Some attachments or links are no longer accessible after server migration

Summary

Solution

Problem

After completing a server migration, some attachments or links are no longer accessible after the migration is completed. The base URL may have been changed in the process.

Cause

There are a couple reasons this could happen:

Cause #1: Base URL was changed

For example:

If the links to attachments or pages are hard-coded (e.g., [Alias|http://oldServerName/download/attachments/<page id>/attachment.extension]), the links used will no longer be accessible because they point to the old URL. Often these links will have been created because of the bug CONF-27754.

Cause #2: Incompatible configuration file

You migrated from Windows to Linux, and copied over the confluence.cfg.xml file from Windows. You might see something like this:

Notice the backslash before "attachments". This backslash is used in the Windows version of confluence.cfg.xml, in a few different properties:

• attachments.dir

• confluence.webapp.context.path

• lucene.index.dir

However, using this configuration file in a Linux platform will result in attachments not being found, among other things, due to the backslashes.

Resolution

Choose the resolution based on the cause:

Cause #1

Prevention

When creating links to attachments, use proper wiki markup instead of hard coding the URL.

Manual Correction

If you only have a few links broken due to this, the best way is to fix the links one by one, as described in Links.

Database Operation

Because this is a database update, please make sure that you perform a database backup prior to executing the queries below. Note the queries below are specific to PostgreSQL; if you are using a different database, please consult your DBA:

1. Create a Content Migration table

   1 2 3 4 5 6 7 8 9 10 create table contentmigration ( bodycontentid BIGINT NOT NULL, body TEXT, PRIMARY KEY (bodycontentid) ); insert into contentmigration select bodycontentid, substring(body from '\\|http://oldServerName/download/attachments/.*/') from bodycontent; delete from contentmigration where body is null;

   These queries will search for all occurrences of links that contain the URL you want to replace, insert it into a newly created table called 'contentmigration' and delete any content where the body is NULL.

2. Confirm that contentmigration is populated successfully:

   1 select * from contentmigration;

3. Run the query below to make an update:

   1 2 3 update bodycontent set body = (select regexp_replace(body, '\\|http://oldServerName/download/attachments/.*/', '|^', 'g')) where bodycontentid in (select bodycontentid from contentmigration);

4. Check if the content has been updated as you want them to be:

   1 2 select * from bodycontent where bodycontentid in (select bodycontentid from contentmigration);

5. Fix Attachments with spaces in the name

   You may see that some attachments where their name contain spaces will have pluses '+' in them. Try changing them manually if only a few have these.

   If you have plenty of affected links, you can consider running the query below - bear in mind that the query will also affect any occurrences of pluses '+' outside of the affected links:

   1 2 3 update bodycontent set body = (select regexp_replace(body, '\\+', ' ', 'g')) where bodycontentid in (select bodycontentid from contentmigration);

6. In an effort to clean up the database, after confirming that the updates are acceptable, run this last query to remove contentmigration:

   1 drop table contentmigration;

7. Flush all Confluence caches afterward.

RELATED DATABASE DOCUMENTATION

PostgreSQL 8.4 String Functions

MySQL 5.1 String Functions

Oracle 10g