Sorry for resurrecting the thread, but I'm seeing the same in my Rasperry Pi 4GB, and given how slow disk swap is in this device, it makes it a problem.
I'm using three mounts, all with the following options:
[Unit]
Description=GDrive-1
AssertPathIsDirectory=/mnt/GDrive-1
Wants=network-online.target
After=network-online.target
[Service]
Type=notify
User=pi
Group=pi
KillMode=none
ExecStart=/usr/bin/rclone mount \
--config=/home/pi/.config/rclone/rclone.conf \
--allow-other \
--fast-list \
--buffer-size=256M \
--vfs-cache-mode=writes \
--dir-cache-time=24h \
--cache-dir=/mnt/2TB/rclone \
GDrive-1-Crypt:/ /mnt/GDrive-1
ExecStop=/bin/fusermount -uz /mnt/GDrive-1
Restart=on-failure
RestartSec=5
[Install]
WantedBy=default.target
I've set vm.swappiness to 10. My system has around 1GB of ram usage after boot. Once rclone starts activity, I immediately see swap usage even though there still is plenty of ram available. This happens even with vm.swapiness set to 0.
After a few hours of uploading files on just one of the mounts, my usage looks like this:
total used free shared buff/cache available
Mem: 3.6Gi 2.1Gi 82Mi 41Mi 1.4Gi 1.3Gi
Swap: 3.6Gi 858Mi 2.8Gi
This doesn't seem to be that bad, but the problem lies on rclone constantly using swap, even though there is plenty of ram available. Not using swap brings up oom killer.
I've mitigated the issue by using zram-swap, however this is not ideal.