Bug fix/documentation request: background target doesn't work?

[presto8]

I set up a 3-drive filesystem with hdd.hdd1, hdd.hdd2, and hdd.hdd3. The filesystem's targets were all left to the default values (none). metadata-replicas was set to 3, all other replica values set to 1. I then created a subvolume and set the foreground and background targets for the subvolume to hdd.hdd3. Finally, I ran bcache data rereplicate.

I then unmounted the filesystem, removed hdd.hdd1 and hdd.hdd2, and re-mounted the filesystem with only hdd.hdd3 and -o very_degraded. I expected the files in the subvolume to be available since the foreground and background target were set to hdd.hdd3. However, trying to access the files in the subvolume resulted in many I/O errors.

The documentation at https://bcachefs.org/GettingStarted/ states "For a multi device filesystem, with sda1 caching sdb1: [...] This will configure the filesystem so that writes will be buffered to /dev/sda1 before being written back to /dev/sdb1 in the background, and that hot data will be promoted to /dev/sda1 for faster access."

However, this does not seem to be the case. Is this feature working or is my configuration incorrect in some way? Or if the feature is not working, should the documentation be updated to reflect that?

[jpsollie]

I do not think bcachefs stores metadata on a bg target in a multi-tiered filesystem.
you can check this with "bcachefs fs usage (mountpoint)" and check of your bg target has btree data.
if not, you need to set metadata_target explicitly with --metadata-target=hdd1,hdd2,hdd3

[presto8]

Actually, the metadata was replicated! I was able to do a "ls" on the directory and all of the files showed up. But, when trying to access the files (like cp file /dev/null), then file i/o errors were observed. I didn't explain that as clearly as I could have. But it appears that the data has not been moved to the target. I am not sure if there is anyway to get finer detail on where data lives, on a per-file or per-sub-volume basis?

[jpsollie]

maybe you should check your dmesg: if any file io fails, it could simply be a checksum error. due to an invalid write