chore: remove druid datasource from the config

[eschutho]

In #19262 I added to the UPDATING.md file that the native druid datasource was going to be deprecated in 2.0. After further consideration, I think it will also be more explicit and better for semver to actually remove the config in this major release.

TESTING INSTRUCTIONS

Native Druid NoSQL tests need to be removed.

ADDITIONAL INFORMATION

• Has associated issue:
• Required feature flags:
• Changes UI
• Includes DB Migration (follow approval process in SIP-59)
  • Migration is atomic, supports rollback & is backwards-compatible
  • Confirm DB migration upgrade and downgrade tested
  • Runtime estimates and downtime expectations provided
• Introduces new feature or API
• Removes existing feature or API

[codecov]

Codecov Report

Merging #19770 (386580d) into master (fdf48c6) will decrease coverage by 0.27%.
The diff coverage is 70.00%.

@@            Coverage Diff             @@
##           master   #19770      +/-   ##
==========================================
- Coverage   66.54%   66.27%   -0.28%     
==========================================
  Files        1714     1712       -2     
  Lines       65102    63964    -1138     
  Branches     6725     6725              
==========================================
- Hits        43321    42389     -932     
+ Misses      20069    19863     -206     
  Partials     1712     1712              

Flag	Coverage Δ
hive	53.65% <60.00%> (+0.75%)	⬆️
mysql	81.98% <70.00%> (+0.01%)	⬆️
postgres	82.02% <70.00%> (+<0.01%)	⬆️
presto	53.50% <60.00%> (+0.74%)	⬆️
python	82.45% <70.00%> (+0.01%)	⬆️
sqlite	81.77% <70.00%> (+<0.01%)	⬆️
unit	48.63% <60.00%> (+0.57%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...tend/src/visualizations/FilterBox/controlPanel.jsx	37.50% <ø> (ø)
superset/cli/update.py	0.00% <ø> (ø)
superset/config.py	91.33% <ø> (-0.09%)	⬇️
superset/models/dashboard.py	75.96% <ø> (+1.06%)	⬆️
superset/utils/dict_import_export.py	77.27% <ø> (+2.27%)	⬆️
superset/views/core.py	77.37% <ø> (+0.29%)	⬆️
superset/datasets/commands/importers/v0.py	91.47% <50.00%> (-2.15%)	⬇️
superset/dashboards/commands/importers/v0.py	91.87% <100.00%> (+1.93%)	⬆️
superset/initialization/__init__.py	91.39% <100.00%> (-0.33%)	⬇️
superset/security/manager.py	95.45% <100.00%> (+0.18%)	⬆️
... and 10 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fdf48c6...386580d. Read the comment docs.

[john-bodley]

I thought for Superset 2.0 we were just going to mark this as deprecated and actually remove it in Superset 3.0? This change could be seen as too aggressive for some institutions and ideally we should strive not to unnecessarily churn users.

[eschutho]

I thought for Superset 2.0 we were just going to mark this as deprecated and actually remove it in Superset 3.0? This change could be seen as too aggressive for some institutions and ideally we should strive not to unnecessarily churn users.

I think the original plan was to remove native Druid support for SIP68, and in trying to get ahead of that work, to mark it as "no longer supported" in 2.0. I believe there is likely going to be something coming before 3.0 that will break this functionality. Do you think that we need to give people more time to switch over to Druid Sql? If absolutely necessary, we can try not to break native Druid as part of the SIP68 work, but I think it's going to be complicated. Or we just hold off on finishing the SIP68 work until 3.0. cc @betodealmeida

I figured the only reason that we didn't actually remove the functionality in 2.0 was the time it would take to write this PR.

[villebro]

I thought for Superset 2.0 we were just going to mark this as deprecated and actually remove it in Superset 3.0? This change could be seen as too aggressive for some institutions and ideally we should strive not to unnecessarily churn users.

@john-bodley I understand the concern here, but at the same time the community has had ample time to prepare for this (link to the SIP #6032 ). Maybe we could float this via the dev list - if there's strong opposition maybe we can postpone until 3.0, but otherwise just rip the bandaid off in 2.0?

My personal opinion: the native Druid connector is already missing some key functionality, so I'm actually surprised if it's even possible to use it in a production setting anymore. So I'd vote for removing it in 2.0 or at least making it abundantly clear that it's no longer maintained and may stop functioning all together at any given point after 2.0.

[betodealmeida]

This is awesome! I left a few minor comments.