[OpenAFS] AFS CopyOnWrite problem

[Marc Santoro]

We have two OpenAFS fileservers. After a reboot, we can no longer create
or remove files in certain directories on one of the non-replicated
volumes on one of the servers, as well as removing or renaming these
directories. Files in that directory can be changed (though not removed or
renamed), and files/directories in subdirectories work as expected.

The failed commands print "No space left on device". There is plenty of
space left.

Ex:
$ touch a
touch: cannot touch `a': No space left on device
$ touch foo/a (where foo is a pre-existing empty directory)
$ rm foo/a
$ rmdir foo
rmdir: `foo': No space left on device

Both AFS fileservers are running 1.2.11, on Linux 2.4.25 (Debian
packages).

in the FileLog, we see:

Thu Mar 18 14:06:44 2004 CopyOnWrite failed: Partition /vicepa that
contains volume 536870918 may be out of free inodes(errno = 17)

Once for each error. The filesystem has plenty of free inodes; files can
be created in other directories.

Both servers have been rebooted, and "salvage"d, to no avail. Salvage
found some inode/vnode version discrepancies, but nothing else.

Help! :)