This ticket may improve Nephron performance slightly.
The information contained in the convo_key is also available in separate fields. Nephron already relies on these fields when calculating its aggregations.
Convo-Key can be removed from
- the flow proto-bufs (raw & summary)
- from ES raw flows (needs check if ES can aggregate based on several fields)
- from ES summary flows
When the convo_key is removed from ES then backwards compatibility must be considered. ATM OpenNMS relies on that field because it extracts parts from the string.
In addition, the grouped_by_key of flow summaries also does only contain information that is available in separate fields. It also seems not to be used anywhere.