
- Add support for boxplot
With all the above, we will ensure that queries that previously worked with histogram data will now also work with exponential_histogram data.
However, in order to allow a truly non-breaking transition, we should ensure that queries also work on mixed data.
E.g. if a field in a data stream is changed from histogram to exponential_histogram, queries ideally should still work even if they cover mixed type data.
For value_count, sum, avg, min and max this will work out of the box.
For the other aggregations this will likely require some investigation on how to achieve this.