That message is more troubling. Unless the files in question have hard
links, we should not have a case where busy inodes are deleted. This
sets up conditions for possible inode mode changes or file corruption --
the server has deleted the object and may reuse the inode number, but
locally we still have an active copy.
If the extra inode count is from a hard link, there's no problem -- the
object will still exist on the server because of its other link.
2.1.64 has fixes to avoid this problem in all the cases I've identified
so far, but apparently there are more to be tracked down.
Regards,
Bill