feat: shorter timeout on test_connection

[betodealmeida]

SUMMARY

When we run test_connection when adding a database, the response might error only after 1-2 minutes, depending on the parameters. The error comes back only when the network connection from the driver times out, and because this requests runs in a separate thread and in a 3rd party library we can't adjust that timeout.

We have a @timeout context manager in our code base, but they don't work here because SQLAlchemy runs the ping command in a separate thread. To enforce a shorter timeout on ping I used the func_timeout library, which spins off and monitors its own thread where the target function runs.

I also improved the error message (see screenshots below), and added logic to disable the "test connection" button while the test is in flight. In my initial attempt I added a spinner, but because the main thread is occupied while the test is in flight the spinner doesn't spin.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

Before, note that an invalid connection to druid://google.com:8082/druid/v2/sql/ times out only after ~90 seconds:

After, with TEST_DATABASE_CONNECTION_TIMEOUT set to 2 seconds just for testing:

TESTING INSTRUCTIONS

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 #18001 (ae30904) into master (de3d397) will decrease coverage by 0.94%.
The diff coverage is 66.66%.

❗ Current head ae30904 differs from pull request most recent head b603df8. Consider uploading reports for the commit b603df8 to get more accurate results

@@            Coverage Diff             @@
##           master   #18001      +/-   ##
==========================================
- Coverage   67.07%   66.12%   -0.95%     
==========================================
  Files        1609     1569      -40     
  Lines       64903    61641    -3262     
  Branches     6866     6233     -633     
==========================================
- Hits        43535    40762    -2773     
+ Misses      19502    19280     -222     
+ Partials     1866     1599     -267     

Flag	Coverage Δ
hive	?
javascript	50.86% <66.66%> (-2.91%)	⬇️
mysql	82.07% <ø> (-0.12%)	⬇️
postgres	82.12% <ø> (-0.12%)	⬇️
presto	?
python	82.21% <ø> (-0.48%)	⬇️
sqlite	81.81% <ø> (-0.12%)	⬇️

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

Impacted Files	Coverage Δ
...-chart-controls/src/sections/advancedAnalytics.tsx	33.33% <ø> (ø)
...s/legacy-plugin-chart-country-map/src/countries.ts	100.00% <ø> (ø)
...legacy-plugin-chart-partition/src/controlPanel.tsx	25.00% <ø> (ø)
...gins/legacy-plugin-chart-rose/src/controlPanel.tsx	50.00% <ø> (ø)
.../plugins/legacy-preset-chart-nvd3/src/Bar/index.js	66.66% <ø> (ø)
...gins/legacy-preset-chart-nvd3/src/DistBar/index.js	66.66% <ø> (ø)
...ins/legacy-preset-chart-nvd3/src/DualLine/index.js	66.66% <ø> (ø)
...gins/legacy-preset-chart-nvd3/src/NVD3Controls.tsx	95.83% <ø> (ø)
...plugins/legacy-preset-chart-nvd3/src/ReactNVD3.jsx	0.00% <ø> (ø)
...c/SqlLab/components/TemplateParamsEditor/index.tsx	75.00% <ø> (-9.00%)	⬇️
... and 538 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 de3d397...b603df8. Read the comment docs.

[john-bodley]

@betodealmeida out of interest in a slow ping problematic? I gather adding and thus testings a database connection—except if it is used for offline monitoring—is somewhat of a rare event.

Additionally are there specific database engines which may take minutes to ping? Asking to learn as the default SQLAlchemy ping is merely a SELECT 1 per here.

[betodealmeida]

@betodealmeida out of interest in a slow ping problematic? I gather adding and thus testings a database connection—except if it is used for offline monitoring—is somewhat of a rare event.

We saw a lot of errors when people are trying to setup their databases, with a lot of people simply giving up. Because of that we're trying to make the process less error-prone, or at least quicker.

Additionally are there specific database engines which may take minutes to ping? Asking to learn as the default SQLAlchemy ping is merely a SELECT 1 per here.

When the DB connects successfully the ping is fast. The problem is that if the URL is incorrect it might take a long time for the test to fail, which only happens when the Superset <--> DB connection times out. If it takes more than 30 seconds to run SELECT 1 it's better to fail early, IMHO, because the SQLAlchemy URI is probably incorrect.