yote_zip

yote_zip, 1 year ago

Seems like there’s some bitrot in the middle of the file, and whatever you’re using to play back the original file just skips it and doesn’t care enough to halt playback. You might try looking for ways to restore as much of the file as possible with something like this, assuming the mkv is a unique copy that you can’t get anywhere else.

Edit: I’m also curious if this file lives on an XFS/BTRFS/ZFS filesystem. The reflink property of these filesystems may be the reason that you can copy within the same folder without it throwing an error.

yote_zip, 1 year ago

Fair enough. I would at least try to get the damaged file off of the disk so you can potentially fix it later, or just have it available to play in its broken state. For the future you should probably be running monthly BTRFS scrubs to detect bitrot sooner, and potentially you should have some backups or data redundancy so you can repair the bitrot when it’s detected.

yote_zip, 1 year ago

It goes without saying but the number of errors you should get on a scrub is ideally 0. Bitrot happens from time to time which is why you should keep some data redundancy/backups so you can repair it when it’s detected, but that number seems higher than normal. Your disk may be going bad if you’re getting that many read errors; I’m not sure. I believe you’re already backing up data off this drive but yeah I would get everything important off the drive ASAP, then run a SMART short test and a SMART long test to see if that reports that anything is wrong. The disk may be fine but better to be safe than sorry.

Back to the video file, I’m assuming it was not one of the ones that BTRFS fixed automatically? The only real options for data recovery are to rescue the file minus the bad blocks with e.g. ddrescue (which I don’t personally have familiarity with) or something similar, or to encode through the errors with ffmpeg if it will let you.

yote_zip, 1 year ago

Okay cool. I would be wary of that drive just in case, and I would definitely schedule weekly SMART short tests and monthly BTRFS scrubs on it if you go with BTRFS in the future. EXT4/XFS/etc do not have a concept of data checksums, which means they can’t scrub and check for bitrot - this might be problematic if you find that your disk starts causing bitrot because you won’t know where it’s happening.

I follow Backblaze’s rules on detecting impending drive failure:

• SMART 5: Reallocated_Sector_Count.
• SMART 187: Reported_Uncorrectable_Errors.
• SMART 188: Command_Timeout.
• SMART 197: Current_Pending_Sector_Count.
• SMART 198: Offline_Uncorrectable.

If any of these SMART metrics are higher than 0 I’d expect failure soon and take precautions.

yote_zip, 1 year ago

Try this answer. I guarantee there is a way to read the file front to back while skipping errors, but I run so much data redundancy that I don’t have any experience with it.