https://bugzilla.novell.com/show_bug.cgi?id=880599
https://bugzilla.novell.com/show_bug.cgi?id=880599#c13
James Carter changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEEDINFO |ASSIGNED
InfoProvider|jimc@math.ucla.edu |
--- Comment #13 from James Carter 2014-06-15 01:11:28 UTC ---
Curse, curse, my i686 test machine crashed and nobody is on site due to parents
attending graduation ceremonies taking up all the parking.... You don't want
to know. Anyway, I attempted to do tests on other i686 machines, similar but
not identical. I did a lot of trials (dd if=/dev/zero
of=/net/$HOST/mathtest/oink.dat bs=1M) with both kernel 3.11.10-11.1 and
3.11.10-0, The NFS client had 3.11.10-11.1 and mounted with Nfsvers=3 in
/etc/nfsmount.conf. The target filesystem was ext3 but was actually a 1Gb file
attached to a loop device, on both servers, whereas on Ocelot the bare metal
was ext3. Hiss, boo.
On 3.11.10-11.1, 77 trials. Most got EDQUOT (quota exceeded). Two got EIO
(I/O error). 14 got Stale NFS Filehandle, but those shouldn't count because I
removed the target file on the server before the client had closed it (even
though the server was refusing to write anything because the user was over
quota). (I messed up a test script.) I was hoping to get consistent
misbehavior, which could then be shown to be cured by the test kernel, but
that's not how it worked out.
On 3.11.10-0, 12 trials, most got EDQUOT, one got EIO (I/O error).
In all EDQUOT outcomes, writing was cut off exactly at the quota limit,
although the files were sparse, i.e. the last written block was further along,
sometimes quite a bit, i.e. over 1Gb when the quota was 750Mb.
So the conclusion has to be, kernel 3.11.10-0 is no worse than 3.11.10-11.1,
and given that you fixed a bug, we should move forward with this patch. But
I'm afraid we haven't seen the last of the NFS problems. Next week when I can
get at the original test machine (Ocelot) I'll re-test, and I'll post something
if the results are significantly different from the above.
If I hit something in the future and can come up with a procedure to reproduce
it, I'll open a new bug.
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.