Copy vs. Direct to Cloud questions

[nikjft]

Just purchased a couple license to Duplicacy having tested it out and found it to be a very robust solution. Now I’m working on getting backups all ship-shape.

I want a backup process that has my various PCs/Macs backing up to two places - a Linux-based server that is only sometimes available (when they’re at my house - which they aren’t a lot of the time), and also to the cloud (I’ve chosen b2 on that side of things). The server also backs up to the same storage - both locally (backups are on a second HDD) and to b2.

I have two ways to approach this: everything backs up to two locations, or everything backs up to the server, which then copies its backups to b2. Both could work, but the former seems more efficient in a whole lot of ways.

First off, is this possible? SFTP/network drives use a different structure from cloud servers, so does “copy” even work in that situation?

I have also already seeded my backups to BackBlaze (pulled the server’s HDD and ran a normal backup of its own files). If I switch to a backup-then-copy, do I need to re-seed with a copy command or can I somehow get these back in sync in a copyable fashion?

Are there differences in bandwidth or speed between a copy and a backup? My biggest limiter is bandwidth at home, where I have < 1 mbit upstream.

Just looking to understand all the pieces and parts here.

[gchen]

Although either way should work, I would suggest backing up everything to the server, and running the copy command from the server to copy backups to B2.

This is mostly because this way you get identical backups on your server and B2. If a client backs up to two locations, if some files change between these two backups then each location will get a different backup.

The copy command is designed to handle different directory structures. On the other hand, you can’t use other apps (such as rclone) to copy backups from your server to B2, unless you ‘flatten’ the chunk directory first.

Another advantage of the copy command over copying by third-party apps is that you can copy a subset of backups instead of all of them. Say, you may want to copy only backups that has the tag nightly.

The copy command should be pretty efficient after Jeff’s PR (https://github.com/gilbertchen/duplicacy/pull/154). However, another user reported slowness (https://github.com/gilbertchen/duplicacy/pull/154). At this time, it is unclear if it was caused by the use of a different repository id on the storage.