In fuse work I've done and seen, it is possible to extract decent
performance from fuse, but it isn't free. By default, FUSE only
performs I/O 4K at a time. With the direct_io mount option, you get the
benefit of only a single copy from userspace to userspace effectively
for I/O buffers and t
f getting your email to the list raises the bar...which
> might not be all-bad. :-)
>
> Thanks,
> -sam
>
> On Nov 7, 2007, at 2:39 PM, Bradley W Settlemyer wrote:
>
> > There is a registered user: [EMAIL PROTECTED]
> >
> > The mail sent to walt was from [
There is a registered user: [EMAIL PROTECTED]
The mail sent to walt was from [EMAIL PROTECTED]
I don't have good answers on how to help users debug this problem as it
can be subtle to detect I guess. Mailing the list owner probably won't
help as those accounts are so swamped with spam that my sp
Sam Lang wrote:
How about just add PVFS_sys_dist_get_IO_server_aliases() and
PVFS_sys_dist_set_IO_server_mapping(). Under the hood each
distribution orders the IO servers it'll use according to the
mapping then does business as usual. This instead of the random
starting point and sequential o
Julian Martin Kunkel wrote:
Hi,
So I think we're mostly trying to work out what our API should really
be, whether we should extend the distro functionality vs. going totally
to hints, and if we go to hints what that API should look like, right?
probably the distribution needs a function which i
Murali fixed it. Bad config file. When I was using pvfs2-genconfig I
must have somehow made a dedicated metaserver rather than just 2 data
servers, which is what I was trying to do.
Sorry about the interruption.
Thanks
Brad
Bradley W Settlemyer wrote:
Hello
I'm having a problem
Hello
I'm having a problem with detecting the number of IO servers -- it
doesn't make any sense. Here is a program sample:
[EMAIL PROTECTED]:~:71> software/pvfs2/bin/pvfs2-cp pvfs2.patch
/parl/bradles/software/pvfs2/mnt/foo.3
BWS: num_io: -1075277896
BWS get_num_io: 1
BWS - num_dfiles: 1
(which I just mailed to, but I'm not sure are appropriate for
the main romio tree).
Thanks for the note, I will begin reintegrating with the head now.
Cheers,
Brad
Robert Latham wrote:
On Fri, Jul 21, 2006 at 02:48:49PM -0400, Bradley W Settlemyer wrote:
Hello
The last 3 or 4 times that
Hello
The last 3 or 4 times that I've moved my work into sync with the head
of the source tree, the head has been broken. I want to be a little
more cautious this time with my work -- I find a bunch of bugs, but they
mostly turn out to not be due to my changes, and it takes me several
days
Btw, if we're talking about usability issues as well, I would suggest
getting rid of the home button, and instead making the big image up top
be a link to the homepage. I always expect header images like that to
take me to the 1st page.
Cheers,
Brad
Walt Ligon wrote:
hey guys
Brad has mad
I used it in the server for my thesis also.
Phil Carns wrote:
The server is definitely not using the msgpairarray code right now. I
think it would be handy to keep it in the common area, though, because
it could be used for server to server communication in the future.
It actually started ou
cal to physical offsets exactly like you
want. It's easy enough to check, should just require a single line
change in the sys-create.sm client state machine.
Cheers,
Brad
John Bent wrote:
On Wed, 28 Jun 2006, Bradley W Settlemyer wrote:
Hmm, its an interesting idea. You could try modifyi
So, umm, is CVS broken again?
Todays server log:
[D 06/15 10:37] About to retrieve attributes for handle 1048576
[D 06/15 10:37] (0x893c008) rmdirent (prelude sm) state: perm_check
(status = 0)[D 06/15 10:37] reading dirdata handle (coll_id =
0x781495c5, handle = 1048576, key = de (3), val_b
I've seen similar behavior to this using Myrinet (jazz) also. I haven't
ever gotten it with TCP though, so maybe mine was due to something a
little different (basically, it looked like a message with the tag was
sent, but never delivered). Are you certain that its just a BMI tcp issue?
Cheer
me either. I'll keep debugging.
-sam
On Jun 12, 2006, at 3:19 PM, Bradley W Settlemyer wrote:
It no longer core dumps when I do a pvfs2-ls on a freshly created
filesystem, but I still cannot get pvfs2-ls to function on a just
created file system.
Is it me or you?
Cheers,
brad
ging.
-sam
On Jun 12, 2006, at 3:19 PM, Bradley W Settlemyer wrote:
It no longer core dumps when I do a pvfs2-ls on a freshly created
filesystem, but I still cannot get pvfs2-ls to function on a just
created file system.
Is it me or you?
Cheers,
brad
Sam Lang wrote:
Brad,
I think this
keyval strings. I had the wrong values for the lengths.
I've committed a fix, can you update and try again?
-sam
On Jun 12, 2006, at 2:38 PM, Bradley W Settlemyer wrote:
I'm not certain whether its something I've done or what, but it's
pretty screwed up. I just did a pv
or re-create your storage spaces.
I can't think of anything else that would bite you...the nightly tests
seem to have passed last night...
-sam
On Jun 12, 2006, at 2:03 PM, Bradley W Settlemyer wrote:
I just did a cvs update this morning, not having much luck with the
filesystem
I just did a cvs update this morning, not having much luck with the
filesystem.
Is pvfs2-cp or pvfs2-touch working for anyone else at the moment?
Create is returning me an invalid argument problem.
[D 14:58:55.314670] Handle created: 4294967297
[D 14:58:55.314828] (0x86a2f10) create (FR sm) s
Hello
In trying to set the dfile count, I'm finding that the attribute mask
is set to
attr.mask = 0x3f
during file creation.
I apparently need there to be a 1 in the 7th bit position (<<25) in
order to set the dfile count (based on pvfs2-types.h). I am passing in
the number of dfiles I
Assert is also redefined in that same ib file. So we may think about
changing that one also if the def from assert.h won't cut it, or at
least using a #undefine there.
I also have to tweak the iba headers, but that seems to be limited to
just adding a const to make so messages go away.
Brad
Pete,
It appears that compile errors I'm running into with PVFS2 and
Silverstorm/Mellanox infiniband are definitions in the ib code that
duplicate definitions in kernel headers. Specifically:
#define likely(x) __builtin_expect(!!(x), 1)
#define unlikely(x) __builtin_expect(!!(x),
I'm not certain how to use --with-ib.
So vapi.h is at:
/usr/include/iba/vapi/vapi.h
so I did
setenv CFLAGS "-I/usr/include/iba/vapi"
and then just configured like
./configure --with-ib=/usr
However, the libraries appear to be in /lib64, so I'm still not clear on
what I'm doing. When I try
Hello,
I'm trying to build PVFS2 with IB support using Mellanox/Silverstorm
drivers. Is this a supported configuration?
Cheers,
Brad
___
Pvfs2-developers mailing list
Pvfs2-developers@beowulf-underground.org
http://www.beowulf-underground.org/mail
Hello PVFS/PVFS2 mailing list users,
The mailing lists have been unavailable due to system failures. We
have attempted to restore the mailing lists, however any mail or
administrative requests sent in the last several days may have been
rejected or lost.
Sorry about the inconvenience.
25 matches
Mail list logo