Backup Failure: Cloudberry – Synology Case Confusion

My Cloudberry backup started failing a couple days ago, saying it couldn’t find two file. It was a minor problem but it took some time to figure out. It was a lesson in mixing and matching file systems.

Frustrated Man tile

Frustrated Man tileHad an interesting problem with my Cloudberry Backup for Windows Home Server that took longer to resolve than it should have, even though it was minor. To set the stage, I use rsync to backup my website to my Synology 1511+ NAS running DSM 4. The web server and the Synology NAS both run Linux and a case sensitive file system (hint, hint). I then use Cloudberry Backup for Windows Home Server to backup those files to Amazon S3 (Paranoid? Yup). Cloudberry connects to the Synology NAS through a Windows type share.

When the backup failed I checked the logs like a good troubleshooter and the logs said that 2 files couldn’t be found. Both were cache files on the server that only get updated once a week and sure enough the problems started after the last cache rebuild. I tried accessing both the original and backup copies of the files by pasting in the long path from the logs. All the files were there.

I then deleted the backup copied of the files in case it was an access issue. Both files deleted easily. The next backup resulted in the same errors and the files weren’t replaced. So next step was to refresh the Cloudberry database. Maybe what it thought was backed up was out of sync with reality. These are time consuming processes to run, but only a few clicks of my time to start. Same backup error in the logs.

Since these were cache files (static copies of dynamic content to reduce server load) they weren’t important so I called it quits for the night. Naturally the errors were there in the morning. But this time I read the console status and the error was some paths were not accessible. The logs themselves still said the two specific files were missing.

An inaccessible directory with many files is different than two missing specific files. This time rather then cutting/pasting the path to the file I browsed directory by directory to the path on the web server. And there was the problem. I had two directories with the same name, one in lower case and one in upper case characters. The lower case name was the one being backed up and with content. The upper case one was empty. Upper and lower case are unique in Linux but not in Windows. And since Cloudberry was accessing the files through what it considered a Windows share it was totally confused as was Windows Explorer when I had browsed to it. Synology and my web server had no problems differentiating case. I went in through the Synology File Station software (it’s version of File Explorer) and deleted the empty upper case directory names. I did the same on my web server. The backup ran error free.

So it looks like my caching software created an upper case directory name for some still unknown reason. It hasn’t been changed in a long while so no clue as to why it did it now. If it repeats the problem in a week at least I’ll know how to quickly fix it. And if it continues it may be easier to exclude the cache directory from backups than it would be to find out why the directory is being created.

The OS’s and files systems don’t always play well together, sometimes I’m amazed they work at all.