Hard reload of dashboard required to see the incoming data when they did not initially exist

[MichaelKatsoulis]

Currently when opening a dashboard when there are no data streams yet, then all visualizations show the missing field error.
Example from Kubernetes dashboard where the integration is installed but agent is not running yet.

When agent starts running and data start flowing in elasticsearch, the data still do not appear in the visualizations Eno matter how many minutes pass. Auto/Manual refresh have no impact at all.

The only way to see the data and fill the dashboard is by reloading the whole dashboard page via the browser.

This can be very confusing for a user.
After discussions it seems that this is expected and the reason is that a refresh of the dashboard (either by manually or due to auto-refresh) only fetches the data again, but not the field lists for each involved data view as this would be very wasteful.

A suggested improvement could probably involve special handling of the "no fields" case, so a refresh also reloads the field list (at a potentially high cost for CCS data views).

[elasticmachine]

Pinging @elastic/kibana-visualizations @elastic/kibana-visualizations-external (Team:Visualizations)

[elasticmachine]

Pinging @elastic/kibana-data-discovery (Team:DataDiscovery)

[elasticmachine]

Pinging @elastic/kibana-presentation (Team:Presentation)

[nreese]

Maps has a similar issue open about reloading data views for data streams to fetch new fields, #143877

[mattkime]

A suggested improvement could probably involve special handling of the "no fields" case, so a refresh also reloads the field list (at a potentially high cost for CCS data views).

This is a solid plan. Also, CCS performance issues have been fixed. Let me know if there's anything I can do to help from a data views perspective.

[ThomThomson]

Removing the Presentation team label from this as it looks like this will be accomplished at the Data View level.

[markov00]

The error messaging will partially be solved by #151526
Currently there is no clean way to detect when a field is populated with the first data point except for polling index mappings until all fields are available in the dashboard

[markov00]

We are wondering why during the integration asset setup/installation step we are not updating the index mappings with the required fields used by the dashboards. If we did that, instead of using dynamic mappings when the first document containing that field reaches ES we could fix that problem at the root level.
Is there any reason why we don't update the index mapping together with the installation of an integration asset?