Mounted OneDrive: Invalid Parent Path

[Ole (Ole Frost)]

@TheEpicSnowWolf

thank you so much for the analysis! This has brought me a huge step forward toward getting this resolved, at least for my personal use case here.

Your are welcome, happy to hear!

@TheEpicSnowWolf

So first things first, to answer your question, I am honestly not exactly sure how old my Microsoft account is, I know that I started using OneDrive in 2016 which would have been after the SkyDrive rename, but it is not unlikely that my Microsoft account itself is in fact older. Do you know of any quick way to check that? I have not (at least to my knowledge) set up the drive as shared or did anything special to it, should be pretty much just the out-of-the-box OneDrive that every personal M365 subscriber gets.

Thanks, sounds like something rclone should be able to handle without issues. The OneDrive backend was added to rclone in 2015, so I would expect it has handled lots of old Microsoft accounts before.

This just further strengthens my suspicion that the bug lies in the relatively new polling code.

@TheEpicSnowWolf

So, the good news is, after removing the lock this time and trying to run a backup with the poll intervall set to 0, I actually got a backup to complete successfully, which is amazing!

Great, can I persuade you to create a GitHub bug report based on the information in your above post:
https://forum.rclone.org/t/mounted-onedrive-invalid-parent-path/35106/7

@TheEpicSnowWolf

However, when borg tried to clean up the repository by removing the lock, it ran into an I/O error, which is at least a new one.

Hmm, not sure, but it looks like a timing issue where the underlying rclone rmdir sometimes reaches OneDrive before the rclone deletefiles have completed. Perhaps triggered by Borg doing rapid renames/deletes of temp files.

I did some research and it seems like Borg and rclone mount doesn't always play well together, here are some links:
https://forum.rclone.org/t/best-practice-for-integrating-rclone-and-borgbackup/26166
https://github.com/borgbackup/borg/issues/5324

Based on the pattern I see in your logs, it may be possible to work around (some of) these issues by keeping more data in the local mount cache and writing less often to OneDrive. Here are some parameters you may want to modify/add to your mount command to do this:

--vfs-cache-mode=full --vfs-write-back=10m --vfs-cache-max-age=48h --vfs-cache-max-size=10G

You can probably guess my intent when reading about the parameters here:
https://rclone.org/commands/rclone_mount/#vfs-file-caching

It is a quick guess based on incomplete knowledge and therefore probably need some testing and adjustments to fit your machine and usage - and is given without any guarantee of success