On Fri, 7 Nov 2008, Sage Weil wrote:
> Ah, ok, makes sense. Can you describe the workload leading up to the rm a
> bit? Is it this something you can reproduce? I can see where it bugged,
> but that's unfortuantely a long way off from the root of the problem
> (which looks like a missing ref cou
On Fri, 7 Nov 2008, Jerker Nyberg wrote:
> On Fri, 7 Nov 2008, Sage Weil wrote:
> > Do you mean that 'rm' segfaulted, or one of the daemons segfaults, or the
> > kernel oopsed?
>
> The command 'rm' segfaulted. I can still reboot the machine with the
> command "/sbin/reboot -f -n".
>
> The file s
On Fri, 7 Nov 2008, Sage Weil wrote:
> Do you mean that 'rm' segfaulted, or one of the daemons segfaults, or the
> kernel oopsed?
The command 'rm' segfaulted. I can still reboot the machine with the
command "/sbin/reboot -f -n".
The file system, all of it it seems, can still be accessed from th
On Fri, 7 Nov 2008, Jerker Nyberg wrote:
> On Fri, 31 Oct 2008, Sage Weil wrote:
> > If the problem is reproducible, the OSD debug output when cosd is
> > started with '--debug_osd 20 --debug_ms 1' would help us out!
>
> I now run the unstable branch of Ceph, a couple of days old. I wrote a
> co
On Fri, 31 Oct 2008, Sage Weil wrote:
> If the problem is reproducible, the OSD debug output when cosd is
> started with '--debug_osd 20 --debug_ms 1' would help us out!
I now run the unstable branch of Ceph, a couple of days old. I wrote a
couple of GB on all nodes and then a "rm" segfaulted o
On Fri, 31 Oct 2008, Carlos Maltzahn wrote:
> Thank you for running these tests and sharing your results. What
> switch are you using?
The switch is a HP Procurve 2824.
--jerker
-
This SF.Net email is sponsored by the Mob
On Fri, 31 Oct 2008, Jerker Nyberg wrote:
> Hi all,
>
> I ran a simple write throughput benchmark for Ceph which might be of
> interest.
>
> http://www.update.uu.se/~jerker/ceph/
>
> Adding one more drive and running two cosd servers on each node does not
> seem to get better performance.
Thi
Hi Jerker,
Thank you for running these tests and sharing your results. What
switch are you using?
Thanks,
Carlos
On Oct 31, 02008, at 4:30 AM, Jerker Nyberg wrote:
>
> Hi all,
>
> I ran a simple write throughput benchmark for Ceph which might be of
> interest.
>
> http://www.update.uu.se/~jer
