[Gluster-devel] df bug in mainline-2.5

[Amar S. Tumballi]

Hi Brent,
 Replies inline..

On 6/26/07, Brent A Nelson <brent at phys.ufl.edu> wrote:
>
> I just started testing the new code, which seems to be going rather well
> so far.
>
> Two things:
>
> 1) df is reporting incorrect sizes.  The sizes are (number of subvolumes
> in my unify)*(sizes of df for the namespace cache), rather than the sum of
> the sizes of the subvolumes.  The filesystem itself is working fine (so
> far), however, and writes are occuring on my subvolumes.


strange, according to code, the df fop (statfs) is sent only to subvolumes
and not to namespace at all. So, it should give sum of subvolumes itself.
Can you confirm this?

2) How can I best tweak the alu scheduler such that multiple concurrent dd
> processes, started at approximately the same time, will go to different
> disks rather than the same one?  I believe this used to do the trick
> reasonably in the old code:
>    option alu.orderdisk-usage:read-usage:write-usage:open-files-usage:disk-speed-usage
>    option alu.disk-usage.entry-threshold 2GB
>    option alu.disk-usage.exit-threshold 60MB
>    option alu.stat-refresh.interval 10sec


This should work fine even now (as nothing related to schedulers is changed
other than min-free-disk option is changed to %), as alu schedules nodes in
rr pattern if all the nodes fit in the limits.

Thanks,
>
> Brent
>
> PS Which performance translators should perform reliably in the new code
> at the moment?


other than io-threads (and stat-prefetch, which is not yet ready at all),
you can use other performance translators.

-bulde

_______________________________________________
> Gluster-devel mailing list
> Gluster-devel at nongnu.org
> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>



-- 
Amar Tumballi
http://amar.80x25.org
[bulde on #gluster/irc.gnu.org]