I have added the following note to the documentation to indicate that you should use the UTF-8 equivalent character type and collation when running PostgreSQL on Windows.

I will leave this issue open, and update the description to make it about improving the upgrade check to recognise this collation.

Well I currently run production confluence on postgres 9.6 with the collation 'C' because the postgres server I use offers nothing else (except for posix, which is an alias to 'C'). 

What are the current issues with the 'C' collation? Any pointers, where I can look?

Found it: https://confluence.atlassian.com/doc/troubleshooting-character-encodings-167194.html

And the issue is: Some characters are not available as upper cases, so postgres keeps them in lower cases.

Hey, I'm one of the devs working on this. To give y'all an update:

• utf8 is incorrectly reporting as being invalid as per CONFSERVER-58052. I updated that case, if you have utf8 you're definitely supported, no dramas. I know some folks are ending up here from that problem, hence me leaving a comment about it.
• The windows-1252 locale is unfortunately not UTF-8, it's a Windows special extension of ISO-8859-1 (aka LATIN1) with different entries. It's also single byte, not multi-byte as per UTF-8, so characters after 127 (hex 0x7F) have different representations as UTF-8 has 2 bytes and 1252 only has 1. https://www.i18nqa.com/debug/utf8-debug.html shows us 128 - 255 and how they can differ. This can cause problems as per below:

  postgres=# select * from (values ('€'), ('‰')) as example;
   column1
  ---------
   ?
   %
  (2 rows)
  

  Because 1252 has no character representations for those certain characters it'll either ? or try to match to the closest.

• PostgreSQL is fairly 'magical' in that it has automatic character set conversion as in https://www.postgresql.org/docs/9.6/multibyte.html (23.3.3) which will translate 1252 to UTF-8, provided the client encoding is set correctly. For the PostgreSQL JDBC driver it does this by default and doesn't like it if you change it.
• Encoding in the database is often fixed as UTF-8, however collate and character type in PostgreSQL will have an impact on ordering, and also string operations such as initcap(), upper(), lower() and so on, as PostgreSQL uses system-level implementations. This means they definitely make a difference so just looking at encoding unfortunately won't 100% guarantee things are working as expected.
• When PostgreSQL is installed it uses whichever default locale is configured, and the 'cluster' limits databases to be created under that (unless you use a separate template). This makes changing the locale a bit finnicky but it's possible, provided the locale exists. If it doesn't exist, it'll error when trying to create a database without one.
• As UTF-8 doesn't fully exist on Windows currently (Server 2019 only has a Beta implementation of it), we can't set UTF-8 collation
• When I tested PostgreSQL 11 on 2019 with UTF-8 it also wasn't implemented in PostgreSQL yet
• C is suggested against completely as it sorts differently, as in https://confluence.atlassian.com/doc/database-setup-for-postgresql-173244522.html. Yes Jira is OK with it, Confluence doesn't like it though.

At this current state we're looking at verifying if 1252 is OK and if so will add it to the white list. We're also fixing CONFSERVER-58052 at the same time. Will keep you updated when we have some progress, thanks so much for your patience while we get this fixed up for everyone!

I think that the test for postgres should check for the "ENCODING" of the database. 

I use collation "C" with ENCODING "utf-8" on Postgres. 

Which, by the way, works perfect with jira. Will test with latest confluence.

Here are the default postgres settings that I have for the database:

       ENCODING = 'UTF8'
       TABLESPACE = pg_default
       LC_COLLATE = 'C'
       LC_CTYPE = 'C'
       CONNECTION LIMIT = -1; 

By the way: Using the default postgres tar.gz installation, you only get collations "C" and "posix" where posix is another name for "C"....

After updating my docker setup to 6.15.1 I receive the following error, too. Which I find seems to be weird because it is utf8, isn't it?

Database: The database collation 'en_US.utf8' is not supported by Confluence. You need to use 'utf-8'.

RUn into the same problem after updating to latest version (btw - i did the update after your warning email a few days ago)!
What is ist with you? The Problem exists for a while and not even a word from anybody@Atlassian?!?!?
Shame!

I just upgraded to the latest version due to the two vulnerabilities as well and received this error too

@IT Department, same here. Just upgraded today after recent Atlassian announcement of two vulnerabilities.

Warning shows:

Hi, Recently i upgraded Jira software from 7.12 to 8.0.2. After upgraded i got some warning message like "Your database is using an unsupported collation: English_United.1252 but i checked postgres db encoding is UTF-8. It's running in windows. I just ignored and its working fine. I am little worried for future. Any this need to change from our side. Please let me know.

.