I am currently synchronizig a Kopia server repo to a Wasabi S3 bucket. The sync is probably halfway through and kopia.repo/kopia.maintenance have already been synched. However, looking at the bucket, the two files don’t have the suffix .f which might be deliberately the case to prevent anyone from messing with the repo until the synch has finished.
If that’s the case, then it would be no wonder, that I cannot connect to the repo and I am getting a password refused error.
Well, the S3 (that includes Wasabi), GCS, B2 and Azure backends don’t use .f suffix nor directory structures because the underlying stores provide atomic writes and can scale horizontally. ‘.f’ is only needed to distinguish partially-written files from fully written ones (have .f) and directory structure helps reduce local filesystem load (also not needed on cloud providers)
To sync repositories between filesystem and cloud you need to use kopia repo sync-to and not regular file sync tool (well you can, but you must flatten the structure and remove ‘.f’ extension from the destination). Sync-to provides this automatically - it has atomic writes and incremental synchronization.
Yes, I know… and I am actually running kopia repo-sync-to. I was under the impression, that I would get a fully operable repo, to which I could connect my KopiaUI and e.g. restore files from, when I am not on my local LAN.
Hi @jkowalski , can you please either confirm or deny, that it’s possible to actually login to a cloned repo? My sync-to S3 job has finished (after 6 days), but I cannot login to that repo, since Kopia refuses to accept the repo password from the source repo and I thought that this should work.
Sure - I first tried that in Kopia UI, but since that wasn’t working, I tried with kopia on CentOS as well…
[root@kopia repos]# kopia repo connect s3 --secret-access-key=<super secret key> --access-key=<access key> --endpoint=s3.eu-central-1.wasabisys.com --bucket=kopiast7a
Enter password to open repository:
failed to open repository: invalid repository password
I tried to enter the original repo password, which I used to initially create the repo on disk… As you can see, there’s not much to it…
I can get you a screenshot, since I don’t have the FS mounted anywhere, but I can confirm that the files kopia.maintenance, kopie.repository and a whopping 22k of blobs are in the bucket.
If that would be the case, then sync-to shouldn’t work either, should it?
And this shouldn’t work as well…
[root@kopia repos]# kopia repository sync-to s3 --dry-run --max-upload-speed=1048576 --secret-access-key=<secret key> --access-key=<access key> --endpoint="s3.eu-central-1.wasabisys.com" --bucket=kopiast7a
Synchronizing repositories:
Source: Filesystem: /mnt/kopia/ST7A
Destination: S3: s3.eu-central-1.wasabisys.com kopiast7a
NOTE: By default no BLOBs are deleted, pass --delete to allow it.
Looking for BLOBs to synchronize...
Found 21701 BLOBs in the destination repository (501.1 GB)
Found 24721 BLOBs (570.8 GB) in the source repository, 3040 (69.7 GB) to copy
Found 0 BLOBs to delete (0 B), 21681 in sync (501.1 GB)
I also ran tcpdump when trying to connect and it showed traffic to the appropriate hosts…