On 2013-10-03 16:18, Ric Wheeler wrote: > On 10/03/2013 09:17 AM, Christoph Hellwig wrote: >> On Thu, Oct 03, 2013 at 09:12:24AM -0400, Ric Wheeler wrote: >>>>>> Which in-tree or soon in-tree filesystem do you care about? And why >>>>>> don't we see pnfs support for it submitted instead of the fairly useless >>>>>> gfs2 support? >>>> I picked gfs2 as the initial use case for simplicity and ease of review. >>>> If there is a rough consensus that it's useless and not worthy of inclusion >>>> then the one we care about the most is exofs that has a more complete pnfs >>>> implementation. >>>> >>>> Benny >>>> >>> I don't see having GFS2 supported as a base for pNFS as useless. >>> Christoph, is this a concern about GFS2 being too complicated for >>> normal deployment or a lack in the pNFS support on top of it? >> Fairly useless was specific to the particular implementation: >> >> - which in the stipped down version here only supports DS access for >> reads >> - which in the previous version showed worse performance than always >> going through the MDS >> >> I don't have a problem with using GFS2 by itself, but any implementation >> proposed should actually show signifiant real life benefits before it >> gets merged. >> The question is what is the minimum value for submitting upstream... The thing pnfs over dlm/gfs2 is missing mostly is supporting read/write layout. One could use them load balancing, e.g. by either redirecting to a node holding an exclusive lock on the file, if there is one, or dlm_ino_hash in its absence. Benny > > Makes sense, thanks! > > Ric > > -- > To unsubscribe from this list: send the line "unsubscribe linux-nfs" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
