On Wed, 6 Jun 2012, Travis Rhoden wrote: > Thanks folks. > > > * 2.6.18 is crazy-old. > Certainly. I'm hoping I only need to work with CentOS6, then I at > least get 2.6.32. I wouldn't try to do RBD with these kernels, > though. I would be more likely to have an Ubuntu 12.04 node mount an > RBD and export with NFS. But as you mention, that has it's own > issues. > > > Your other options are re-exporting NFS or CIFS, or trying to use ceph-fuse. > ceph-fuse would only be for use with cephfs, no? Or can ceph-fuse > interface to RBD or radosgw? Just the fs. > I'm probably just slightly ahead of the curve here. I am trying to > shoehorn ceph into my scenario (exporting a single filesystem to lots > of cluster nodes) without using cephfs since it's not quite ready. > Even though I lose parallel file access when using NFS (all traffic > would go through that one NFS node), I was hoping that mounting an > RBD, formatting it, and exporting it with NFS would avoid the ESTALE > issues (as opposed to using cephfs, mounting via ceph-fuse, and then > exporting _that_ with NFS). Oh, yeah, exporting an xfs/ext4/whatever on top of rbd via NFS will certainly work! sage > > I actually have a phone call with inktank tomorrow. I'll ask some > more questions then. > > - Travis > > On Wed, Jun 6, 2012 at 12:27 PM, Sage Weil <sage@xxxxxxxxxxx> wrote: > > On Wed, 6 Jun 2012, Travis Rhoden wrote: > >> Sorry for the re-post. First attempt got eaten by the mailing list > >> because it wasn't plain text. My first attempt! > >> > >> > With CephFS not quite ready for production, what would be the > >> > recommended way for sharing Ceph to older operating systems? > >> > Specifically, I want to share a file system (i.e. mount a common > >> > directory) on several CentOS5 and CentOS6 nodes. I don't think I can > >> > mount an RBD from more than one node, so what is my best bet? > >> > > >> > Mount an RBD on a file "head" and export it via NFS? > >> > > >> > I feel like that is the only sane choice, but wanted to ask in case I > >> > am missing something better. > > > > Two things: > > > > * 2.6.18 is crazy-old. It's possible the libceph and rbd bits could be > > merged in, but getting the fs to work with ancient VFS will be a world of > > pain. > > > > * NFS reexport works in basic scenarios, but should not be used in > > production as there are various conditions that can lead to ESTALE. > > Making it work is somewhere on the mid-term roadmap, but it'll be a while. > > > > Your best bet is to run cephfs on a modern kernel. But be warned, we > > aren't ready to support the fs just yet. Soon! > > > > sage > > > > > >> > >> - Travis > >> > > >> > > >> > On Wed, Jun 6, 2012 at 9:58 AM, Sam Zaydel <sam.zaydel@xxxxxxxxxxx> wrote: > >> >> > >> >> I think you will be able to build from source the client bits that you > >> >> need, but I saw this from an earlier similar discussion mentioning > >> >> Python version, so I would make sure that python > 2.5 is used, > >> >> realistically 2.6 should be the norm today. > >> >> > >> >> http://comments.gmane.org/gmane.comp.file-systems.ceph.devel/5717 > >> >> > >> >> On Tue, Jun 5, 2012 at 6:16 PM, <Eric_YH_Chen@xxxxxxxxxx> wrote: > >> >> > Dear All: > >> >> > > >> >> >ÿÿÿÿ My ceph cluster is installed on Ubuntu 12.04 with kernel 3.2.0. > >> >> > > >> >> >ÿÿÿÿ But I have a situation that I may need to mount cephfs on centos 5.7, which with kernel 2.6.18. > >> >> > > >> >> > (that is to say , I only need to execute mount.ceph on the centos 5.7, not the whole ceph system) > >> >> > > >> >> > I want to find a solution that can provide reliable and high availability share file system. > >> >> > > >> >> > Is it possible to do this? Or any other recommend way? Ex: export NFSÿÿ > >> >> > > >> >> > Thanks! > >> >> > > >> >> > > >> >> > -- > >> >> > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > >> >> > the body of a message to majordomo@xxxxxxxxxxxxxxx > >> >> > More majordomo info at http://vger.kernel.org/majordomo-info.html > >> >> > >> >> > >> >> > >> >> -- > >> >> Inked by Sam Zaydel > >> >> -- > >> >> To unsubscribe from this list: send the line "unsubscribe ceph-devel" 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 ceph-devel" 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 ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > >
