The xentools (which can be more restrictively referred to as the guest agent) are usually independent of the PV drivers. Most modern linux kernels include PV drivers - see http://wiki.xen.org/wiki/Xen_Linux_PV_on_HVM_drivers, so if you boot them in HVM mode they will negotiate up to use the PV drivers. This isn't true for Windows and therefore you have to install the drivers on Windows - which are provided packaged with the guest agent too.
There are many ways to confirm that the guest is using PV drivers - such as running "dmesg | egrep -i 'xen|front'. This should then show lines such as "Initialising Xen virtual ethernet driver". If you have any questions, let us know! Thanks, Bob > -----Original Message----- > From: Bryan Solie [mailto:bso...@gmail.com] On Behalf Of Bryan Solie > Sent: 06 August 2013 14:30 > To: Bob Ball > Cc: openstack@lists.openstack.org; #OpenStack External Email > Subject: Re: [Openstack] Cinder basic questions > > Thanks Bob. > > I will monkey with the iscsi a little more to understand how the xenserver > relationship works. Thanks for the leads. > > But one followup question: So all I need is xentools installed on the guest > for it to be considered paravirtualized? > > Bryan > > On Aug 6, 2013, at 8:47 AM, Bob Ball <bob.b...@citrix.com> wrote: > > > Hi Bryan, > > > > The volumes should appear in the guest under /dev/xvd<num> when they > are added. > > > > As long as the guest has PV drivers it should work - so HVM guests (such as > Windows or if you have installed Linux as HVM) with PV drivers will also see > these volumes added (although clearly with Windows they don't show up > under /dev!) > > > > If using iSCSI then the target should appear on the xenserver host as a PBD > - so "xe pbd-list" should show one up. The code that creates this is > https://github.com/openstack/nova/blob/master/nova/virt/xenapi/volume > _utils.py#L51 - although that's not too obvious because we actually create a > XenServer SR and allow that to create the iscsi connection. > > > > You're right that xensm isn't supported in the current code and the docs > need to be updated for Havana, but you don't have to use XenAPINFS - that > SR is primarily for setups using host aggregates where you want to do in- > pool migration between hosts. > > > > Bob > > > >> -----Original Message----- > >> From: Bryan Solie [mailto:bso...@well.com] > >> Sent: 06 August 2013 13:17 > >> To: openstack@lists.openstack.org > >> Subject: [Openstack] Cinder basic questions > >> > >> I am running OpenStack grizzly on Ubuntu 12.04 with XenServer. > >> > >> I have tried configuring Cinder with a few different backends, but am not > >> able to get the end-to-end attachment process working. > >> > >> I have tried LVM/iscsi driver: The volumes can be created. I attach them, > >> and it returns successfully. There is an iscsi target on the Cinder host. > There > >> is no sign that iscsi target is logged in from the instance I have > >> attached it > to, > >> and there's no indication from the CLI that it is actually attached to the > >> instance. > >> > >> I have tried it with xensm storage backend, which I found in the Grizzly > docs. > >> But this backend no longer seems supported by the code. Not sure if this > is > >> out of sync somehow; I don't know the history. > >> > >> I have tried it with XenapiNFS, and again, the volumes seem to be > created, > >> and I attach them without error, but they are not obviously visible from > the > >> instance. I can't find an iscsi target when I create the volume in this > >> case; > I > >> don't know whether there is supposed to be one with this driver or not. > >> Code suggests no, but I am a Python newbie. > >> > >> The one that seems most successful for me is the LVM/iscsi driver, but > how > >> do I get the volume to automatically login to the created iscsi target? > >> > >> Is there a requirement to use XenapiNFS if I am on xenserver? What are > the > >> advantages? And again, how do I get created volumes to be visible from > >> inside the instance? > >> > >> The doc says with xenserver I can attach volumes only to paravirtualized > >> guests. I have looked at the PV-bootloader for the instances I have > created, > >> and it is blank. As I understand it, they need to be on pygrub to be > >> paravirtualized? Again, I am a little new on this, and I have not been > >> able > to > >> get much out of google. > >> > >> Any help would be appreciated. > >> > >> Bryan > >> _______________________________________________ > >> Mailing list: http://lists.openstack.org/cgi- > bin/mailman/listinfo/openstack > >> Post to : openstack@lists.openstack.org > >> Unsubscribe : http://lists.openstack.org/cgi- > >> bin/mailman/listinfo/openstack _______________________________________________ Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack Post to : openstack@lists.openstack.org Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
