Index pattern creation UX: Implement wizard flow

[cjcenizal]

Let's implement a creation flow like the Visualize wizard:

1. Select a type of index pattern you want to create. This is an opportunity to surface explanatory text per type, to help the user understand what the options are and make an informed decision.
2. Let the user configure the index pattern by filling out a form.
3. Show the user the new index pattern with a "success" notification.

Note: In the creation process can be simplified to the point where there are no "types" of index patterns, then we can remove the first step from this wizard.

[uboness]

Wizard is too big of a word here...

[uboness]

Before any work is done on this one, we need to have a clearer idea of where we want the whole "index pattern" thing to head in the future (Eg, folding it into a somewhat bigger thing like "data sources")

[debadair]

IMHO, I think this is kind of missing the point. Surfacing the concept of an "index pattern" up front assumes that the user is already familiar with how their data is stored in Elasticsearch. That's a big assumption for someone who's just getting started. We've reached the point where "index pattern" is a completely new concept that many users don't have much context for. We need to make it easier for people who just know that there's data in Elasticsearch they want to explore.

If we can tell them that the index pattern they specify doesn't match any Elasticsearch indices, why can't we simply suggest possible configurations based on the indices that are available? Instead of forcing people to think about the different types of index patterns, why don't we ask them what they know about their data in terms that are more familiar? Is it time-series data? Does each document have a timestamp? Which of these fields contains the timestamp? It looks like your data rolls over to a new index on a daily/weekly/monthly schedule. Do you want to build visualizations based on all of those indices?

At it's heart, Kibana is data-centric. At it's outer edges, it should be more user-centric.

[cjcenizal]

Good points @debadair! When we tackle this issue, I'd like to reassess where we want to go, using your comments and questions as a conversation starter.

[uboness]

@debadair I totally agree with you... and this is part of a larger discussion and vision that we need to align on. This is exactly why I'd defer big changes before we have this point in the horizon where we all feel we should direct the ship to. As part of this, we need to take into account what knowledge we can extrapolate from ES that can help us understand what the user has and what to should suggest. Also, we terminology is also something we can open up for the discussion - What's index pattern? shall we just call it "Data Source"? What else is part of a data source?... there's a lot to discuss, and I don't think we need to figure it all right now, but we should at least be somewhat aligned on where we want to go. There's little point in making big changes to this screen without this alignment IMO.

[Bargs]

Related #7585

[alexfrancoeur]

@uboness @debadair I agree with this idea as well. Introducing the "data source" concept might address some of my comments in the meta ticket. With how easy it is to get Elastic Cloud up and running these days, not all users necessarily do the correct amount due diligence setting up the stack (to be fair, they shouldn't have to). I would imagine choosing or setting up a data source would be the step before configuring an index pattern, regardless of nomenclature. Whatever approach we take needs to be a better experience for new users but we also need to make sure that it is not a hinderance for existing / power users.

Just my two cents. Looking forward to diving into this topic more.

[cjcenizal]

This has been addressed. Closing.