I have a large kopia repository in Gdrive. Everything worked fine until yesterday I could not access the repository anymore and got the following error message:
Error: error connecting to repository: unable to list sources: unable to find manifest entries: unable to load manifest contents: error loading manifest content: error getting cached content: failed to get blob with ID qa2328e949a262602138a0c377e1af092-s8209e3abbb2ab48d110: BLOB not found.
I have no clue what caused this for everything was running fine the day before. The only thing changed was upgrading kopia to the new version.
Can you try downgrading Kopia to the version that worked and see if the error persists? That is an easy way to rule out if this was caused by a regression. Downgrading is easy, just download the version you had previously. The only thing to note is you cannot downgrade to below v0.9.0 because of the repo change.
Otherwise, are you using the gdrive repo or connecting to gdrive through the rclone repo? When you say you cannot access the repo, do you mean you literally cannot connect to it or that you can connect and you just get errors when trying to run a snapshot or a restore? If the latter, there could be data corruption. Try kopia snapshot verify. One thing that may have happened is a snapshot was running when you upgraded and that causes some issue with blobs that Kopia thinks should be there but are not.
I reinstalled the previous version v0.11.2 but that did not solve the issue.
I am using the rclone repo.
Normally I use the KopiaUI interface and using this I cannot connect to the rclone repo and it gives the stated error message.
But just now I tried using the Kopia command line commands and that gave the following results:
kopia repository connect from-config --token , connects to the rclone repo without a problem.
kopia blob list; gives no errors.
kopia content list; gives no errors.
kopia index list; gives no errors.
kopia manifest list; ERROR unable to find manifests: unable to load manifest contents: error loading manifest content: error getting cached content: failed to get blob with ID q886d874d6a8b1df0149ec4dbe8ae24b8-s129a53ef8c78c766110: BLOB not found.
I have tried runnng kopia snapshot fix invalid-files --verify-files-percent=100 --commit. But that did not help. It gave the following result:
C:\Users\Ruud>kopia snapshot fix invalid-files --verify-files-percent=100 --commit
2022-07-13T20:13:54.021704+02:00 Listing blobs…
2022-07-13T21:47:43.194737+02:00 10000 blobs…
2022-07-13T22:56:55.656566+02:00 20000 blobs…
2022-07-13T23:41:19.251981+02:00 Listed 26227 blobs.
2022-07-13T23:41:19.252225+02:00 Listing all snapshots…
2022-07-13T23:41:47.081074+02:00 ERROR error running maintenance: error running maintenance: unable to get maintenance params: error looking for maintenance manifest: unable to load manifest contents: error loading manifest content: error getting cached content: failed to get blob with ID q9f929af62689a12d7b6405816434c9f5-s3926780f93563acb110: BLOB not found
2022-07-13T23:41:49.365427+02:00 ERROR unable to list snapshot manifests: unable to find snapshot manifests: unable to load manifest contents: error loading manifest content: error getting cached content: failed to get blob with ID qa2328e949a262602138a0c377e1af092-s8209e3abbb2ab48d110: BLOB not found
No sorry, I did glance at this, but it looked rather complicate. Since I did not have any time today, I decided to run the kopia snapshot fix first, because that could run overnight and share the results. I’ll certainly look into this more closely tomorrow.
I followed all steps from your link. Found the erroneous index file, but removing this index file from the remote repo did not solve the problem. However I did notice something else which seems odd to me. All remote index files are numbered sequentially from 1_0 thru 1_f, 2_0 thru 2_f, etc. But in my erroneous index file some indexes are missing. The current list is: 5_0, 5_2, 5_3, 5_4, 5_6, 5_7, 5_8, 5_9, 5_a, 5_d, 5_e all dated 11 july 2022. But missing 5_1, 5_5, 5_b, 5_c, and possibly 5_f.
I suppose that this could cause the problems. Is there any way to rebuild the index?
Try kopia index recover --parallel=10 --commit. You could also consider including the --delete-indexes command, but I would recommend downloading your indexes to your computer first, in case something goes wrong and you want to put back the ‘old’ ones. See index recover | Kopia
Try running a snapshot and maintenance after that. If still errors, you could also try kopia snapshot fix invalid-files --verify-files-percent=100 --parallel=10 --commit again after you try recovering the index.
I have run kopia index recover --parallel=10 --commit --delete-indexes. The run ended without errors, but did not solve the problem. The above mentioned index list is exactly the same with the same files “missing”. Apparently these files are not missing and the index is correct this way.
Next I ran kopia snapshot fix invalid-files --verify-files-percent=100 --parallel=10 --commit . This ended in error unable to get maintenance params: error looking for maintenance manifest: unable to load manifest contents: error loading manifest content: error getting cached content: failed to get blob with ID q6a0d05972b5f739d45a46ce481fcda0d-s086049b6da019c69110: BLOB not found
I checked this blob in the rclone repo and q6a_0d0 is empty.
Then I checked all other blobs mentioned in the previous error messages, and all are empty in the rclone repo.
Since I understood from other discussions that kopia will never ever write empty files to the repo, this clearly is in error. But how to solve this?
Looking for zero size folders is not a standard function in Gdrive and doing this manually with this number of folders is a nightmare…
But even then, there appear to be so many empty folders. And I fear that deleting so many folders corrupt things even more.
I tries that as first. But with the same problem.
No strange dates to be found.
I tried this too, but after a while I stashed away so many index files that I might as well delete all of them So i put them all back again.
I agree that it is better to wipe the repo and start over. But that raises the question how to do that. Can I just delete all content from Gdrive? Wipe the content with kopia content delete? But leaving the root folder intact in Gdrive will result in the same problems I have now?
I do not think Kopia has a command to delete a whole repository easily. You could do content delete or snapshot delete or manifest delete, but you would need to do that to each individual respective item.
On the other hand, whenever I need to clear a repo, I literally just go to the platform and delete all the Kopia files in the repo. Presumably, there is a specific folder on Gdrive that contains the Kopia repo. Just log into Gdrive and delete that whole folder. Recreate the folder again in Gdrive. Then reconnect to Gdrive with Kopia. You are literally starting a new repo from Kopia’s perspective. Note that this means you will lose all your policies as well, so you will need to recreate them.
I’ve deleted the old repo and created successfully a new one. Recreated the snapshots and policies and it’s running at this moment.
Pity that we could not save the old repo but so be it. There is no data loss since the repo just contained backup files which I also have on a Kopia repo on an external drive. It just takes a long time to upload all data anew.