Sorry, this question might have been answered several times already. However, I went through a number of articles about this issue, but was not able to resolve it with the information found in these articles.
My backup is giving the following error:
Unexpected difference in fileset version 2: 2019-05-12 19:16:11 (database id: 8), found 899913 entries, but expected 899914
According to what I found in other posts, I shoud delete the version in question via the command line. When I try to do this, I am getting the following output:
C:\Users\Tournesol>“C:\Program Files\Duplicati 2\Duplicati.CommandLine.exe” delete “file://Y:\backup fast alles 2\” --version=2
Verschlüsselungspassphrase eingeben: ********
Database file does not exist: C:\Users\Tournesol\AppData\Local\Duplicati\IYQQHANOJQ.sqlite
Update “22.214.171.124_beta_2019-07-14” detected
Looking into the graphical user interface, I see that the database mentioned for this backup is a different one:
Any help is highly appreciated. I am running into this error from time to time. In earlier cases, I tried to repair the database (which does not help) and to recreate it (which never finishes). I always ended up in removing the whole backup.
There are a whole lot of options that need to be set appropriately to do a delete from a Command Prompt for a GUI backup job. Doing the delete in the GUI Commandline for the backup will solve that by supplying options for a backup command, then all you must do is edit backup into delete. One method is to change Command to backup and change Commandline arguments box into --version=2 or Add advanced option.
Almost everything uses a database. I think your delete just invented one because you gave it no –dbpath.
Whether or not this fixes the original problem isn’t known, but at least it may let you see if delete can help.
Thanks for your quick reply. Your instructions helped me to execute the delete command. It is a bit puzzling though that the -dbpath option is not mentioned in the documentation of that command.
Unfortunately, deleting the version did not fix the problem. After I had deleted version 2 and ran the backup again, I got a message about unexpected difference in fileset in version 1 and after deleting version 1, a similar message for version 0.
Deleting version 0 did not help either. After deleting version 0 and running the backup again, I got the very same message about version 0 (same values for “found” and “expected”).
I might have heard of SQL DB savvy users editing the database, but I wouldn’t suggest that for everyone…
Possibly you found or tried some other things people attempted, while you were looking for articles earlier?
One thing that might be helpful at least for future reference would be if you’re willing to post a bug report so someone can see what your database looks like. Pathnames are sanitized for privacy against ordinary DB browsers, but a deep forensic analysis might still find them. There’s also a privacy defect in next true beta. 126.96.36.199 beta is just 188.8.131.52 beta plus a warning about Amazon Cloud Drive going away. Since you’ve seen this issue for awhile, you might also consider trying to narrow it down to see if you can find what causes it.
Bug report is running for almost five hours now. I fear that it is hanging. I am going to delete this backup in the weekend and create a new one.
On the next occurance of this problem, I will try again to create a bug report. Thank you for your help, so far.
Another question in this context: When making a backup of the database, I saw that it is stored in %appdata% on the local disk and not on the backup destination. Is the database not needed for restoring files?
It is not absolutely necessary for restore (I recently did a restore without), but it significantly speeds up the metadata processing that is needed for restore.
If you want to back up the local database(s), you’d better create a separate backup job that only backs up these databases, except its own (because it is the one in active use). So if you are in a disaster situation and do not have access to the local databases, the restore operation from this separate job is faster and safer than from some other backup set.
You can check your job log from the backup before your problem to see if compact shows any statistics.
If so, one avoidance of a repetition is to set –no-auto-compact=true (which prevents re-use of free space) or install a fixed version. 184.108.40.206 did well over two months of testing, however 220.127.116.11 canary is the latest canary. If it does well (probably over a shorter period of testing), then it will likely lead to beta at some point.
Note that due to database format updates, downgrading (e.g. to 18.104.22.168 which is basically 22.214.171.124) is hard, but upgrade to next feature beta (or experimental) should work. You should also change your Settings to a more stable channel (Beta for example) if you install Canary. You can change to Canary temporarily if you want to upgrade that way (instead of using full installer), then change it back before a broken canary ships.
EDIT: Upgrading would be to avoid future problems. Current problem is dealt with (or not) as shown earlier.
Thanks for the information provided. I set -no-auto-compact to true for all my backup jobs and will see within a few months, whether this helps. The corrupted backup, I am going to delete and recreate. (I have a lot of redundancy in my backup jobs).
Currently, I am running Duplicati - 126.96.36.199_beta_2019-07-14 and am on the default (beta) upgrade channel. I am not sure whether or not you are advising to upgrade to 188.8.131.52 canary.
Advising upgrade to canary is always a tough call, especially for one so new. I’d feel quite comfortable advising 184.108.40.206 (and I ran my personal backup on that to avoid too-frequent “Unexpected difference”) however in general canary is bleeding-edge not-very-proven stuff, so staying on it can get “surprising”, though sometimes the features and fixes outweigh any accidental breakage. Testers are also needed.
With your level of redundancy, a good plan might be to start on relatively-well-proven 220.127.116.11, upgrade shortly to 18.104.22.168 if 22.214.171.124 is good, and stay on that until next experimental (probably) or actual beta.
Your idea of just burning some extra storage will also work (assuming this is the issue – you can study logs), then hopefully before the storage waste gets too high you can get into safer channel than canary.