--=_22a8f6a92661e1f6532f09f6ce07eee9 Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=UTF-8 Le 2015-10-12 14:04, Nir Soffer a écrit :
On Mon, Oct 12, 2015 at 11:14 AM, Nico <gluster@distran.org> wrote:
Yes, engine will let you use such volume in 3.5 - this is a bug. In 3.6 you will not be able to use such setup.
replica 2 fails in a very bad way when one brick is down; the application may get stale data, and this breaks sanlock. You will be get stuck with spm that cannot be stopped and other fun stuff.
You don't want to go in this direction, and we will not be able to support that.
For the record, I already rebooted node1; and the node2 took over the existing VM from node 1 and vice-versa. GlusterFS worked fine, oVirt application was still working fine .. i guess it is because it was a soft reboot which stops softly the services. I got another case where i stuck the network on the 2 nodes simultaneously after a bad manipulation on oVirt GUI and i got a split brain. i kept the error at this very moment: root@devnix-virt-master02 nets]# gluster volume heal ovirt info split-brain Brick devnix-virt-master01:/gluster/ovirt/ /d44ee4b0-8d36-467a-9610-c682a618b698/dom_md/ids Number of entries in split-brain: 1 Brick devnix-virt-master02:/gluster/ovirt/ /d44ee4b0-8d36-467a-9610-c682a618b698/dom_md/ids Number of entries in split-brain: 1 This file was having same size on both nodes; so it was hard to select one. Finally i chose the younger one and all was back online after the heal. It is this kind of stuff you are talking about with 2 nodes ? For now, I don't have budget to take a third one; so i'm a bit stuck and disappointing. I've a third device but for backup, it has lot of storage but low cpu abilities (no VT-X) so i can't use it as hypervisor. I could maybe use it as a third brick but is it possible to have this kind of configuration ? 2 actives nodes as hypervisor and 1 third only for gluster replica 3 ? Cheers Nico --=_22a8f6a92661e1f6532f09f6ce07eee9 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=UTF-8 <html><head><meta http-equiv=3D"Content-Type" content=3D"text/html; charset= =3DUTF-8" /></head><body style=3D'font-size: 10pt; font-family: Verdana,Gen= eva,sans-serif'> <p>Le 2015-10-12 14:04, Nir Soffer a écrit :</p> <blockquote type=3D"cite" style=3D"padding: 0 0.4em; border-left: #1010ff 2= px solid; margin: 0"><!-- html ignored --><!-- head ignored --><!-- meta ig= nored --> <div class=3D"pre" style=3D"margin: 0; padding: 0; font-family: monospace"> <blockquote type=3D"cite" style=3D"padding: 0 0.4em; border-left: #1010ff 2= px solid; margin: 0">On Mon, Oct 12, 2015 at 11:14 AM, Nico <<a href=3D"= mailto:gluster@distran.org">gluster@distran.org</a>> wrote:</blockquote> <br /> Yes, engine will let you use such volume in 3.5 - this is a bug. In = 3.6 you will<br /> not be able to use such setup.<br /> <br /> replica 2 fa= ils in a very bad way when one brick is down; the<br /> application may get= <br /> stale data, and this breaks sanlock. You will be get stuck with spm<= br /> that cannot be<br /> stopped and other fun stuff.<br /> <br /> You do= n't want to go in this direction, and we will not be able to support that= =2E<br /> </div> </blockquote> <p> </p> <p>For the record, I already rebooted node1; and the node2 took over the ex= isting VM from node 1 and vice-versa.</p> <p>GlusterFS worked fine, oVirt application was still working fin= e .. i guess it is because it was a soft reboot which stops soft= ly the services.</p> <p>I got another case where i stuck the network on the 2 nodes simultaneous= ly after a bad manipulation on oVirt GUI and i got a split brain.</p> <p>i kept the error at this very moment:</p> <p>root@devnix-virt-master02 nets]# gluster volume heal ovirt info split-br= ain<br />Brick devnix-virt-master01:/gluster/ovirt/<br />/d44ee4b0-8d36-467= a-9610-c682a618b698/dom_md/ids<br />Number of entries in split-brain: 1</p> <p>Brick devnix-virt-master02:/gluster/ovirt/<br />/d44ee4b0-8d36-467a-9610= -c682a618b698/dom_md/ids<br />Number of entries in split-brain: 1</p> <p> </p> <p> </p> <p>This file was having same size on both nodes; so it was hard to select o= ne. Finally i chose the younger one and all was back online after the heal= =2E</p> <p>It is this kind of stuff you are talking about with 2 nodes ?</p> <p> </p> <p>For now, I don't have budget to take a third one; so i'm a bit stuck and= disappointing.</p> <p>I've a third device but for backup, it has lot of storage but low cpu ab= ilities (no VT-X) so i can't use it as hypervisor.</p> <p>I could maybe use it as a third brick but is it possible to have this ki= nd of configuration ? 2 actives nodes as hypervisor and 1 third only for gl= uster replica 3 ?</p> <p>Cheers</p> <p>Nico</p> <p> </p> <p> </p> <div> </div> </body></html> --=_22a8f6a92661e1f6532f09f6ce07eee9--