Problem:
When I copy to an S3 remote, it fails with this error
2020/09/03 17:20:51 ERROR : test2.txt: corrupted on transfer: MD5 hash differ "f127881627718748c34ec1ce7888d6f9" vs "d57f66dee17fef1d018429a223c859c6"
If I drag and drop it to a mount, Windows says "Copying 1 Item". It completes with no errors, but it does not show up in S3 (due to the Hash error obviously). Well, it shows for a split second in S3, then is removed.
This remote was working before, but now I can't copy anything to this remote. Even an empty file will not transfer.
The files are not open or being edited! I'm 100% sure.
Rclone Version:
rclone v1.51.0
- os/arch: windows/amd64
- go version: go1.13.7
Yes, actually. I think you're on the right track! I thought maybe that was handled automatically, because it was working before. (I'm pretty sure my other bucket that is working is also encrypted, but not sure b/c I don't have console access on that one).
Ok, the --ignore-checksum flag works for mount as well.
So really we're just masking some bug, eh?
It doesn't seem like the missing kms key was the issue. To prove that I wasn't crazy that it was working before (with no KMS key configured), I have removed the encryption settings from the config and it is working fine... with only adding the --ignore-checksum. The file are intact in the AWS GUI.
this is not a bug, as rclone is working as expected based on the documentation.
however,
perhaps setting --s3-upload-cutoff=0 is a work around, perhaps you can test that.
https://github.com/rclone/rclone/issues/1824
"We noticed that once we enabled default encryption on s3 buckets, small files failed to move with rclone. Once a file is large enough to transfer via multipart uploads, the problem goes away"
https://rclone.org/s3/#multipart-uploads
"rclone switches from single part uploads to multipart uploads at the point specified by --s3-upload-cutoff . This can be a maximum of 5GB and a minimum of 0 (ie always upload multipart files)."
Let me make sure I understand your question. Are you asking if upload_cutoff = 0 is an acceptable solution for me? Yes, I think that's fine. I'm more comfortable with that than disabling checksum!
My bucket has been configured to use KMS encryption on the AWS side from the beginning. But no mention of it in Rclone config.
Even now I still have the same setup, with only modifying the upload-cutoff.
During our troubleshooting, I did temporarily add server_side_encryption = aws:kms and the key.
That didn't work so I removed it. It automatically uses that anyway, as it is taking the default from the bucket config. It is the same when I use python boto3 or CLI, you only need to specify the KMS key if you want to use one other than the bucket's default.