Duplicati fails to start on new Windows install

Yes, and for any readers who wonder how I got there (maybe you did the same to get it):

Tracing attempt (awaiting information on installation, and DB location and contents) was:

finds that this part was a Duplicati-generated message, and finds the English text for it.
Next, find where the message is in the code:

Next, find out what uses the message:

Next, find that this more explanatory error message is just passed by Duplicati, so try a search engine.
Guess at English translation, or get help guessing from translate.google.com or similar translation tool.
The translation often won’t give the exact English words that message author wrote, but pieces may fit.

Next, use SQL expertise (which I lack sufficiently) to try to see why the exception might have happened.
Maybe the DBNull is also being generated by some intermediate layer that connects C# to SQLite code.

As always, this would be easier to debug (e.g. with a debugger) if it was known how to set up the issue, therefore try to find the installation and run history. Clearing up the issue might just take removing some database, but understanding the issue to avoid it in the future might mean taking close look at database.

Unless the guess below can do it. I would have guessed a bad database, e.g. one not actually initialized

Permission problem can be somewhat ruled in or out by learning whether the user profile is accessible. Technically advanced users can use Sysinternals Process Monitor to actually see attempted DB action.
Somewhat simpler is Sysinternals Process Explorer Find handle or DLL to see what Duplicati used:

Duplicati.GUI.TrayIcon.exe 6784 File C:\ProgramData\Duplicati\duplicati-2.0.5.114_canary_2021-03-10\RUN\Duplicati-server.sqlite

is where I expect my server database to be, because I use –server-datafolder to allow multiple Duplicati.

What I’m trying to find out now is where the Duplicati databases are, and next step may be what’s inside.
DB Browser for SQLite is good to actually look, but some guessing can be made just from very-tiny size.

That theory just faded after some test (sometimes it’s not worth chasing a path without some guidance).
Hiding Duplicati-server.sqlite and starting Duplicati rebuilt it and gave me First run setup, as it should:

image

and trying to trick it with empty Duplicati-server.sqlite was also handled right, so issue cause is unclear.
There are also per-job sqlite databases with randomized names, but I don’t know if a backup was done.