Re: [ovirt-users] gluster performance oVirt 3.4

Underlying FS is XFS GlusterFS - glusterfs-3.5.0-2.el6 I'm using CentOS, if this is problem could RHEL packages be used? why CentOS packages are compiled without native glusterfs support? Citējot Paul Robert Marino <prmarino1@gmail.com> :
What's the underlying filesystem for gluster is it XFS? What version of gluster are you using? What distro are you using and if its not RHEL or Fedora are you using a version of QEMU with gluster support compiled in keep in mind the versions with CentOS and Scientific Linux do not include Gluster native support compiled in.
-- Sent from my HP Pre3
On May 11, 2014 5:40, Vadims Korsaks <truckk@inbox.lv> wrote:
Citējot Vijay Bellur <vbellur@redhat.com> :
On 05/11/2014 02:04 AM, Vadims Korsaks wrote:
HI!
Created 2 node setup with oVirt 3.4 and CentOS 6.5, for storage created 2 node replicated gluster (3.5) fs on same hosts with oVirt. mount looks like this: 127.0.0.1:/gluster01 on
/rhev/data-center/mnt/glusterSD/127.0.0.1:_gluster01
type fuse.glusterfs
(rw,default_permissions,allow_other,max_read=131072)
when i making gluster test with dd, something
like
dd if=/dev/zero bs=1M count=20000
of=/rhev/data-center/mnt/glusterSD/127.0.0.1\:_gluster01/kaka
i'm gettting speed ~ 110 MB/s, so this is 1Gbps speed of ethernet adapter
but with in VM created in oVirt speed is lower than 20 MB/s
why there is so huge difference? how can improve VMs disks speed?
What are your gluster volume settings? Have you applied the following performance tunables in gluster's virt profile:
eager-lock=enable remote-dio=enable
Regards, Vijay
setting were: [root@centos155 ~]# gluster volume info gluster01
Volume Name: gluster01 Type: Replicate Volume ID: 436edaa3-ac8b-421f-aa35-68b5bd7064b6 Status: Started Number of Bricks: 1 x 2 = 2 Transport-type: tcp Bricks: Brick1: 10.2.75.152:/mnt/gluster01/brick Brick2: 10.2.75.155:/mnt/gluster01/brick Options Reconfigured: storage.owner-gid: 36 storage.owner-uid: 36
add your settings settings now it looks
[root@centos155 ~]# gluster volume info gluster01
Volume Name: gluster01 Type: Replicate Volume ID: 436edaa3-ac8b-421f-aa35-68b5bd7064b6 Status: Started Number of Bricks: 1 x 2 = 2 Transport-type: tcp Bricks: Brick1: 10.2.75.152:/mnt/gluster01/brick Brick2: 10.2.75.155:/mnt/gluster01/brick Options Reconfigured: network.remote-dio: enable cluster.eager-lock: enable storage.owner-gid: 36 storage.owner-uid: 36
but this didn't affected performace in any big way should hosts to be restarted?
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users

I've got the same issue using the same versions as Korsaks, using CentOS 6.5, too. I could launch vms with gluster block backend driver using the standard package of libvirt from centos repo. but ovirt does not run the VM with the qemu GFAPI integration, the path of the disk using the fuse mount. <disk type='file' device='disk' snapshot='no'> <driver name='qemu' type='raw' cache='none' error_policy='stop' io='threads'/> <source file='/rhev/data-center/mnt/glusterSD/localhost:VMDATA/0add493f-0a7f-4b32-bcd5-ff25ca504b8b/images/68dbbc67-ea24-45a9-8727-4f85d100d1bb/8fe386f7-2aeb-43c2-bcb0-f76829c876b4'> <seclabel model='selinux' relabel='no'/> </source> <target dev='vda' bus='virtio'/> <serial></serial> <alias name='virtio-disk0'/> <address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x0'/> </disk> or am i wrong? On Mon, May 12, 2014 at 7:41 AM, Vadims Korsaks <truckk@inbox.lv> wrote:
Underlying FS is XFS GlusterFS - glusterfs-3.5.0-2.el6 I'm using CentOS, if this is problem could RHEL packages be used? why CentOS packages are compiled without native glusterfs support?
Citējot Paul Robert Marino <prmarino1@gmail.com> :
What's the underlying filesystem for gluster is it XFS? What version of gluster are you using? What distro are you using and if its not RHEL or Fedora are you using a version of QEMU with gluster support compiled in keep in mind the versions with CentOS and Scientific Linux do not include Gluster native support compiled in.
-- Sent from my HP Pre3
On May 11, 2014 5:40, Vadims Korsaks <truckk@inbox.lv> wrote:
Citējot Vijay Bellur <vbellur@redhat.com> :
On 05/11/2014 02:04 AM, Vadims Korsaks wrote:
HI!
Created 2 node setup with oVirt 3.4 and CentOS 6.5, for storage created 2 node replicated gluster (3.5) fs on same hosts with oVirt. mount looks like this: 127.0.0.1:/gluster01 on
/rhev/data-center/mnt/glusterSD/127.0.0.1:_gluster01
type fuse.glusterfs
(rw,default_permissions,allow_other,max_read=131072)
when i making gluster test with dd, something
like
dd if=/dev/zero bs=1M count=20000
of=/rhev/data-center/mnt/glusterSD/127.0.0.1\:_gluster01/kaka
i'm gettting speed ~ 110 MB/s, so this is 1Gbps speed of ethernet adapter
but with in VM created in oVirt speed is lower than 20 MB/s
why there is so huge difference? how can improve VMs disks speed?
What are your gluster volume settings? Have you applied the following performance tunables in gluster's virt profile:
eager-lock=enable remote-dio=enable
Regards, Vijay
setting were: [root@centos155 ~]# gluster volume info gluster01
Volume Name: gluster01 Type: Replicate Volume ID: 436edaa3-ac8b-421f-aa35-68b5bd7064b6 Status: Started Number of Bricks: 1 x 2 = 2 Transport-type: tcp Bricks: Brick1: 10.2.75.152:/mnt/gluster01/brick Brick2: 10.2.75.155:/mnt/gluster01/brick Options Reconfigured: storage.owner-gid: 36 storage.owner-uid: 36
add your settings settings now it looks
[root@centos155 ~]# gluster volume info gluster01
Volume Name: gluster01 Type: Replicate Volume ID: 436edaa3-ac8b-421f-aa35-68b5bd7064b6 Status: Started Number of Bricks: 1 x 2 = 2 Transport-type: tcp Bricks: Brick1: 10.2.75.152:/mnt/gluster01/brick Brick2: 10.2.75.155:/mnt/gluster01/brick Options Reconfigured: network.remote-dio: enable cluster.eager-lock: enable storage.owner-gid: 36 storage.owner-uid: 36
but this didn't affected performace in any big way should hosts to be restarted?
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users

On Mon, May 12, 2014 at 2:58 PM, Tobias Honacker <tobias@honacker.info>wrote:
I've got the same issue using the same versions as Korsaks, using CentOS 6.5, too. I could launch vms with gluster block backend driver using the standard package of libvirt from centos repo. but ovirt does not run the VM with the qemu GFAPI integration, the path of the disk using the fuse mount.
[snip]
or am i wrong?
This same question asked many times latest months. If you search through archives we remained here: http://lists.ovirt.org/pipermail/users/2014-April/023825.html BTW, Itamar, are there any bug trackers for this still remaining problems with live migration? Could we put anything on the wiki? If so, where, so that I can fill this up? Gianluca

On 05/12/2014 09:09 AM, Gianluca Cecchi wrote:
On Mon, May 12, 2014 at 2:58 PM, Tobias Honacker <tobias@honacker.info <mailto:tobias@honacker.info>> wrote:
I've got the same issue using the same versions as Korsaks, using CentOS 6.5, too. I could launch vms with gluster block backend driver using the standard package of libvirt from centos repo. but ovirt does not run the VM with the qemu GFAPI integration, the path of the disk using the fuse mount.
[snip]
or am i wrong?
This same question asked many times latest months. If you search through archives we remained here: http://lists.ovirt.org/pipermail/users/2014-April/023825.html
BTW, Itamar, are there any bug trackers for this still remaining problems with live migration? Could we put anything on the wiki? If so, where, so that I can fill this up?
I'd put it under ovirt.org/Troubleshooting (I think we have such a page), pointing to "Troubleshooting_LiveMigration" page

<span style=3D"font-family:Prelude, Verdana, san-serif;"><br><br></span><s=
--Alternative_=_Boundary_=_1399919906 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On the Gluster site there is a QEMU repo I'll send you the link latter toda= y. But essentially CentOS doesn't include the Gluster client or libraries s= o they can't compile against it.<br>I'm not sure why CentOS and Scientific = Linux don't include it but I think it might just be an over site since Glus= ter 3.4 is new set of packages which were added in the latest release of RH= EL 6.<br>In truth you can simply unpack the source RPM and rebuild it with = Gluster support its not difficult but I don't remember the flags you need t= o pass the rpmbuild command off the top of my head. That said I'd probably = use the ones off the Gluster site any way. <br>If you are using G= luster 3.5 the ones included in RHEL are incompatible because Gluster still= has large API changes between minor releases and they were compiled agains= t 3.4. <br><br>Also Gluster 3.5 is a brand new release so I wouldn't r= ule out the possibility of a bug. So if the Gluster enabled QEMU = rpms don't help you may need to inquire on the Gluster mailing list.<br><br= pan id=3D"signature"><div style=3D"font-family: arial, sans-serif; font-siz= e: 12px;color: #999999;">-- Sent from my HP Pre3</div><br></span><span styl= e=3D"color:navy; font-family:Prelude, Verdana, san-serif; "><hr align=3D"le= ft" style=3D"width:75%">On May 12, 2014 8:58, Tobias Honacker <tobias@ho= nacker.info> wrote: <br><br></span><div dir=3D"ltr">I've got the sam= e issue using the same versions as Korsaks, using CentOS 6.5, too.<div>I co= uld launch vms with gluster=C2=A0<span style=3D"color:rgb(51,51,51);font-fa= mily:arial,sans-serif;font-size:13px;line-height:17px">block backend driver= =C2=A0using the standard package of libvirt from centos repo.</span></div> <div><font color=3D"#333333" face=3D"arial, sans-serif"><span style=3D"line= -height:17px">but ovirt does not run the VM with the qemu=C2=A0</span></fon= t><span style=3D"color:rgb(0,0,0);font-family:Helvetica,Arial,sans-serif;fo= nt-size:12px;line-height:18px">GFAPI integration, the path of the disk usin= g the fuse mount.</span></div> <div><font color=3D"#000000" face=3D"Helvetica, Arial, sans-serif"><span st= yle=3D"font-size:12px;line-height:18px"><br></span></font></div><div><font = color=3D"#000000" face=3D"Helvetica, Arial, sans-serif"><span style=3D"font= -size:12px;line-height:18px"><div> <disk type=3D'file' device=3D'disk' snapshot=3D'no= 39;></div><div>=C2=A0 =C2=A0 =C2=A0 <driver name=3D'qemu' typ= e=3D'raw' cache=3D'none' error_policy=3D'stop' io= =3D'threads'/></div> <div>=C2=A0 =C2=A0 =C2=A0 <source file=3D'/rhev/data-center/mnt/glus= terSD/localhost:VMDATA/0add493f-0a7f-4b32-bcd5-ff25ca504b8b/images/68dbbc67= -ea24-45a9-8727-4f85d100d1bb/8fe386f7-2aeb-43c2-bcb0-f76829c876b4'><= /div><div>=C2=A0 =C2=A0 =C2=A0 =C2=A0 <seclabel model=3D'selinux'= ; relabel=3D'no'/></div> <div>=C2=A0 =C2=A0 =C2=A0 </source></div><div>=C2=A0 =C2=A0 =C2=A0 &l= t;target dev=3D'vda' bus=3D'virtio'/></div><div>=C2=A0 = =C2=A0 =C2=A0 <serial></serial></div><div>=C2=A0 =C2=A0 =C2=A0 = <alias name=3D'virtio-disk0'/></div><div>=C2=A0 =C2=A0 =C2=A0= <address type=3D'pci' domain=3D'0x0000' bus=3D'0x00= ' slot=3D'0x06' function=3D'0x0'/></div> <div>=C2=A0 =C2=A0 </disk></div></span></font><div><br></div></div><d= iv><br></div><div>or am i wrong?</div><div class=3D"gmail_extra"><br><br><d= iv class=3D"gmail_quote">On Mon, May 12, 2014 at 7:41 AM, Vadims Korsaks <s= pan dir=3D"ltr"><<a href=3D"mailto:truckk@inbox.lv" target=3D"_blank">tr= uckk@inbox.lv</a>></span> wrote:<br> <blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1p= x #ccc solid;padding-left:1ex">Underlying FS is XFS<br> GlusterFS - glusterfs-3.5.0-2.el6<br> I'm using CentOS, if this is problem could RHEL<br> packages be used? why CentOS packages are compiled<br> without native glusterfs support?<br> <br> Cit=C4=93jot Paul Robert Marino <<a href=3D"mailto:prmarino1@gmail.com" = target=3D"_blank">prmarino1@gmail.com</a>> :<br> <div>> =C2=A0What's the underlying filesystem for gluster is<br> it XFS?<br> > =C2=A0What version of gluster are you using?<br> > =C2=A0What distro are you using and if its not RHEL<br> or Fedora are you using a version of QEMU with<br> gluster support compiled in keep in mind the<br> versions with CentOS and Scientific Linux do not<br> include Gluster native support compiled in.<br> ><br> ><br> ><br> > =C2=A0-- Sent from my HP Pre3<br> ><br> </div><div><div>> =C2=A0 On May 11, 2014 5:40, Vadims Korsaks<br> <<a href=3D"mailto:truckk@inbox.lv" target=3D"_blank">truckk@inbox.lv</a=
> wrote:<br> ><br> > =C2=A0Cit=C4=93jot Vijay Bellur <<a href=3D"mailto:vbellur@redhat.c= om" target=3D"_blank">vbellur@redhat.com</a>> :<br> > =C2=A0> On 05/11/2014 02:04 AM, Vadims Korsaks wrote:<br> > =C2=A0> > HI!<br> > =C2=A0> ><br> > =C2=A0> > Created 2 node setup with oVirt 3.4 and<br> > =C2=A0CentOS 6.5, for storage created<br> > =C2=A0> > 2 node replicated gluster (3.5) fs on same<br> > =C2=A0hosts with oVirt.<br> > =C2=A0> > mount looks like this:<br> > =C2=A0> > 127.0.0.1:/gluster01 on<br> > =C2=A0> ><br> ><br> /rhev/data-center/mnt/glusterSD/127.0.0.1:_gluster01<br> > =C2=A0type fuse.glusterfs<br> > =C2=A0> ><br> ><br> (rw,default_permissions,allow_other,max_read=3D131072)<br> > =C2=A0> ><br> > =C2=A0> > when i making gluster test with dd, something<br> > =C2=A0like<br> > =C2=A0> > dd if=3D/dev/zero bs=3D1M count=3D20000<br> > =C2=A0> ><br> ><br> of=3D/rhev/data-center/mnt/glusterSD/<a href=3D"http://127.0.0.1" target=3D= "_blank">127.0.0.1</a>\:_gluster01/kaka<br> > =C2=A0> > i'm gettting speed ~ 110 MB/s, so this is<br> > =C2=A01Gbps speed of ethernet adapter<br> > =C2=A0> ><br> > =C2=A0> > but with in VM created in oVirt speed is<br> > =C2=A0lower than 20 MB/s<br> > =C2=A0> ><br> > =C2=A0> > why there is so huge difference?<br> > =C2=A0> > how can improve VMs disks speed?<br> > =C2=A0> ><br> > =C2=A0><br> > =C2=A0> What are your gluster volume settings? Have you<br> > =C2=A0applied the following<br> > =C2=A0> performance tunables in gluster's virt profile:<br> > =C2=A0><br> > =C2=A0> eager-lock=3Denable<br> > =C2=A0> remote-dio=3Denable<br> > =C2=A0><br> > =C2=A0> Regards,<br> > =C2=A0> Vijay<br> > =C2=A0><br> > =C2=A0setting were:<br> > =C2=A0[root@centos155 ~]# gluster volume info gluster01<br> ><br> > =C2=A0Volume Name: gluster01<br> > =C2=A0Type: Replicate<br> > =C2=A0Volume ID: 436edaa3-ac8b-421f-aa35-68b5bd7064b6<br> > =C2=A0Status: Started<br> > =C2=A0Number of Bricks: 1 x 2 =3D 2<br> > =C2=A0Transport-type: tcp<br> > =C2=A0Bricks:<br> > =C2=A0Brick1: 10.2.75.152:/mnt/gluster01/brick<br> > =C2=A0Brick2: 10.2.75.155:/mnt/gluster01/brick<br> > =C2=A0Options Reconfigured:<br> > =C2=A0storage.owner-gid: 36<br> > =C2=A0storage.owner-uid: 36<br> ><br> ><br> > =C2=A0add your settings settings now it looks<br> ><br> > =C2=A0[root@centos155 ~]# gluster volume info gluster01<br> ><br> > =C2=A0Volume Name: gluster01<br> > =C2=A0Type: Replicate<br> > =C2=A0Volume ID: 436edaa3-ac8b-421f-aa35-68b5bd7064b6<br> > =C2=A0Status: Started<br> > =C2=A0Number of Bricks: 1 x 2 =3D 2<br> > =C2=A0Transport-type: tcp<br> > =C2=A0Bricks:<br> > =C2=A0Brick1: 10.2.75.152:/mnt/gluster01/brick<br> > =C2=A0Brick2: 10.2.75.155:/mnt/gluster01/brick<br> > =C2=A0Options Reconfigured:<br> > =C2=A0network.remote-dio: enable<br> > =C2=A0cluster.eager-lock: enable<br> > =C2=A0storage.owner-gid: 36<br> > =C2=A0storage.owner-uid: 36<br> ><br> ><br> > =C2=A0but this didn't affected performace in any big way<br> > =C2=A0should hosts to be restarted?<br> ><br> > =C2=A0_______________________________________________<br> > =C2=A0Users mailing list<br> > =C2=A0<a href=3D"mailto:Users@ovirt.org" target=3D"_blank">Users@ovirt= =2Eorg</a><br> > =C2=A0<a href=3D"http://lists.ovirt.org/mailman/listinfo/users" target= =3D"_blank">http://lists.ovirt.org/mailman/listinfo/users</a><br> <br> _______________________________________________<br> Users mailing list<br> <a href=3D"mailto:Users@ovirt.org" target=3D"_blank">Users@ovirt.org</a><br=
<a href=3D"http://lists.ovirt.org/mailman/listinfo/users" target=3D"_blank"=
http://lists.ovirt.org/mailman/listinfo/users</a><br> </div></div></blockquote></div><br></div></div>
--Alternative_=_Boundary_=_1399919906--

On Mon, May 12, 2014 at 8:38 PM, Paul Robert Marino <prmarino1@gmail.com>wrote:
On the Gluster site there is a QEMU repo I'll send you the link latter today. But essentially CentOS doesn't include the Gluster client or libraries so they can't compile against it. I'm not sure why CentOS and Scientific Linux don't include it but I think it might just be an over site since Gluster 3.4 is new set of packages which were added in the latest release of RHEL 6. In truth you can simply unpack the source RPM and rebuild it with Gluster support its not difficult but I don't remember the flags you need to pass the rpmbuild command off the top of my head. That said I'd probably use the ones off the Gluster site any way. If you are using Gluster 3.5 the ones included in RHEL are incompatible because Gluster still has large API changes between minor releases and they were compiled against 3.4.
Also Gluster 3.5 is a brand new release so I wouldn't rule out the possibility of a bug. So if the Gluster enabled QEMU rpms don't help you may need to inquire on the Gluster mailing list.
If I remember correctly, due to problems with snapshots (now solved I think) and with live migration (not yet solved) the developers hard coded the fact that the xml domain definition is the suboptimal one. So, even if you change gluster repo, the qemu command line generated by oVirt will be the same, so without the use of libgfapi. Gianluca

----- Original Message -----
From: "Gianluca Cecchi" <gianluca.cecchi@gmail.com> To: "Paul Robert Marino" <prmarino1@gmail.com> Cc: "users" <users@ovirt.org> Sent: Monday, May 12, 2014 5:33:41 PM Subject: Re: [ovirt-users] gluster performance oVirt 3.4
On Mon, May 12, 2014 at 8:38 PM, Paul Robert Marino <prmarino1@gmail.com>wrote:
On the Gluster site there is a QEMU repo I'll send you the link latter today. But essentially CentOS doesn't include the Gluster client or libraries so they can't compile against it. I'm not sure why CentOS and Scientific Linux don't include it but I think it might just be an over site since Gluster 3.4 is new set of packages which were added in the latest release of RHEL 6. In truth you can simply unpack the source RPM and rebuild it with Gluster support its not difficult but I don't remember the flags you need to pass the rpmbuild command off the top of my head. That said I'd probably use the ones off the Gluster site any way. If you are using Gluster 3.5 the ones included in RHEL are incompatible because Gluster still has large API changes between minor releases and they were compiled against 3.4.
Also Gluster 3.5 is a brand new release so I wouldn't rule out the possibility of a bug. So if the Gluster enabled QEMU rpms don't help you may need to inquire on the Gluster mailing list.
If I remember correctly, due to problems with snapshots (now solved I think) and with live migration (not yet solved) the developers hard coded the fact that the xml domain definition is the suboptimal one. So, even if you change gluster repo, the qemu command line generated by oVirt will be the same, so without the use of libgfapi.
I recently tested with F20 hosts and virt pkgs from the fedora-virt-preview repo, and the libgfapi support remained disabled. The changelog for libvirt 1.2.2 mentions a gluster/qemu/snapshot fix, I'd hoped this would do the trick, but I couldn't figure out how to re-enable libgfapi to test it... Jason
Gianluca
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
participants (6)
-
Gianluca Cecchi
-
Itamar Heim
-
Jason Brooks
-
Paul Robert Marino
-
Tobias Honacker
-
Vadims Korsaks