Problem with cyrillic links

[Main User]

Hello.

We're trying to migrate from 6.4.0 to 6.15.2 using full xml backup. Articles and attachments were imported correctly, but all links containing cyrillic characters are broken. When editing an article, you can easily insert a link to any of those articles, but after saving it becomes broken too.

OS: CentOS 7
DB: PostgreSQL 9.6.3

DB params are fine:

select datname,pg_encoding_to_char(encoding) as encoding,datcollate,datctype FROM pg_database WHERE datname = 'confluencedb';
confluencedb UTF8 en_US.utf8 en_US.utf8

Raw format doesn't show anything strange:

<ac:link><ri:page ri:content-title="Заявка на исправление" /></ac:link>

It looks completely the same in the database.

Encoding test page (/admin/encodingtest.action) works fine too with one small exception of "The file name has been mangled". I don't think that matters.

I also tried to follow this article - https://confluence.atlassian.com/confkb/links-to-attachments-and-pages-that-contains-non-latin-characters-are-randomly-broken-966676202.html
Symptoms are pretty similar, however all attachments are working fine. I tried reinstalling confluence numerous times, recreating database from scratch, exporting and reimporting everything - nothing helps.

Temporary solution is to open an article and immediately save it, that way all links to it starts working again, but after a day or so it returns to a previous, broken state.

I am completely lost, please help.

[Main User · Accepted solution]

So we moved to xwiki (it can import your data from confluence just fine), since there is no support for atlassian products. Thanks for nothing, i guess.

[Gonchik Tsymzhitov]

Wow, I meet with the same error

[Gonchik Tsymzhitov]

Finally I fixed via adjusting DB for the content table

[Tom Hillartion]

Hi!
I encountered a similar problem (Can you tell me what exactly you did in the DB? Thanks

[Diego]

Hey, hello there!
Great to have you here in the Community. So, as I understand Andrew all links with Cyrillic characters are broken after some time in Confluence. Is that correct?

This behaviour is indeed similar to the one described under Links to attachments and pages that contains non-latin characters are randomly broken.

However you said that all the steps described there have been tried and you have also tried a new installation from scratch but the behaviour is the same.

With the above in mind, could you clarify some points for us? Here they are:

1. What locales are supplied with your system?
   1. You should be able to check this with the command locale -a
2. Are all types of links broken or just the ones pointing to pages in Confluence?
3. Do links with Latin characters behave as expected?
4. What happens if you change the page name to Latin characters and then link to it?

Further, could you provide screenshots of:

1. The broken links
2. The source from a working link
3. The source from a broken link
4. The output from locale -a command

You can check the source of your page with the source editor in Confluence:

Confluence Source Editor

Please, keep in mind that the Community is an open place. Any information posted here can be seen by anyone. So take care not to disclose any sensitive information with the screenshots!

Looking forward to your reply Andrew!

[Main User]

Hello.

1. locale -a show all standard stuff, including en_US.utf8 and ru_RU.utf8, full output is available here.

2. Only links pointing to internal pages are broken. Attachments and external links work just fine.

3. Links with latin characters behave as expected.

4. Changing page name to latin fixes the problem permanently.

I made you a video showing everything in detail (including page sources), it's high quality and you can just stop it anywhere you want.

Steps i do in that video:

1. Showing main page full of broken links.

2. Opening random page (one of those broken links).

3. Starting to edit it and immediately save it without any changes (source stays the same).

4. Returning back to main page, link to page edited in #3 miraculously fixes itself. Main page is not touched in any way.

After some time (1 day max) it will become broken again. That's why i think it's an indexing problem, some timed operation brakes it after completion. I tried to reproduce it manually by running tasks from "Scheduled Jobs" but failed.

[Gonchik Tsymzhitov]

Hm, it something with caches. 

I reproduce it. 

links became broken after cleaning cache.