I have a common use case of summarizing event-based indices into session-based indices. I currently achieve this by fetching data from elasticsearch to a compute environment, summarizing the data by session and then ingesting the result into a summary index. When this analysis involves high-frequency events, it becomes prohibitively expensive to transfer and process large volumes of events.
It would be great if the output of an aggregation could be ingested as a new index without having to leave the cluster. I had expected that the reindex api could be used for this type of job, but was dismayed that the source query does not support aggregations.
BTW - I’ve seen notes that suggests the xpack watcher api can be repurposed for this type of job. It also looks like the rollup api can do something similar but only supports a limited set of summary statistics.
From an api standpoint, I would have expected es to support a general purpose “reduce” api that works with arbitrary pipeline aggregations.
Pinging @elastic/es-search-aggs
@bitkowitz We discussed this feature request in FixitFriday and would like to ask you to share more information about your use-case. In particular, what would the aggregation look like in your case, and what kind of post-processing would you need to perform?
Leaving the discuss label so that it will be discussed again when you share more information.
No additional feedback, closing
It's too bad this was closed as I have a use case for this feature and was surprised and disappointed to find out aggregations are not supported by _reindex or by the Logstash Elasticsearch input plugin. I can provide more details if anyone is interested.
Most helpful comment
It's too bad this was closed as I have a use case for this feature and was surprised and disappointed to find out aggregations are not supported by _reindex or by the Logstash Elasticsearch input plugin. I can provide more details if anyone is interested.