Fedora QA Meeting -- no meeting scheduled

Adam Miller maxamillion at fedoraproject.org
Tue Dec 22 19:36:02 UTC 2009

On Tue, Dec 22, 2009 at 12:55 PM, James Laska <jlaska at redhat.com> wrote:
> I've not heard of the scratch file solution, that might be something to
> consider.  I'm fairly comfortable with iSCSI, which is surprisingly not
> difficult to setup for testing
> (https://fedoraproject.org/wiki/Scsi-target-utils_Quickstart_Guide).  If
> there are other 'do-it-yourself' solutions that don't require hardware,
> they might be worth exploring.

I apparently goobered my vocab on this one. Its called a "sparse
file." You essentially do the following in order to thin provision
space on your drive:

$ dd if=/dev/zero of=fake_image bs=1 count=0 seek=8G

Once the sparse fie is created, you can use du to verify how much
space is allocated to it:

$ du -sh fake_image
0 fake_image

$ du -sh –apparent-size fake_image
8.0G fake_image

losetup /dev/loop0 fake_image

You can now just export /dev/loop0 as a block device over iscsi using
the usual utilities.


> Hardware is definitely a factor whenever we host install-based test
> events.  For example, the RAID test event wasn't able to draw out the
> hardware raid (or BIOS raid) we had hoped for.  I gather that part of
> the problem is people don't realize whether the hardware on the system
> supports RAID.  If they do, it's probably in production because it's
> expensive.  It also might be their only system, so it might be a bit
> daunting to mess with on your *only* production system.
> In some way, I wish we had more virt-based solutions that would allow
> testers to mimic a true hardware setup (RAID, SANs, FCoE, iSCSI iBFT
> etc...).  While this wouldn't let us explore edge conditions with
> specific hardware setups, it might at least shake out the obvious
> interaction issues.
> Alternatively, are there pockets of interest out there who have
> specialty hardware and would be interested in coming together to
> identify and resolve issues in Fedora?
> Thanks,
> James

VM fake RAID could be an interesting concept. I wonder if we could
ping the KVM guys about this to see if it would be horribly difficult
to pull off or if it was just something that hasn't been implemented
because there hasn't been a "practical" use case for it since you
generally have the RAID under the VMs and not in them.

Food for thought :)


()  ascii ribbon campaign - against html e-mail
/\  www.asciiribbon.org   - against proprietary attachments

More information about the fedora-test-list mailing list