orphaned tombstones...

[taichi]

Hiya's,

Wondering if someone could provide some ideas to resolve some lingering tombstones in which the actual data file has been deleted. I've run fsck and fsck -o which identifies the issue (as pasted below), but does not seems to actually resolve the issue by deleting the invalid tombstones. How might I go about resolving this? I could go through and delete them manually but I'm a bit confused why fsck doesn't resolve the problem itself.

Thanks,
Rhys

greyhole.log:

Jun 28 19:05:19 7 fsck_file: Now working on task ID 219871: fsck_file archive/My Docs/RGV/Kit/page36.jpg
Jun 28 19:05:19 7 fsck_file: Loading tombstones for archive/My Docs/RGV/Kit/page36.jpg... Got 1 tombstones.
Jun 28 19:05:19 4 fsck_file: WARNING! No copies of this file are available in the Greyhole storage pool.
Jun 28 19:05:19 7 fsck_file: Saving 1 tombstones for archive/My Docs/RGV/Kit/page36.jpg
Jun 28 19:05:19 7 fsck_file: Saving tombstones in /mnt/drive2/gh/.gh_graveyard/archive/My Docs/RGV/Kit/page36.jpg
Jun 28 19:05:19 7 fsck_file: Saving tombstones (backup) in /mnt/drive3/gh/.gh_graveyard_backup/archive/My Docs/RGV/Kit/page36.jpg

and in the tombstone:


a:1:{i:0;O:8:"stdClass":3:{s:4:"path";s:49:"/mnt/drive2/gh/archive/My Docs/RGV/Kit/page36.jpg";s:9:"is_linked";b:0;s:5:"state";s:4:"Gone";}}


2.6.35.6-45.fc14.x86_64
samba-3.5.5-68.fc14.1.x86_64
hda-greyhole-0.9.2-1.x86_64

[lrevxl]

You've got the idea backward for orphaned files. That option looks for files which don't have any tombstones and creates tombstones for them, it does not look for tombstones without file copies. I don't don't think there's a way to clear out barren tombstones. It might be a good feature to request on the github.

You can always manually delete the tombstones and its backup from the drives.

[taichi]

Yeah, I understand what the real orphans are. I was just using the term as it still applies in the strict sense of the word.

I'm a little bewildered that functionality doesn't exist as part of the fsck. But if it doesn't, it doesn't. Might have to put forward the suggestion like you suggest since it should be relatively easy to implement - although there may be caveats which would cause permanent data loss. Maybe that's why it hasn't been done.

Thanks anyway, at least I'm not going crazy