Yes, I would just use the stock kernel. The DKMS route is a different aproach to packaging kernel modules. You don't need DKMS if you use the method I suggested. Ideally, to avoid confusion, they would not be housed in the same directory on the download site as the binary rpms.
The DKMS packages basically just contain source code. The kernel modules are compiled at package install time, and then again every time a new kernel is installed. Both methods, binary rpms and DKMS rpms, have their advantages and disadvantages. Sometimes it just comes down to a matter of preference. Chris On 06/26/2016 06:09 PM, Yu Chen wrote: > Hi, Christopher > > Thanks a lot for your help! > > So you are saying, I should just use stock kernel without lustre patch > from now on? And what is that lustre-dkms rpm then? Seems I don't need > it ... ? > > For now, I am using the patched kernel, and rebuilt everything according > to your suggestion, and it worked! But the lustre system mounted as > ready-only on the MDS, is that right? > > I will do more testing tomorrow. > > Thanks again for your time and help, really appreciated! > > Chen > > On Fri, Jun 24, 2016 at 4:15 PM, Christopher J. Morrone > <morro...@llnl.gov <mailto:morro...@llnl.gov>> wrote: > > Yes, it is all a lot harder than it should be at this point. We're > slowly getting in packaging changes that should make it a little easier > in future releases. But fear not, it is doable. > > There are other ways, but here is how I would do it with 2.8: > > Skip the installation of the lustre-patched kernel. It is only actually > only required for ldiskfs testing these days. (If all goes well, there > will be no lustre-patched included with the Lustre 2.9 release packages) > Remove that and just use the stock CentOS 7 kernel. > > Next install zfs (and its associate spl packages). You can find > pre-build binary or DKMS packages here: > > https://github.com/zfsonlinux/zfs/wiki/RHEL-%26-CentOS > > Or you can compile them yourself using the instructions here: > > https://github.com/zfsonlinux/zfs/wiki/Custom-Packages > > We are going to compile lustre next, so you will also need to install > the following zfs development packages: > > libzfs2-devel kmod-zfs-devel kmod-spl-devel. > > Next download the lustre source rpm from here: > > > > https://downloads.hpdd.intel.com/public/lustre/lustre-2.8.0/el7/server/SRPMS/ > > The package name is > "lustre-2.8.0-3.10.0_327.3.1.el7_lustre.x86_64.src.rpm", which makes it > seem like you have to have that specific kernel. But don't worry, that > was just an artifact of the old, broken packaging system. We are hoping > to fix that before the Lustre 2.9 is released. > > Next you should be able to generate binary lustre rpms with this > command: > > rpmbuild --rebuild --with zfs --without ldiskfs > lustre-2.8.0-3.10.0_327.3.1.el7_lustre.x86_64.src.rpm > > I am just guessing that you don't need ldiskfs since you are talking > about zfs. You can decide if you want to compile it or not. > > If you hit any problems let us know. Maybe we'll use this discussion to > work up a new page of documentation in the wiki. > > Chris > > On 06/24/2016 07:48 AM, Yu Chen wrote: > > Hello, > > > > I am new to lustre system. Was trying to setup a lustre on Centos 7 with > > zfs. > > > > When came to the step to start lustre.service from systemctl, got the > > errors: > > .... > > > > ------- > > > > osd_zfs: disagrees about version of symbol dmu_objset_space > > > > osd_zfs: Unknown symbol dmu_objset_space (err -22) > > > > ------ > > > > LustreError: 158-c: Can't load module 'osd-zfs' > > > > LustreError: 48041:0:(genops.c:318:class_newdev()) OBD: unknown type: > > osd-zfs > > > > ..... > > > > > > > > Here is what I have done: > > > > * I installed kernel-3.10.0-327.3.1.el7_lustre.x86_64, > > * Then configured and installed zfs from archive.zfsonlinux.org > <http://archive.zfsonlinux.org> > > <http://archive.zfsonlinux.org>(0.6.5.7-1) > > * Then downloaded and installed lustre rpms from el7.2.1511 from: > > > https://downloads.hpdd.intel.com/public/lustre/latest-feature-release/ > > * Then *zpool*create mdt01-pool /dev/sdb > > * mkfs.lustre --reformat --fsname=lustre01 --mgs --mdt > > --backfstype=zfs --index=0 --mgsnode=10.1.12.1@o2ib > > --servicenode=10.1.12.1@o2ib mdt01-pool/mgsmdt01 > > * echo "mds1 - lustre01-MDT0000 zfs:mdt01-pool/mgsmdt01" > > /etc/ldev.conf > > * echo "options lnet networks=o2ib(ib0)" > > /etc/modprobe.d/lustre.conf > > * systemctl start lustre.service > > > > This is where it failed. > > > > Then I think it might be because of the zfs version. So downloaded > > lustre-2.8.0-3.10.0_327.3.1.el7_lustre.x86_64.src.rpm, and tried to > > rebuild from it. But after rebuilding, there were no lustre-osd-zfs and > > lustre-osd-zfs-mount rpms? > > > > > > Would greatly appreciate any suggestions! > > > > > > Thanks in advance! > > > > > > yc > > > > > > _______________________________________________ > > lustre-discuss mailing list > > lustre-discuss@lists.lustre.org > <mailto:lustre-discuss@lists.lustre.org> > > http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org > > > > _______________________________________________ > lustre-discuss mailing list > lustre-discuss@lists.lustre.org <mailto:lustre-discuss@lists.lustre.org> > http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org > > . _______________________________________________ lustre-discuss mailing list lustre-discuss@lists.lustre.org http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
