Hey Murali, Yeah, just a thought. Oh well. NFSD does a lot of checking with regards to export settings which can drastically change behavior.
Dean On 9/21/07, Murali Vilayannur <[EMAIL PROTECTED]> wrote: > Dean, > I don't think this will have any effect. > NFSd makes use of a set of vfs APIs which invoke appropriate FS driver > callbacks. > NFSd settings won't have any effect on the behavior of the vfs api's though. > thanks, > Murali > > On 9/21/07, Dean Hildebrand <[EMAIL PROTECTED]> wrote: > > One thing to check is how you are exporting PVFS2. The "correct" way is > > to export synchronously, since exporting a file system asynchronously > > gives no integrity guarantees. cat /etc/exports to see how you are > > exporting pvfs2. I don't know if this affects nfsd using aio or not > > though... > > > > Dean > > > > Emmanuel Florac wrote: > > > Le Fri, 21 Sep 2007 07:08:51 -0700 > > > "Murali Vilayannur" <[EMAIL PROTECTED]> écrivait: > > > > > > > > >> I did not realize that there is a need or use case for nr_segs > 1. > > >> > > > > > > Mmmh... Shouldn't "the right behaviour" be falling back ( even > > > silently ) to synchronous IO instead of rejecting the whole IO ? Would > > > it be difficult to do ? I'll look into this :) > > > > > > > > > > > > _______________________________________________ > > Pvfs2-users mailing list > > [email protected] > > http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users > > > _______________________________________________ Pvfs2-users mailing list [email protected] http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
