On Oct 22, 2009, at 7:11 AM, Bernhard Gschaider <bgschaid_lists at ice-sf.at > wrote: > >>>>>> On Tue, 20 Oct 2009 12:36:06 -0700 >>>>>> "AY" == Akemi Yagi <amyagi at gmail.com> wrote: > > AY> On Tue, Oct 20, 2009 at 11:51 AM, Bernhard Gschaider > AY> <bgschaid_lists at ice-sf.at> wrote: > >>> I've got a fileserver currently running under 5.3 with the >>> /home-partition being an XFS-filesystem. I use the kmod-xfs >>> from extras. It works great ;) >>> >>> Now: as I understand it from the release-notes the 5.4 kernel >>> has XFS already built-in. Right? Or is it just a kmod-package >>> ("technology preview") >>> >>> Now my question is: are there any recommendations for an >>> upgrade-procedure? I mean, I can probably manage, but I'll want >>> to minimize downtime > >>> BTW: when doing "yum list updates" I don't see any >>> "kernel*"-packages in the list. Is this because the last kernel >>> from the 5.3-updates has the same build-numer (164 I think)? >>> And is the 5.4-base-kernel the same as the latest >>> 5.3-updates-kernel? > > AY> The -164 kernel is indeed from 5.4 and has xfs as a built-in > AY> kernel module. If you are already running this kernel, that > AY> indicates all is well and no further action is needed. > > AY> Could you show us the output returned by: > > AY> ls -l `find /lib/modules -name xfs.ko` > > Thanks for the hint: I did it (I'll spare you the listing). The > -164-kernel ist the first one where according to "rpm -qf <path>" the > module is "owned" by the kernel package. All the other instances of > xfs.ko point to a module "owned" by kmod-xfs. > > So obviously I'm not using the kmod-xfs anymore (I'm relieved that the > last kernel-update worked without a clash) > > Thanks again for clearing that up > > Bernhard > > BTW: yes. It is a x86_64-machine What version of XFS is supplied? Is it still true that RH doesn't supply xfsprogs and xfsdump so we need to use the ones in 'extras'? If so, wouldn't it be risky using the RH supplied kernel driver with third party supplied maintenance apps in case RH were to backport any XFS internal structs that the user progs don't know how to handle, or vice versa. I think I'll stick with the complete CentOS XFS bundle for a while and see what RH does. -Ross