Unexpected rewrite of a file in Google Cloud Storage

What is the problem you are having with rclone?

When rclone updates modification time in destination (Google Cloud Storage), it seems that it rewrites an object. This is unexpected because a metadata update is enough. Moreover, rewriting causes some troubles:

  • It can incur the early deletion and data retrieval charges for objects stored in a particular storage classes.
  • It creates a non-current version of an object if versioning is enabled for the bucket.

Here we see that rclone created another version of the object with the same conent (same hashes).

Run the command 'rclone version' and share the full output of the command.

rclone v1.64.0
- os/version: Microsoft Windows 10 Home 21H2 (64 bit)
- os/kernel: 10.0.19044.3086 (x86_64)
- os/type: windows
- os/arch: amd64
- go/version: go1.21.1
- go/linking: static
- go/tags: cmount

Which cloud storage system are you using? (eg Google Drive)

Google Cloud Storage

The command you were trying to run (eg rclone copy /tmp remote:tmp)

rclone copy . gcs:bucket1/MD

The rclone config contents with secrets removed.

type = google cloud storage
project_number = ***REDACTED***
service_account_file = ~/REDACTED.json
bucket_policy_only = true
location = us-central1

A log from the command with the -vv flag

2023/10/05 20:16:09 DEBUG : rclone: Version "v1.64.0" starting with parameters ["C:\\Users\\REDACTED\\scoop\\apps\\rclone\\current\\rclone.exe" "copy" "." "gcs:bucket1/MD" "-vv"]
2023/10/05 20:16:09 DEBUG : Creating backend with remote "."
2023/10/05 20:16:09 DEBUG : Using config file from "C:\\Users\\REDACTED\\scoop\\apps\\rclone\\current\\rclone.conf"
2023/10/05 20:16:09 DEBUG : fs cache: renaming cache item "." to be canonical "//?/Q:/temp"
2023/10/05 20:16:09 DEBUG : Creating backend with remote "gcs:bucket1/MD"
2023/10/05 20:16:15 DEBUG : GCS bucket bucket1 path MD: Waiting for checks to finish
2023/10/05 20:16:15 DEBUG : MD-100A.mp3: Modification times differ by -550.5124ms: 2022-05-07 22:13:37.5505124 +0530 IST, 2022-05-07 22:13:37 +0530 IST
2023/10/05 20:16:15 DEBUG : MD-100A.mp3: md5 = cd24e01aadc04db30954d5458983db88 OK
2023/10/05 20:16:17 INFO  : MD-100A.mp3: Updated modification time in destination
2023/10/05 20:16:17 DEBUG : MD-100A.mp3: Unchanged skipping
2023/10/05 20:16:17 DEBUG : GCS bucket bucket1 path MD: Waiting for transfers to finish
2023/10/05 20:16:17 INFO  : There was nothing to transfer
2023/10/05 20:16:17 INFO  :
Transferred:              0 B / 0 B, -, 0 B/s, ETA -
Checks:                 1 / 1, 100%
Elapsed time:         7.2s

2023/10/05 20:16:17 DEBUG : 4 go routines active

Rclone used to use PATCH to update the modtime of objects, but it switched to a COPY over itself.

From the change

gcs: fix need for elevated permissions on SetModTime - fixes [#3493](https://github.com/rclone/rclone/issues/3493)

Before this change we used PATCH on the object to update the metadata.

Apparently this requires the "full_control" scope which Google were
unhappy with in their oauth review.

This changes it to update the metadata by copying the object ontop of
itself (which is the way s3 works).  This can be done with normal

Ultimately this was caused by rclone's consent screen failing review by google

Now it might be that it doesn't need those permissions any more or there is a better way of doing it - any suggestions appreciated!

See the source code here

Thank you for the explanation. I'll try to research the current situation with the permissions and get back.

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.