[Gluster-devel] Regression tests: Should we test non-XFS too?

[James]

On Mon, 2014-04-14 at 15:50 -0400, Ric Wheeler wrote:
> On 04/07/2014 08:55 AM, Kaleb S. KEITHLEY wrote:
> > On 04/05/2014 01:23 PM, Vijay Bellur wrote:
> >> On 04/05/2014 10:27 PM, Justin Clift wrote:
> >>> On 05/04/2014, at 5:17 PM, James wrote:
> >>>> On Sat, Apr 5, 2014 at 11:37 AM, Justin Clift <justin at gluster.org>
> >>>> wrote:
> >>>>> So, are people interested in us running the tests on other
> >>>>> brick filesystem types, such as ext4? (or whatever else)
> >>>>
> >>>> Yes, absolutely, but I think it's btrfs that will matter, not ext4.
> >>>
> >>>
> >>> Cool.  Running the tests on CentOS 6.5 VM's at the moment, so will need
> >>> to investigate btrfs for that. :)
> >>
> >> We need to increase the number of variables for our regression runs.
> >> Hence running on both ext4 and btrfs would be nice to have.
> >
> > Seconded. I almost hesitate to say it, but we should probably test on zfs too.
> 
> The ZFS kernel modules cannot be legally shipped with either RHEL or CentOS - 
> end users (according to some lawyers) can download and run it themselves.
> 
> Testing ZFS it iffy for those of us at Red Hat at least :)
> 
> >
> >>
> >>>
> >>> If btrfs on CentOS 6.5 isn't a go-er (no idea), we'll need to get the
> >>> tests running cleanly on something where it is.  Maybe Fedora 20?
> >>> (again, no idea, would have to check) :)
> >>
> >> btrfs on Fedora 20 is probably a better bet than CentOS 6.5.
> >>
> >
> > btrfs is in the kernel. There's a 
> > /lib/modules/2.6.32-431.11.2.el6.centos.plus.x86_64/kernel/fs/btrfs/btrfs.ko 
> > on my CentOS 6.5 box. You can get btrfs-progs from epel.
> >
> > A cursory look at the diff between the RHEL6.5 .../fs/btrfs source and the 
> > Fedora .../fs/btrfs source doesn't look too different; if we're worried about 
> > the relative stability of btrfs. It's my understanding that it was only the 
> > lack of an fsck utility that was hindering its adoption in RHEL.
> >
> > I'd say we ought to test on both RHEL/CentOS and Fedora. With our limited 
> > resources we just need to prioritize accordingly.
> >
> > And if there's some concern about the diff between btrfs-progs-0.20.0 in 
> > RHEL/CentOS versus btrfs-progs-3.12-1 in Fedora, I suspect we can build 
> > btrfs-progs-3.12-1 for RHEL and CentOS.
> >
> 
> I would use a recent Fedora or (eventually) RHEL7 versions of the kernel to test 
> btrfs. The btrfs code in Fedora (and RHEL7) tracks upstream well, RHEL6/CentOS6 
> is pretty old and rife with bugs (data loss/performance/you name it).
> 
> Btrfs is tech preview in RHEL6.x so you should not need to get anything from 
> EPEL - the utilities should be there.

So your message motivated me to start working on a patch for btrfs in
Puppet-Gluster. I think this is an easy way to get btrfs+gluster in the
hands of users, testers, developers. I made a feature bug:

https://bugzilla.redhat.com/show_bug.cgi?id=1094860

with WIP patch:
https://github.com/purpleidea/puppet-gluster/tree/feat/btrfs

Turns out choosing UUID's is missing in btrfs:
https://bugzilla.redhat.com/show_bug.cgi?id=1094857

which blocks this for now. Would love to know if you know the correct
person to address this missing feature.

Note hacking on this is now quite easy with vagrant-libvirt thanks to:
https://github.com/purpleidea/vagrant-libvirt/commit/be2042537e4f8f3e59a7880bf3e09358252be252
https://github.com/purpleidea/vagrant-libvirt/commit/72fbedaaca00302b433a2cbb25983b036f9a9619

HTH,
James

> 
> Testing ext4 on RHEL6.x/CentOS should be fine though,
> 
> Ric
> 
> 
> 
> 
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at nongnu.org
> https://lists.nongnu.org/mailman/listinfo/gluster-devel

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20140506/83c3f091/attachment-0002.sig>