Good information!
Tip: You can use "rclone config show" or "rclone config show myRemote:" to print the (decrypted) config file, or the config of a single remote.
(@Animosity022 : I suggest we add this to the support template and “How to ask for assistance”)
I have confirmed your above observations with this config:
[myGoogleDriveUnion]
type = union
upstreams = myGoogleDrive1: myGoogleDrive2:
action_policy = lfs
create_policy = lfs
It has the same behaviour; it first fills the remote with least free space and then errors when it is out of space.
Strictly speaking, this is correct according to the “lfs” policy description: “Create category: Pick the upstream with the least available free space.” Rclone does indeed pick the remote with least free space (E.g. 0 bytes) and then tries to create.
The policy you are looking for is: “Create category: Pick the (first) upstream with enough free space to create the file.” But unfortunately, this doesn’t exist.
I therefore see no possibility to fulfil your wish for sequentially filled upstreams, sorry.
The rclone union is inspired by trapexit/mergerfs having similar policy descriptions.
Do you know if mergerfs or other similar file systems can do sequential fills?