Restore from old multisite installation failing

Has something changed in the restore process which would affect backups from multisite environments?

Attempted to move a site between servers today. Old server was a multisite install, with nginx in front handling HTTPS encapsulation and force_https is turned on.

New server is a standalone version of the standard install, using Let’s Encrypt.

Checking the backup file the original uploads are present, but they don’t appear to be restored to the filesystem on the new server. References to images across the site refer to the old multisite ‘secondsite’, so paths point to /uploads/secondsite/original rather than /uploads/default/original.

Attempted restore via gui, and command line for good measure. Plan to dig into this further, but figured it was worth posting in #support for now and will move to #bug if someone can confirm this is still a supported scenario.

This is something that previously worked, but I’ve not done it for at least a year, maybe longer.

5 Likes

So the files are there but in the wrong place? And the database points to default rather than secondsite? Sounds like a bug, but I guess you can just rename the directory? Or did the backup not include the uploads at all?

They’re in the backup archive, but never make it to the filesystem.

1 Like

That’s bizarre! Does sound like a bug. And neither system has S3 uploads configured?

Correct. I’ll do some more testing later.

1 Like

That’s unexpected. Any chance I can have a look at the backup file? Otherwise, if you want to debug this yourself, take a look at https://github.com/discourse/discourse/blob/master/lib/backup_restore/uploads_restorer.rb.

3 Likes

The willingness of @team to poke into stuff like this is really awesome. @tgxworld helped me resolve an edge case issue last week. :tada:

3 Likes