On Mon, Oct 07, 2013 at 06:09:21AM -0400, Anjana Kar wrote: > Is it possible to configure the MDT as ldiskfs and the OSTs with zfs > in lustre 2.4? The server is running a lustre kernel on a Centos 6.4 > system, has both lustre-osd-ldiskfs and lustre-osd-zfs rpms installed. > The MDT is up as ldiskfs, but get an error trying to configure the ost: > > mkfs.lustre --fsname=lustrefs --reformat --ost --backfstype=zfs ..... > > mkfs.lustre FATAL: unable to prepare backend (22) > mkfs.lustre: exiting with 22 (Invalid argument)
I suspect that the mkfs.lustre error is unrelated to the presence of an ldiskfs MDT. We need to see the complete command line parameters and output to help you diagnose that problem. It's often helpful to try creating the pool and dataset "by hand" with zpool create and zfs create, using the same names and devices. It may reveal the problem with an error message that is masked by mkfs.lustre. Also check dmesg to see if there was an error loading the ZFS module stack. Having said all that, as Andreas said, you may run into other problems running the ZFS and ldiskfs OSDs concurrently on the same node. Using separate MDS and OSS nodes is more likely to work, but this configuration has not seen much testing. So please report on your experience if you go this route. Thanks, Ned _______________________________________________ Lustre-discuss mailing list [email protected] http://lists.lustre.org/mailman/listinfo/lustre-discuss
