[ClusterLabs] GFS2 + KVM with qcow2 recommended IO mode

FeldHost™ Admin admin at feldhost.cz
Wed Feb 28 10:39:10 EST 2018


Hello, thank you for reply. I use GFS2 on compute nodes and VMs using qcow2 images sits on it.

I just found this https://bugzilla.redhat.com/show_bug.cgi?id=1356632 <https://bugzilla.redhat.com/show_bug.cgi?id=1356632>

but no explanation.

> On 28 Feb 2018, at 16:14, Andrew Price <anprice at redhat.com> wrote:
> 
> On 28/02/18 14:03, FeldHost™ Admin wrote:
>> Hello all,
>> I have GFS2 cluster with KVM using qcow2 images. What is recommended IO mode? threads or native? I will be happy for recommendation and explanation.
> 
> You'd likely get a better answer from the QEMU folks, but:
> 
> As I understand it, with io=native qemu submits IO using OS-based aio (io_submit()) and with io=threads it handles its own async io using threading. To see what works best for you, it might be a good idea to do some perf testing of your intended workload with each i/o mode.
> 
> This talk covers the subject https://www.youtube.com/watch?v=Jx93riUF5_I
> 
> You didn't mention whether you're using qcow2 _for_ gfs2 but be aware that recent versions of libvirt will prevent non-raw images being made shareable due to the potential for corruption. See https://bugzilla.redhat.com/show_bug.cgi?id=1511480 for details.
> 
> Hope this helps,
> 
> Andy

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/users/attachments/20180228/f80c7f2c/attachment-0002.html>


More information about the Users mailing list