I have been using Synology Cloud Sync for a while to sync data from my NAS to S3.
Is it possible to switch from Cloud Sync to rclone and have it not reupload data? One problem I can see is that the two products use different metadata tags to track source timestamps.
--checksum is in use but the source and destination have no hashes in common
(Source is SMB, target is S3. Not sure of the problem.)
The --size-only option appears to work ok, but maybe not the ideal option.
Surprising (to me) is that Synology Cloud Sync and rclone both use the S3 metdata tag of x-amz-meta-mtime. Guess there are differences in precision or something.
Yes, although rclone is probably getting that date from the object metadata tag on the S3 side, right... (which if true means maybe it is interpreting the cloud sync timestamp properly?)
[aws-rclone]
type = s3
provider = AWS
access_key_id = XXX
secret_access_key = XXX
region = us-west-2
acl = private
[nas]
type = smb
host = XXX
user = XXX
pass = XXX
I’ll take a look at the guide you posted.. thanks!
There are 589 files in this folder, and rclone wants to xfer 501 of them due to the modification timestamp being more than 1ms out of sync. The amount out of sync varies wildly:
Modification times differ by 19522h56m28s: 2019-10-26 21:23:36 -0700 PDT, 2022-01-17 15:20:04 +0000 UTC
Modification times differ by 15588h17m38s: 2020-04-07 20:02:26 -0700 PDT, 2022-01-17 15:20:04 +0000 UTC
Modification times differ by 3.6994964s: 2025-10-31 00:00:25.3005036 -0700 PDT, 2025-10-31 07:00:29 +0000 UTC
Modification times differ by 1.5182517s: 2025-11-03 13:00:08.4817483 -0800 PST, 2025-11-03 21:00:10 +0000 UTC
Modification times differ by 3218h18m58s: 2021-09-05 06:01:06 -0700 PDT, 2022-01-17 15:20:04 +0000 UTC
Modification times differ by 7976h18m57s: 2021-02-18 23:01:07 -0800 PST, 2022-01-17 15:20:04 +0000 UTC
Modification times differ by 1.1254861s: 2025-11-03 01:00:11.8745139 -0800 PST, 2025-11-03 09:00:13 +0000 UTC
Modification times differ by 1.5920738s: 2025-11-03 23:00:13.4079262 -0800 PST, 2025-11-04 07:00:15 +0000 UTC
Modification times differ by 2602h19m28s: 2021-09-30 22:00:37 -0700 PDT, 2022-01-17 15:20:05 +0000 UTC
Modification times differ by 1.3693084s: 2025-10-29 12:00:20.6306916 -0700 PDT, 2025-10-29 19:00:22 +0000 UTC
Modification times differ by 19522h56m29s: 2019-10-26 21:23:36 -0700 PDT, 2022-01-17 15:20:05 +0000 UTC
Modification times differ by 1348h17m1.944625s: 2021-11-22 03:03:03.055375 -0800 PST, 2022-01-17 15:20:05 +0000 UTC
Modification times differ by 1.6245446s: 2025-07-22 22:24:54.3754554 -0700 PDT, 2025-07-23 05:24:56 +0000 UTC
Modification times differ by 844.4459ms: 2025-10-29 00:00:24.1555541 -0700 PDT, 2025-10-29 07:00:25 +0000 UTC
Modification times differ by 1.5546133s: 2025-11-03 09:00:16.4453867 -0800 PST, 2025-11-03 17:00:18 +0000 UTC
I'll look at the metadata tag more closely, but maybe Synology Cloud Sync adjusted how they tag at some point. I have been using it for almost 8 years...
I just realized something. A few years back I switched from B2 to S3, and I almost certainly used rclone to facilitate the transfer. That's why some of the S3 objects have the x-amz-meta-mtime metadata tag, not because Synology Cloud Sync uses it.
I don't see any metadata tag for the objects uploaded in the past few years, so I guess Cloud Sync doesn't do it that way at all.
Checksum option sounds good, but I need to see if I can get it to work on the SMB side. (I haven't looked at your how-to yet.)
Looks like about 780MB, or 40% of the total folder size, would need to be transferred. This folder is quite small compared to the others. I definitely don't want to reupload data. I sync several TBs right now with Cloud Sync.
I checked your how-to and that makes sense. One question: will rclone update the tagging on the S3 side so that it doesn't have to verify checksums next time?
The answer is no. Using --checksum it will realize it doesn't have to upload the file again, but there is no effort to update the metadata tag on the remote side. As such, I will have to continue to use the --checksum option going forward, which is less than ideal.
Would be nice if it could somehow just update the metadata tag so that future rclone runs don't need to calculate checksums for the source side.