Max-Sum
(Max Sum)
October 17, 2022, 2:44pm
1
I am using rclone as backend for Kubernetes, csi-rclone . The cache works well but didn't guarantee write back before unmount. This would cause changes to lost when a workload is transferred to another machine.
I proposed:
a command to purge vfs cache or;
an mounting option to force write back before unmount.
ncw
(Nick Craig-Wood)
October 17, 2022, 3:53pm
2
Actually someone made an issue about your second proposal very recently (I don't think that was you was it?)
opened 06:52PM - 12 Oct 22 UTC
enhancement
HI @ncw
I wonder if that would be possible when using
(Asynchronous writes)
…
```bash
--vfs-cache-mode=full/writes
--vfs-write-back=xxs
```
when gracefully stop rclone via **SIGTERM** or via `umount`
in case of dirty blocks (data not yet uploaded to the backend)
set Fuse to Read only and block rclone process shutdown until all blocks are uploaded to backend.
Also do we have any kind of metric or way to know if we have dirty pending blocks? and the amount?
Happy to contribute on this Feature request but I would need a little bit of guideline 🚀
I wrote some ideas on how to implement it there - I'd be interested in your opinions.
Purge as in guarantee all dirty files are uploaded? Or Purge as in just delete everything?
Max-Sum
(Max Sum)
October 17, 2022, 4:19pm
3
This is exactly what I want.
Purge as all dirty block uploaded
ncw
(Nick Craig-Wood)
October 17, 2022, 10:19pm
4
I see. I think vfs/sync would be a better name.
system
(system)
Closed
December 16, 2022, 10:19pm
5
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.