Confluence attachment backups - server ZIP backup vs directory backup

[Jiri Pik]

If I ask Confluence to backup the attachments directory, it levels out the hierarchy of attachments so that all the IDs are at most in 2 levels.

When I, however, backup the attachments directory manually - tar / gzip, the directory structure below attachments starts with ver003 and then a multiple levels of numbers.

1. What is the mapping between these two file organizations?

2. Would Confluence restore from both types of backups?

3. If I browse the Confluence, all links are indeed of the one level hierarchy e.g.  /download/attachments/39563733/mypicture.png. Where would this picture be located in the ver003 hierarchy?

[AnnWorley · Accepted solution]

It's my understanding that when  the XML export is restored, it is restored back to the hierarchy under ver003 as described in Hierarchical File System Attachment Storage

When the files are not stored in that structure you see errors when trying to access attachments in Confluence. We have a script for moving the attachments back to the expected folders when a mismatch occurs between where Confluence is looking for the attachments and where they really are: How to resolve missing attachments in Confluence

[Jiri Pik]

Ann: THANK YOU.