Missing blobs - what are next steps? (not using rclone)

Hello. I am doing SFTP backups. I see a lot of ‘missing blob’ posts that mention rclone, but I am not using rclone.

I logged into my server and ran verify:

screen -S kopia-verify -d -m bash -c "kopia snapshot verify --verify-files-percent=100 --file-parallelism=10 --parallel=10"

And the output included,

ERROR error processing use@host:/backup/path@2025-01-19 16:51:54 PST/.cache/kopia/content-logs/kopia-20250110-180454-196071-server-start.0.log: object Ix9acf8210cb6104b8a5cbe445276aae1e is backed by missing blob p7a48b72d6e056a411825e64d418675ac-sd22b78f3bc02f75d130

ERROR object Ix9acf8210cb6104b8a5cbe445276aae1e is backed by missing blob p7a48b72d6e056a411825e64d418675ac-sd22b78f3bc02f75d130

Based on things I read on this message board and Verifying Validity of Snapshots/Repositories and Trying to Repair Any Corruption | Kopia I looked for 0 size or outdated files:

find . -size 0 # found one!
trash ./p31/e33/0f9a9204fa27f506645eef5ad87-s0e70f3b811b20d1512f.f.tmp.b1b21a07642776e8

find . -newermt "2026-01-01" # future found nothing
find . ! -newermt "2024-01-01" # past found nothing

I then re-ran

screen -S kopia-verify -d -m bash -c "kopia snapshot verify --verify-files-percent=100 --file-parallelism=10 --parallel=10"

And it found a NEW missing blob:

ERROR error processing user@host:/backup/path@2025-01-19 16:51:54 PST/.cache/icon-cache.kcache: object Ixf1954f519d7f98c672ee027b16805183 is backed by missing blob pc3fa259830785b2ef6f0358db3c4f774-sd22b78f3bc02f75d130

ERROR object Ixf1954f519d7f98c672ee027b16805183 is backed by missing blob pc3fa259830785b2ef6f0358db3c4f774-sd22b78f3bc02f75d130

But I note this is the same snapshot: 2025-01-19 16:51:54 PST.

I’m not sure what to do now. If the missing files are just a log and .cache file, I don’t think I need them. But are these missing blobs indicative of something more serious?

I’ve read these posts:

Should I do this?

kopia snapshot fix invalid-files --commit=true

Some of the posts mention doing that, but that it does not actually repair anything and the next snapshot may still have things missing. Is there some way to make sure the next snapshot is complete, even if old ones are not? Should I just delete that old snapshot?

Thanks,

-k.

Because errors were both from the same snapshot, I tried deleting that snapshot in the KopiaUI, and then

kopia maintenance run --full

Which reported more errors:

Running full maintenance...
Looking for active contents...
Looking for unreferenced contents...
GC found 71035 unused contents (11.9 GB)
GC found 15232 unused contents that are too recent to delete (3 GB)
GC found 2359035 in-use contents (1.2 TB)
GC found 59 in-use system-contents (50.2 KB)
Rewriting contents from short packs...
unable to rewrite content "2183aaf357c7787b5d7794643ed92f53": unable to get content data and info: error getting cached content from blob "pf49eb78295aa77db459cc092e97830cb-sd22b78f3bc02f75d130": failed to get blob with ID pf49eb78295aa77db459cc092e97830cb-sd22b78f3bc02f75d130: BLOB not found
unable to rewrite content "7533f39b823d9a691c3048c079b0f99c": unable to get content data and info: error getting cached content from blob "pc3fa259830785b2ef6f0358db3c4f774-sd22b78f3bc02f75d130": failed to get blob with ID pc3fa259830785b2ef6f0358db3c4f774-sd22b78f3bc02f75d130: BLOB not found
Total bytes rewritten 382.2 MB
Finished full maintenance.

This looks like a backup of Kopia’s cache directory which is just a waste. Best practice is to exclude cache directories from backups by adding a CACHEDIR.TAG file (see specification) and setting --ignore-cache-dirs=true in the (global) policy.

Yes. It has fixed multiple errors for me in the past and was recommended by the author of Kopia multiple times.

You should definitely not delete any files in your repository manually. If you have to manually edit your repository, renaming or moving the file in question is the better approach.