Indexing with ?require_alias=true should create missing data streams

[DaveCTurner]

The ?require_alias=true parameter to the indexing APIs is there to let users block indexing before the cluster is properly configured, particularly preventing auto-creation of an index if the intended target was an alias which points to a different index. However, today it also prevents auto-creation of a data stream, even if there is a template installed which shows how the data stream should be created. If there's an appropriate template then the cluster is properly configured, so IMO we should auto-create the missing data stream when processing an indexing request with ?require_alias=true.

[elasticsearchmachine]

Pinging @elastic/es-data-management (Team:Data Management)

[DaveCTurner]

Andrei rightly points out that data-stream aliases make this a little more complicated. For starters the name ?require_alias suggests a data-stream alias might be required, and it's possible that different users would want different behaviour in that regard. But also we could in principle satisfy "requires a data-stream alias" by creating one of several different data streams each with the same alias in their template. Although it seems invalid for all those templates to specify that their alias is the write alias, we could reject in that case.

Perhaps we need to expand the parameter to have more values than just ?require_alias and ?require_alias=true and ?require_alias=false. Or perhaps we do need another parameter after all.