Invalid checksum, how do I fix this?

During a backup a full maintenance was done and I got

Rewriting contents from short packs...
unable to rewrite content "221e7d9b291385853988adf8c573289c": unable to get content data and info: invalid checksum at p8e1243cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b offset 5017640 length 2268/2268: decrypt: unable to decrypt content: cipher: message authentication failed
unable to rewrite content "30e68ee9a8746c121b66dcdef9770031": unable to get content data and info: invalid checksum at p8e1243cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b offset 5020054 length 814/814: decrypt: unable to decrypt content: cipher: message authentication failed
unable to rewrite content "7bda9ff16b1df4a2c8a84ae85b41acb6": unable to get content data and info: invalid checksum at p8e1243cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b offset 5019908 length 146/146: decrypt: unable to decrypt content: cipher: message authentication failed
unable to rewrite content "828129b51404b77e93ee9dd9b63f60eb": unable to get content data and info: invalid checksum at p8e1243cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b offset 5014849 length 2791/2791: decrypt: unable to decrypt content: cipher: message authentication failed
unable to rewrite content "fcc834c840fdfafb1b97dcc06ee92152": unable to get content data and info: invalid checksum at p8e1243cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b offset 5020868 length 1128/1128: decrypt: unable to decrypt content: cipher: message authentication failed
Finished full maintenance.
ERROR error running maintenance: error running maintenance: error rewriting contents in short packs: failed to rewrite 5 contents

I looked at repository repair from-config but was unsure if this is the way to go.

My repository is on pCloud and (fuse) mounted locally at /home/manfred/pCloudDrive.

How do I fix this?

I hope I do not have to create a secondary repository.

This would be really hard as I have an upload bandwidth of 10 Mbit/s and the repository has a size of 1TB. On pCloud I have left 780GB.

That is usually an indication of data corruption either in storage (hard disk, ssd) or in memory (faulty RAM).

Looks like the file p8e1243cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b is corrupted.

The storage is in pcloud, and mapped to my local system via a fusemount. I guess that somehow during a backup there occured a corruption.

I have no idea how to fix it without duplicating the repository which is not easily possible to do.

Your statement “That is usually an indication of data corruption…” made me think twice.

I didn’t believe that

  • my memory is corrupt
  • my local ssd is corrupt
  • pcloud storage is corrupt

but something must be corrupt.

So I checked the sha256 checksums of the 3cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b.f file

It was different when mounted via the pCloud fusermount and when fusermounted via rclone.

So, my conclusion was that the pCloud cache was corrupt.

I stopped the pcloud service, and deleted the ~/.pcloud directory. Then I started pCloud service again and ran kopia maintenance run --full again which now succeeded.

Oh, that’s very interesting. Great find, would it be worth following up upstream (rclone, fusermount, pcloud) to get to the bottom of corruption?

BTW. You should be able to skip fusermount and directly use rclone provider from Kopia.

Not sure how to get to the root of the problem.

It could have happened that I had some network problems on December 14 (that was the date the 3cc4a96fbd35e4da55d244bb6d0-sba5a0776993e286910b.f has been changed last time.

My guess is that there was a problem with the pcloud service which didn’t handle things properly. So, I would need to ask the pcloud guys, but I have no idea how to track things down as the cache is now ok again.

I will play with rclone to see how it works in conjunction with kopia.