Hi guys, i''ve just created/ran my first virtual machine on/from CentOS 5.4 Dom0. I installed it onto drbd device, using an Debian iso image from http://stacklet.com. I''ve also extracted the kernel and initrd out to my Dom0 and configured it like this: kernel = "/bootvirts/vmlinuz-2.6.26-2-xen-amd64" ramdisk = "/bootvirts/initrd.img-2.6.26-2-xen-amd64" disk = [''drbd:r0,sda1,w''] root = "/dev/sda1 ro" What i''m wondering now is, how the kernel updating will work or how it''s supposed to work? User of this virtual machine will be able to update the kernel on drbd device (in his virtual machine) but not in /bootvirts/ on Dom0, as it''s not accessible from within virtual machine. Or am I mistaken? Thank you for your time, Regards, M. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon, Nov 30, 2009 at 9:36 PM, Martin Gombac( <martin@isg.si> wrote:> kernel = "/bootvirts/vmlinuz-2.6.26-2-xen-amd64" > ramdisk = "/bootvirts/initrd.img-2.6.26-2-xen-amd64" > disk = [''drbd:r0,sda1,w''] > root = "/dev/sda1 ro" > > What i''m wondering now is, how the kernel updating will work or how it''s > supposed to work? User of this virtual machine will be able to update the > kernel on drbd device (in his virtual machine) but not in /bootvirts/ on > Dom0, as it''s not accessible from within virtual machine. Or am I mistaken?... and that is exactly why by default RHEL/Centos uses pygrub :) With pygrub/pvgrub, domU kernel is located on domU filesystem, so that users can update it themselves. -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> ... and that is exactly why by default RHEL/Centos uses pygrub :) > With pygrub/pvgrub, domU kernel is located on domU filesystem, so that > users can update it themselves.I didnt get that domU filesystem you mean to say in the installed guest OS _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hi, i don''t think CentOS 5.4 pvgrub/pygrub support drbd disk which is essential for my HA cluster with non interruptive failover. Will look into it. Regards, M. Fajar A. Nugraha wrote:> On Mon, Nov 30, 2009 at 9:36 PM, Martin Gombac( <martin@isg.si> wrote: > >> kernel = "/bootvirts/vmlinuz-2.6.26-2-xen-amd64" >> ramdisk = "/bootvirts/initrd.img-2.6.26-2-xen-amd64" >> disk = [''drbd:r0,sda1,w''] >> root = "/dev/sda1 ro" >> >> What i''m wondering now is, how the kernel updating will work or how it''s >> supposed to work? User of this virtual machine will be able to update the >> kernel on drbd device (in his virtual machine) but not in /bootvirts/ on >> Dom0, as it''s not accessible from within virtual machine. Or am I mistaken? > > ... and that is exactly why by default RHEL/Centos uses pygrub :) > With pygrub/pvgrub, domU kernel is located on domU filesystem, so that > users can update it themselves. >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tue, Dec 1, 2009 at 3:30 PM, Martin Gombac( <martin@isg.si> wrote:> Hi, > > i don''t think CentOS 5.4 pvgrub/pygrub support drbd disk which is essential > for my HA cluster with non interruptive failover. Will look into it.It does not support drbd:/ resource type (at least not last time I check), but it works fine with phy:/dev/drbd/by-res/... (by-res symlinks available starting drbd 8.3). That''s what I currently use. -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Sadly phy:/ is not an option, ''cause using it doesn''t allow live failover/migration of PVs from DRBD. If it did i could have live failover/migration for HVM''s too. That would be so cool. :-) Fajar A. Nugraha wrote:> On Tue, Dec 1, 2009 at 3:30 PM, Martin Gombac( <martin@isg.si> wrote: >> Hi, >> >> i don''t think CentOS 5.4 pvgrub/pygrub support drbd disk which is essential >> for my HA cluster with non interruptive failover. Will look into it. > > It does not support drbd:/ resource type (at least not last time I > check), but it works fine with phy:/dev/drbd/by-res/... (by-res > symlinks available starting drbd 8.3). That''s what I currently use. >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Tue, Dec 1, 2009 at 7:29 PM, Martin Gombac( <martin@isg.si> wrote:> Sadly phy:/ is not an option, ''cause using it doesn''t allow live > failover/migration of PVs from DRBD.Saids who?> If it did i could have live > failover/migration for HVM''s too. That would be so cool. :-)AFAIK the only noticable difference between drbd:/ and phy: is the state of drbd device. On normal circumstances, with drbd:/ one would be primary, the other secondary. This is good because it prevents domU from starting at both dom0 at the same time (which might lead to corruption). However for live migration to work you still need to setup drbd for dual primary mode. With phy:/, both nodes will see the drbd block as primary. You may also need to setup drbd so that it becomes primary on both nodes on startup. Live migration works great. However you need to take extra precaution measures to make sure that at any one time only one dom0 is running a particular domU. -- Fajar> > Fajar A. Nugraha wrote: >> >> On Tue, Dec 1, 2009 at 3:30 PM, Martin Gombac( <martin@isg.si> wrote: >>> >>> Hi, >>> >>> i don''t think CentOS 5.4 pvgrub/pygrub support drbd disk which is >>> essential >>> for my HA cluster with non interruptive failover. Will look into it. >> >> It does not support drbd:/ resource type (at least not last time I >> check), but it works fine with phy:/dev/drbd/by-res/... (by-res >> symlinks available starting drbd 8.3). That''s what I currently use. >> >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Tnx for the info. It seems that i don''t yet know enough about this, so i''ll get back to you later on. :-) I''ll try dual primary mode with phy and HVMs. :-) Regards, M. Fajar A. Nugraha wrote:> On Tue, Dec 1, 2009 at 7:29 PM, Martin Gombac( <martin@isg.si> wrote: >> Sadly phy:/ is not an option, ''cause using it doesn''t allow live >> failover/migration of PVs from DRBD. > > Saids who? > >> If it did i could have live >> failover/migration for HVM''s too. That would be so cool. :-) > > AFAIK the only noticable difference between drbd:/ and phy: is the > state of drbd device. On normal circumstances, with drbd:/ one would > be primary, the other secondary. This is good because it prevents domU > from starting at both dom0 at the same time (which might lead to > corruption). However for live migration to work you still need to > setup drbd for dual primary mode. > > With phy:/, both nodes will see the drbd block as primary. You may > also need to setup drbd so that it becomes primary on both nodes on > startup. Live migration works great. However you need to take extra > precaution measures to make sure that at any one time only one dom0 is > running a particular domU. >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users