It seems I''ve been a bit thick. It''s been pretty obvious recently that Xen isn''t flavour of the month around here, but I assumed there were good reasons for that. Now, rather belatedly, I''ve found http://www.redhat.com/about/news/prarchive/2008/qumranet.html In short, RedHat paid $107 million for Qumranet in September 2008. The acquisition includes KVM. I''ve got 2 years invested in Xen, on FC8, and I can''t help feeling that I''ve been shafted. Am I alone? -Evan
On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote:> It seems I''ve been a bit thick. It''s been pretty obvious recently that > Xen isn''t flavour of the month around here, but I assumed there were > good reasons for that. Now, rather belatedly, I''ve found > > http://www.redhat.com/about/news/prarchive/2008/qumranet.html > > In short, RedHat paid $107 million for Qumranet in September 2008. The > acquisition includes KVM. > > I''ve got 2 years invested in Xen, on FC8, and I can''t help feeling that > I''ve been shafted. Am I alone?Not alone, same situation for me. I am using Xen aprox 2 years too on our university, but with end of F8 there is not a fully functional Xen Dom0 and also DomU kernel for any stable Fedora. KVM is still not a replacement for paravirtualized machines and I think fully virtualized KVM will be slower like a paravirtualized XEN. Also I am missing some howtos for migration to KVM/xenner. Jan ONDREJ (SAL)
On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote:> It seems I''ve been a bit thick. It''s been pretty obvious recently that > Xen isn''t flavour of the month around here, but I assumed there were > good reasons for that. Now, rather belatedly, I''ve found > > http://www.redhat.com/about/news/prarchive/2008/qumranet.html > > In short, RedHat paid $107 million for Qumranet in September 2008. The > acquisition includes KVM. > > I''ve got 2 years invested in Xen, on FC8, and I can''t help feeling that > I''ve been shafted. Am I alone? >Xen (paravirt) is in the process of going into the Mainstream kernel, also RHEL5, which has a number of years left, includes xen - I don''t think Red Hat are going to mess their corporate clients around by removing it. The problem with F8 is that the kernel people could no longer drag an obsolete (2.6.21) kernel around just for xen, and decided to concentrate on helping get it into the mainstream kernel. This has taken longer than expected. RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and Fedora 11 should have, as well. Another thing the Red Hat people are working on is to make the management interface (and the images, I think - please correct me if I''m wrong RH folks) the same for Qemu, KVM and xen, so the impact of moving from one to the other is very small. Shafted?...I don''t think so. We''re just in a blip at the moment. -- Cheers! (Relax...have a homebrew) Neil ...aliquando et insanire iucundum est. -- Lucius Annaeus Seneca
On Wed, Jan 21, 2009 at 11:44:35AM +0200, Neil Thompson wrote:> On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote: > RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and Fedora 11 > should have, as well.Just there is at least one critical problem with Fedora 10 DomU and also one for Fedora 9 DomU, so Fedora 10 kernels can''t run stable. :( https://bugzilla.redhat.com/show_bug.cgi?id=478414 SAL
On Wed, Jan 21, 2009 at 10:51:07AM +0100, Jan ONDREJ (SAL) wrote:> On Wed, Jan 21, 2009 at 11:44:35AM +0200, Neil Thompson wrote: > > On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote: > > RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and Fedora 11 > > should have, as well. > > Just there is at least one critical problem with Fedora 10 DomU and also one > for Fedora 9 DomU, so Fedora 10 kernels can''t run stable. :( > > https://bugzilla.redhat.com/show_bug.cgi?id=478414 >Your problem is with an F8 Dom0...the 2.6.29 kernel that will (hopefully) be in F11 should sort out most of the paravirt problems, from what I can see. I''m not sure what the status of the code in EL5.3 is, maybe the RH people can comment. I don''t think your problem qualifies as "shafted". -- Cheers! (Relax...have a homebrew) Neil ...aliquando et insanire iucundum est. -- Lucius Annaeus Seneca
On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote:> It seems I''ve been a bit thick. It''s been pretty obvious recently that > Xen isn''t flavour of the month around here, but I assumed there were > good reasons for that. Now, rather belatedly, I''ve found > > http://www.redhat.com/about/news/prarchive/2008/qumranet.html > > In short, RedHat paid $107 million for Qumranet in September 2008. The > acquisition includes KVM. > > I''ve got 2 years invested in Xen, on FC8, and I can''t help feeling that > I''ve been shafted. Am I alone?The acqusition of Qumranet has had absolutely *zero* impact on the availability Xen kernels in Fedora. The sole reason for not having Xen support in Fedora 9/10 is that the Dom0 kernel is not yet merged upstream, and this problem existed long long before Qumranet joined Red Hat. When we first shipped Xen in Fedora Core 5 (or was it 4?) none of the Xen code was merged into the mainline Linux kernel tree. For several releases we spent a great deal of time forward porting Xen to newer kernels. When we got to Fedora 9 the guest side was merged into the main kernel, but the host side was not. Unfortunately the Xen host kernel was still on 2.6.18 while Fedora was on 2.6.24 and the kernel was just too old to work with the userspace tools. We did not want to drop Xen Dom0 host from Fedora 9, but we had no viable options to continue with it in the short term. Since that time though, Jeremy Fitzhardinge has done alot of work on getting Dom0 patches in shape for merging in upstream Linux. It it still hard to say just when these will be accepted upstream, but there is a semi-reasonable we''ll be able to turn Xen Dom0 back on in Fedora 11 kernels. While we (Red Hat) think KVM is a very compelling technology, as long as Xen is open source, actively maintained upstream & in mainline Linux kernels, there''s no reason why it shouldn''t be available in Fedora. So once the Dom0 kernel is merged, Fedora users will be able to have a choice between Xen and KVM for many future releases. We have also put effort into developing Xenner which allows paravirt Xen guests to be run under KVM without having to re-configure the guest kernel, giving people a potential migration strategy if they need one. As for RHEL-5, that continues to support Xen, and will do for the entire of its 7 year lifetime. If you don''t want official Red Hat support, there is also the option of using CentOS 5 as a Xen host which again will have Xen support it in for whole of its 7 year lifetime. So while it is definitely unfortunate that we don''t have a Xen Dom0 kernel in Fedora 9/10, we are *not* trying to shaft anyone & will re-introduce Xen Dom0 kernels to Fedora when they become available. Regards, Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
Neil Thompson wrote:> Xen (paravirt) is in the process of going into the Mainstream kernel, also > RHEL5, which has a number of years left, includes xen - I don''t think > Red Hat are going to mess their corporate clients around by removing it.I hope you''re right, but that $107 million is saying something else. -Evan
On Wed, Jan 21, 2009 at 12:04:34PM +0200, Neil Thompson wrote:> On Wed, Jan 21, 2009 at 10:51:07AM +0100, Jan ONDREJ (SAL) wrote: > > On Wed, Jan 21, 2009 at 11:44:35AM +0200, Neil Thompson wrote: > > > On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote: > > > RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and Fedora 11 > > > should have, as well. > > > > Just there is at least one critical problem with Fedora 10 DomU and also one > > for Fedora 9 DomU, so Fedora 10 kernels can''t run stable. :( > > > > https://bugzilla.redhat.com/show_bug.cgi?id=478414 > > > > Your problem is with an F8 Dom0...Not exactly. On my F8 Dom0 Fedora 10 works well with latest fc8 domU kernel. Just when trying to boot and run .fc10. PAE or x86_64 kernel, it fails in less than 48 hours.> the 2.6.29 kernel that will (hopefully) be in > F11 should sort out most of the paravirt problems, from what I can see. I''m > not sure what the status of the code in EL5.3 is, maybe the RH people > can comment.I think it will not be in 2.6.29. Although http://wiki.xensource.com/xenwiki/XenParavirtOps says thatit''s planned for 2.6.29, I can''t see any progres in 2.6.29-rc2 changelog: http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.29-rc2 http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.29-rc1 I think, 2.6.29 is closed for large changes. Or I missed something? SAL
Jan ONDREJ (SAL) wrote:> On Wed, Jan 21, 2009 at 12:04:34PM +0200, Neil Thompson wrote: > >> On Wed, Jan 21, 2009 at 10:51:07AM +0100, Jan ONDREJ (SAL) wrote: >> >>> On Wed, Jan 21, 2009 at 11:44:35AM +0200, Neil Thompson wrote: >>> >>>> On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote: >>>> RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and Fedora 11 >>>> should have, as well. >>>> >>> Just there is at least one critical problem with Fedora 10 DomU and also one >>> for Fedora 9 DomU, so Fedora 10 kernels can''t run stable. :( >>> >>> https://bugzilla.redhat.com/show_bug.cgi?id=478414 >>> >>> >> Your problem is with an F8 Dom0... >> > > Not exactly. On my F8 Dom0 Fedora 10 works well with latest fc8 domU kernel. > Just when trying to boot and run .fc10. PAE or x86_64 kernel, it fails > in less than 48 hours. > >I would really recommend you to change the Dom0 to be CentOS 5.2 or newer, Fedora 8 has been "Dead" about ½-month allready and will not get any updates be it KVM, QEMU or Xen. So the bugs will not be fixed and there will be no security updates for F8. I have some virtual environments running still vith Xen and F7 or F8, but will migrate them soon to CentOS for stability and long term support and I quite much expect them to be still supported 5-years from now. Also I don''t think kvm will be that different or hard to learn if it becomes to that. It actually has paravirtual network drivers for windows from Qumranet which you can get without extra fee so I think if you are having windows clients it could be way to go in future. For xen you need to pay to Novell for that priviledge. So all in all I think for me this aquisition is good news. I think most problems with xen comes from xensource as it''s they only product generating income and for that reason the opensource version seems to get less care than the version you can buy from them. (This is just my opinion so it''s not necessarily so) Greetings, Veli-Pekka
On Wed, 21 Jan 2009, Jan ONDREJ (SAL) wrote:> On Wed, Jan 21, 2009 at 11:44:35AM +0200, Neil Thompson wrote: >> On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote: >> RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and Fedora 11 >> should have, as well. > > Just there is at least one critical problem with Fedora 10 DomU and also one > for Fedora 9 DomU, so Fedora 10 kernels can''t run stable. :( > > https://bugzilla.redhat.com/show_bug.cgi?id=478414That might actually be nothing to do with xen. Dovecot provides a very successful way of hammering the file locking code, and you might just be exposing bugs or race conditions in the underlying kernel rather than the xen specific code. I suggest you check to see if the processes involved are waiting to access locked files. Michael Young
On Wed, Jan 21, 2009 at 09:20:48AM +0100, Jan ONDREJ (SAL) wrote:> KVM is still not a replacement for paravirtualized machines and I think > fully virtualized KVM will be slower like a paravirtualized XEN.KVM is a great replacement for Xen. It''s much easier to use for a start -- no more rebooting into a completely separate kernel^W hypervisor. As long as you have the virtio drivers in the guest, which is the default for all new Linux distros, performance is roughly the same.> Also I am missing some howtos for migration to KVM/xenner.Install a recent Linux kernel in the guest, adjust the configuration file[1], and reboot. You only need Xenner if you want to run the Xen PV guest unchanged (ie. without installing a new guest kernel). Rich. [1] ''virsh edit domname'', and edit the domain type, <os> and <emulator> fields, as detailed here: http://libvirt.org/drvqemu.html -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into Xen guests. http://et.redhat.com/~rjones/virt-p2v
On Wed, Jan 21, 2009 at 01:40:55PM +0200, Veli-Pekka Kestilä wrote:> Jan ONDREJ (SAL) wrote: > >On Wed, Jan 21, 2009 at 12:04:34PM +0200, Neil Thompson wrote: > > > >>On Wed, Jan 21, 2009 at 10:51:07AM +0100, Jan ONDREJ (SAL) wrote: > >> > >>>On Wed, Jan 21, 2009 at 11:44:35AM +0200, Neil Thompson wrote: > >>> > >>>>On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote: > >>>>RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and > >>>>Fedora 11 > >>>>should have, as well. > >>>> > >>>Just there is at least one critical problem with Fedora 10 DomU and also > >>>one > >>>for Fedora 9 DomU, so Fedora 10 kernels can''t run stable. :( > >>> > >>>https://bugzilla.redhat.com/show_bug.cgi?id=478414 > >>> > >>> > >>Your problem is with an F8 Dom0... > >> > > > >Not exactly. On my F8 Dom0 Fedora 10 works well with latest fc8 domU > >kernel. > >Just when trying to boot and run .fc10. PAE or x86_64 kernel, it fails > >in less than 48 hours. > > > > > I would really recommend you to change the Dom0 to be CentOS 5.2 or > newer, Fedora 8 has been "Dead" about ½-month allready and will not get > any updates be it KVM, QEMU or Xen. So the bugs will not be fixed and > there will be no security updates for F8. I have some virtual > environments running still vith Xen and F7 or F8, but will migrate them > soon to CentOS for stability and long term support and I quite much > expect them to be still supported 5-years from now. > > Also I don''t think kvm will be that different or hard to learn if it > becomes to that. It actually has paravirtual network drivers for windows > from Qumranet which you can get without extra fee so I think if you are > having windows clients it could be way to go in future. For xen you need > to pay to Novell for that priviledge. >You can always use "GPLPV" Windows drivers for Xen. They''re open source.> So all in all I think for me this aquisition is good news. I think most > problems with xen comes from xensource as it''s they only product > generating income and for that reason the opensource version seems to > get less care than the version you can buy from them. (This is just my > opinion so it''s not necessarily so) >I think Xensource is putting a lot of effort into opensource Xen. It''s just the dom0/pv_ops mess that''s causing problems atm.. that _should_ get fixed in the near future. -- Pasi
Veli-Pekka Kestilä wrote:> Also I don''t think kvm will be that different or hard to learn if it > becomes to that. It actually has paravirtual network drivers for windows > from Qumranet which you can get without extra fee so I think if you are > having windows clients it could be way to go in future. For xen you need > to pay to Novell for that priviledge.Has anyone tried the Qumranet drivers? My XP clients on Xen are very slow. There are free Windows drivers at http://wiki.xensource.com/xenwiki/XenWindowsGplPv, but I think they still need some development. -Evan
On Wed, Jan 21, 2009 at 1:00 PM, Evan Lavelle <sa212+fcxen@cyconix.com<sa212%2Bfcxen@cyconix.com>> wrote:> Veli-Pekka Kestilä wrote: > > Also I don''t think kvm will be that different or hard to learn if it >> becomes to that. It actually has paravirtual network drivers for windows >> from Qumranet which you can get without extra fee so I think if you are >> having windows clients it could be way to go in future. For xen you need to >> pay to Novell for that priviledge. >> > > Has anyone tried the Qumranet drivers? My XP clients on Xen are very slow. > There are free Windows drivers at > http://wiki.xensource.com/xenwiki/XenWindowsGplPv, but I think they still > need some development. >I tried, they were running OK for me. -- Emre
On Wed, Jan 21, 2009 at 01:40:55PM +0200, Veli-Pekka Kestilä wrote:> I would really recommend you to change the Dom0 to be CentOS 5.2 or > newer, Fedora 8 has been "Dead" about ½-month allready and will not getIt''s not easy to convert my Dom0 to CentOS, but may be it will be only one whing which I can do. But I am still not sure, if it helps. My fc8 domU works well. May be there are some API changes between fc8 xen dom0 and fc10 domU, which causes my problems.> So all in all I think for me this aquisition is good news. I think most > problems with xen comes from xensource as it''s they only product > generating income and for that reason the opensource version seems to > get less care than the version you can buy from them. (This is just my > opinion so it''s not necessarily so)Destination of KVM looks good, but it''s still not ready to replace Xen. May be after 1-2 years of testing it will be better like xen. SAL
On Wed, Jan 21, 2009 at 11:50:39AM +0000, Richard W.M. Jones wrote:> On Wed, Jan 21, 2009 at 09:20:48AM +0100, Jan ONDREJ (SAL) wrote: > > KVM is still not a replacement for paravirtualized machines and I think > > fully virtualized KVM will be slower like a paravirtualized XEN. > > KVM is a great replacement for Xen. It''s much easier to use for a > start -- no more rebooting into a completely separate kernel^W > hypervisor. As long as you have the virtio drivers in the guest, > which is the default for all new Linux distros, performance is roughly > the same. > > > Also I am missing some howtos for migration to KVM/xenner. > > Install a recent Linux kernel in the guest, adjust the configuration > file[1], and reboot. You only need Xenner if you want to run the Xen > PV guest unchanged (ie. without installing a new guest kernel).For F10 there is no need to change domU kernel. It''s same. But after reboot to KVM, my virtual machine has an 8139 network card. Is it paravirtualized? How I can tell my machine to use "virtio" drivers? SAL
On Wed, Jan 21, 2009 at 11:47:55AM +0000, M A Young wrote:> On Wed, 21 Jan 2009, Jan ONDREJ (SAL) wrote: > >> On Wed, Jan 21, 2009 at 11:44:35AM +0200, Neil Thompson wrote: >>> On Wed, Jan 21, 2009 at 08:15:00AM +0000, Evan Lavelle wrote: >>> RHEL (and CentOS) 5.3 has support for Fedora 10 (and above) domUs, and Fedora 11 >>> should have, as well. >> >> Just there is at least one critical problem with Fedora 10 DomU and also one >> for Fedora 9 DomU, so Fedora 10 kernels can''t run stable. :( >> >> https://bugzilla.redhat.com/show_bug.cgi?id=478414 > > That might actually be nothing to do with xen. Dovecot provides a very > successful way of hammering the file locking code, and you might just be > exposing bugs or race conditions in the underlying kernel rather than the > xen specific code. I suggest you check to see if the processes involved > are waiting to access locked files.I know, why dovecot fails. And it''s not only dovecot, also mysql, apache, ... This bug describes only one example, when I was able to make some logs. My problem is, that command "sync" and may be similar kernel function called from dovecot, mysql, ... stays in "D" state. File cache is not stored on my media. I can read all changed data back from cache, but they are not stored on disks. After reboot all data written to cache is lost. May be my english is not perfect, I am sorry. :( SAL
On Wed, Jan 21, 2009 at 01:12:52PM +0100, Jan ONDREJ (SAL) wrote:> On Wed, Jan 21, 2009 at 11:50:39AM +0000, Richard W.M. Jones wrote: > > On Wed, Jan 21, 2009 at 09:20:48AM +0100, Jan ONDREJ (SAL) wrote: > > > KVM is still not a replacement for paravirtualized machines and I think > > > fully virtualized KVM will be slower like a paravirtualized XEN. > > > > KVM is a great replacement for Xen. It''s much easier to use for a > > start -- no more rebooting into a completely separate kernel^W > > hypervisor. As long as you have the virtio drivers in the guest, > > which is the default for all new Linux distros, performance is roughly > > the same. > > > > > Also I am missing some howtos for migration to KVM/xenner. > > > > Install a recent Linux kernel in the guest, adjust the configuration > > file[1], and reboot. You only need Xenner if you want to run the Xen > > PV guest unchanged (ie. without installing a new guest kernel). > > For F10 there is no need to change domU kernel. It''s same. > > But after reboot to KVM, my virtual machine has an 8139 network card. > Is it paravirtualized? How I can tell my machine to use "virtio" drivers?You have to tell the host to give the guest a virtio network card - change the NIC <model type=''virtio''/> as described here: http://libvirt.org/formatdomain.html#elementsNICS The guest needs to have a relatively up to date kernel which has drivers for the virtio network card - that''s included in all recent Linux kernels (virtio_net.ko). Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-top is ''top'' for virtual machines. Tiny program with many powerful monitoring features, net stats, disk stats, logging, etc. http://et.redhat.com/~rjones/virt-top
Pasi Kärkkäinen wrote:> On Wed, Jan 21, 2009 at 01:40:55PM +0200, Veli-Pekka Kestilä wrote: > >> >> Also I don''t think kvm will be that different or hard to learn if it >> becomes to that. It actually has paravirtual network drivers for windows >> from Qumranet which you can get without extra fee so I think if you are >> having windows clients it could be way to go in future. For xen you need >> to pay to Novell for that priviledge. >> >> > > You can always use "GPLPV" Windows drivers for Xen. They''re open source. > >Thanks for the tip, I have quite many Windows hosts running with Xen and have been looking for drivers, but haven''t found these. Have to test them out when I get my new testing machine.>> So all in all I think for me this aquisition is good news. I think most >> problems with xen comes from xensource as it''s they only product >> generating income and for that reason the opensource version seems to >> get less care than the version you can buy from them. (This is just my >> opinion so it''s not necessarily so) >> > > I think Xensource is putting a lot of effort into opensource Xen. > > It''s just the dom0/pv_ops mess that''s causing problems atm.. that _should_ > get fixed in the near future. > >I have noticed that. But main reason for my feeling is that they should have started the push much more early than they did. Or maybe it wasn''t feasible before. Anyway it''s just have been my feeling of the whole thing and I think it''s good to have two competing techs as it will provide that both will advance. Veli-Pekka
On Wed, Jan 21, 2009 at 01:53:19PM +0200, Pasi Kärkkäinen wrote:> I think Xensource is putting a lot of effort into opensource Xen.I''ve heard totally different stories from others. For almost every problem you report, they''ll tell you that "it is fixed in the commercial version". Unfortunately, many companies just "use" open source as marketing instrument, sometimes they even *abuse* the term (i.e. call software with no OSI-compliant license open source -- which is not the case with Xen b.t.w.). -- -- Jos Vos <jos@xos.nl> -- X/OS Experts in Open Systems BV | Phone: +31 20 6938364 -- Amsterdam, The Netherlands | Fax: +31 20 6948204
Evan Lavelle wrote:> Neil Thompson wrote: > >> Xen (paravirt) is in the process of going into the Mainstream kernel, >> also >> RHEL5, which has a number of years left, includes xen - I don''t think >> Red Hat are going to mess their corporate clients around by removing it. > > I hope you''re right, but that $107 million is saying something else. >And http://www.redhat.com/promo/qumranet/ again says Neil''s right. Kind regards, Jeroen van Meeuwen -kanarip
On Wed, Jan 21, 2009 at 12:19:31PM +0000, Richard W.M. Jones wrote:> On Wed, Jan 21, 2009 at 01:12:52PM +0100, Jan ONDREJ (SAL) wrote: > > You have to tell the host to give the guest a virtio network card - > change the NIC <model type=''virtio''/> as described here:Many thanks. Works well.> http://libvirt.org/formatdomain.html#elementsNICSThere is nothing interesant on this URL. Part elementsNICS is not found, but also thanks. On Wed, Jan 21, 2009 at 01:23:06PM +0100, Emre Erenoglu wrote:> you need to read the KVM virtio wiki pages, you need to disable that 8139 > emulated card driver (blacklist), then change some line in guest > configuration file, and probe virtio driver at boot. > > http://wiki.libvirt.org/page/Virtio > > http://kvm.qumranet.com/kvmwiki?action=fullsearch&context=180&value=virtio&titlesearch=Ba%C5%9Fl%C4%B1klar > > find /lib/modules | grep virtio will give you the list of modules (if they > are compiled as modules) > > You also need to have a recent kernel such as 2.6.25 which has built-in > virtio support.Thanks. I have it. virtio_net works well, but I have trouble to boot from virtio_blk. I can add second disk as virto block device, but I can''t boot from first disk. When using sedond disk, everything works well. When booting via grub, this is on console: input: ImExPS/2 Generic Explorer Mouse as /devices/platform/i8042/serio1/input/input3 Setting up hotplug. Creating block device nodes. Creating character device nodes. Loading pata_acpi module Loading ata_generic module Loading virtio_blk module Creating root device. Mounting root filesystem. mount: error mounting /dev/root on /sysroot as ext3: No such file or directory Setting up hotplug. Creating block device nodes. Creating character device nodes. Loading pata_acpi module Loading ata_generic module Loading virtio_blk module Creating root device. Mounting root filesystem. mount: error mounting /dev/root on /sysroot as ext3: No such file or directory I am using UUID in GRUB, my initrd has been regenerated with virtio_blk module. I think, nash can''t create /dev/vda* devices or something similar. Any ideas, how to debug this or how to fix this? SAL
On Wed, Jan 21, 2009 at 02:28:36PM +0100, Jan ONDREJ (SAL) wrote:> virtio_net works well, but I have trouble to boot from virtio_blk. > > I can add second disk as virto block device, but I can''t boot from first > disk. When using sedond disk, everything works well. When booting via grub, > this is on console:Upgrading to using virtio_blk is very complicated. You have to rebuild initrd, and there''s a difficult circular dependency to be resolved when doing this because you need to be using virtio_blk in order for mkinitrd to believe that you need it, although possibly mkinitrd supports some command line argument to override this. I actually gave up at this point. For newly installed guests, recent anaconda just works everything out for you and puts the correct drivers into initrd. Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-df lists disk usage of guests without needing to install any software inside the virtual machine. Supports Linux and Windows. http://et.redhat.com/~rjones/virt-df/
Mark McLoughlin
2009-Jan-21 13:48 UTC
Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
(This is all getting offtopic for fedora-xen, we should really move to fedora-virt) On Wed, 2009-01-21 at 14:28 +0100, Jan ONDREJ (SAL) wrote:> virtio_net works well, but I have trouble to boot from virtio_blk. > > I can add second disk as virto block device, but I can''t boot from > first disk.When switching from IDE to virtio, you need to first build a new initrd in the guest with e.g.: $> mkinitrd --with virtio_pci --with virtio_blk -f /boot/initrd-$(kernelversion) $(kernelversion) You only need to do this once. After that, if a new kernel is installed while you''re booted off a virtio disk, then mkinitrd will include the modules automatically. Cheers, Mark.
On Wed, Jan 21, 2009 at 01:45:52PM +0000, Richard W.M. Jones wrote:> On Wed, Jan 21, 2009 at 02:28:36PM +0100, Jan ONDREJ (SAL) wrote: > > virtio_net works well, but I have trouble to boot from virtio_blk. > > > > I can add second disk as virto block device, but I can''t boot from first > > disk. When using sedond disk, everything works well. When booting via grub, > > this is on console: > > Upgrading to using virtio_blk is very complicated. You have to > rebuild initrd, and there''s a difficult circular dependency to be > resolved when doing this because you need to be using virtio_blk in > order for mkinitrd to believe that you need it, although possibly > mkinitrd supports some command line argument to override this. I > actually gave up at this point.Just tell it to include virtio_blk module explicitly eg mkinitrd --with=virtio_blk .....other option... Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
Emre Erenoglu
2009-Jan-21 14:06 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Wed, Jan 21, 2009 at 2:48 PM, Mark McLoughlin <markmc@redhat.com> wrote:> (This is all getting offtopic for fedora-xen, we should really move to > fedora-virt) > > On Wed, 2009-01-21 at 14:28 +0100, Jan ONDREJ (SAL) wrote: > > virtio_net works well, but I have trouble to boot from virtio_blk. > > > > I can add second disk as virto block device, but I can''t boot from > > first disk. > > When switching from IDE to virtio, you need to first build a new initrd > in the guest with e.g.: > > $> mkinitrd --with virtio_pci --with virtio_blk -f > /boot/initrd-$(kernelversion) $(kernelversion) > > You only need to do this once. After that, if a new kernel is installed > while you''re booted off a virtio disk, then mkinitrd will include the > modules automatically.You will also need to specify /dev/vdX on the kernel root= line and make sure your init script inside your initrd triggers the virtio drivers at boot so that the /dev/vdX are created. -- Emre
> I think it will not be in 2.6.29. Although > http://wiki.xensource.com/xenwiki/XenParavirtOps says thatit''s planned for > 2.6.29, I can''t see any progres in 2.6.29-rc2 changelog: > http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.29-rc2 > http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.29-rc1 > > I think, 2.6.29 is closed for large changes. Or I missed something?Yea, looks like most of the dom0 stuff missed the boat. Some prelimary stuff might be in though (swiotlb patches ran over lkml at least, didn''t check whenever they are actually made it). You can combine a self-compiled pv_ops/dom0 kernel with the Rawhide xen packages and help testing/debugging/get-stuff-upstream this way. You should follow xen-devel for the latest news & patches if you do so. The pv_ops/dom0 kernel has some not-yet debugged storage issues (disk controller either fails after a while or doesn''t work at all), which is the major stumbling block right now. If you are lucky and the box stays up long enougth you can start guests. cheers, Gerd
Hi,> Just tell it to include virtio_blk module explicitly eg > > mkinitrd --with=virtio_blk .....other option...virtio_pci too (which is probably the issue in this case, as the reporter sayed virtio_blk is included already). cheers, Gerd
Richard W.M. Jones
2009-Jan-21 14:45 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Wed, Jan 21, 2009 at 03:06:43PM +0100, Emre Erenoglu wrote:> On Wed, Jan 21, 2009 at 2:48 PM, Mark McLoughlin <markmc@redhat.com> wrote: > > > (This is all getting offtopic for fedora-xen, we should really move to > > fedora-virt) > > > > On Wed, 2009-01-21 at 14:28 +0100, Jan ONDREJ (SAL) wrote: > > > virtio_net works well, but I have trouble to boot from virtio_blk. > > > > > > I can add second disk as virto block device, but I can''t boot from > > > first disk. > > > > When switching from IDE to virtio, you need to first build a new initrd > > in the guest with e.g.: > > > > $> mkinitrd --with virtio_pci --with virtio_blk -f > > /boot/initrd-$(kernelversion) $(kernelversion) > > > > You only need to do this once. After that, if a new kernel is installed > > while you''re booted off a virtio disk, then mkinitrd will include the > > modules automatically. > > > You will also need to specify /dev/vdX on the kernel root= line and make > sure your init script inside your initrd triggers the virtio drivers at boot > so that the /dev/vdX are created.Yes I have to agree with Emre here - I don''t think it''s as simple as just rebuilding mkinitrd. I got that far but gave up later on. /me checks notes ... Yup, I got as far as working out that you would have to edit fstab and possibly /boot/grub/device.map and /boot/grub/menu.lst, before giving up. If anyone would like to fill in the wiki page here on the subject: http://wiki.libvirt.org/page/Virtio I''d like to reiterate that _none_ of this complexity is required when installing a new guest. Anaconda sets up everything for you. Rich. -- Richard Jones, Emerging Technologies, Red Hat http://et.redhat.com/~rjones virt-df lists disk usage of guests without needing to install any software inside the virtual machine. Supports Linux and Windows. http://et.redhat.com/~rjones/virt-df/
Emre Erenoglu
2009-Jan-21 14:51 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Wed, Jan 21, 2009 at 3:45 PM, Richard W.M. Jones <rjones@redhat.com>wrote:> On Wed, Jan 21, 2009 at 03:06:43PM +0100, Emre Erenoglu wrote: > > On Wed, Jan 21, 2009 at 2:48 PM, Mark McLoughlin <markmc@redhat.com> > wrote: > > > > > (This is all getting offtopic for fedora-xen, we should really move to > > > fedora-virt) > > > > > > On Wed, 2009-01-21 at 14:28 +0100, Jan ONDREJ (SAL) wrote: > > > > virtio_net works well, but I have trouble to boot from virtio_blk. > > > > > > > > I can add second disk as virto block device, but I can''t boot from > > > > first disk. > > > > > > When switching from IDE to virtio, you need to first build a new initrd > > > in the guest with e.g.: > > > > > > $> mkinitrd --with virtio_pci --with virtio_blk -f > > > /boot/initrd-$(kernelversion) $(kernelversion) > > > > > > You only need to do this once. After that, if a new kernel is installed > > > while you''re booted off a virtio disk, then mkinitrd will include the > > > modules automatically. > > > > > > You will also need to specify /dev/vdX on the kernel root= line and make > > sure your init script inside your initrd triggers the virtio drivers at > boot > > so that the /dev/vdX are created. > > Yes I have to agree with Emre here - I don''t think it''s as simple as > just rebuilding mkinitrd. I got that far but gave up later on.Well what I did was that I created a special initramfs, with virtio-net and virtio-blk drivers, as well as all of their dependencies (virtio-pci, virti-baloon and possibly one more), edited the init script going inside the initrd image so that it triggers the virtio drivers, then booted up on that ROOT filesystem using the kernel command line. on system side, it''s better to edit fstab but root is anyway mounted. You may need to boot to shell at your initrd to see what''s going on, or probe by hand to see the devices are created etc. You also need to use the correct guest configuration file so that virtio-type devices are created instead of emulated ones. -- Emre
On Wed, Jan 21, 2009 at 02:20:19PM +0200, Veli-Pekka Kestilä wrote:> >>So all in all I think for me this aquisition is good news. I think most > >>problems with xen comes from xensource as it''s they only product > >>generating income and for that reason the opensource version seems to > >>get less care than the version you can buy from them. (This is just my > >>opinion so it''s not necessarily so) > >> > > > >I think Xensource is putting a lot of effort into opensource Xen. > > > >It''s just the dom0/pv_ops mess that''s causing problems atm.. that _should_ > >get fixed in the near future. > > > > > I have noticed that. But main reason for my feeling is that they should > have started the push much more early than they did. Or maybe it wasn''t > feasible before. Anyway it''s just have been my feeling of the whole > thing and I think it''s good to have two competing techs as it will > provide that both will advance. >Yeah well.. IIRC Xen patches were first sent for 2.6.15 kernel. Those were rejected and not integrated into vanilla kernel. I can''t remember the reason for that. Next attempt failed when others wanted to have the paravirt ops (pv_ops) framework instead, which enables Linux kernel PV support for any hypervisor, not just for Xen. Then it took a while to get the generic pv_ops framework done and merged into Linux kernel. And after that it has taken a lot of time to port the Xen domU from the original "Xenlinux" to pv_ops framework. pv_ops Xen domU support has been in mainline Linux kernel since 2.6.23. And now there is active development going on to get the pv_ops dom0 working and merged into Linux. Btw Redhat started this work earlier, and it is now being continued by Xensource (Jeremy Fitzhardinge). http://wiki.xensource.com/xenwiki/XenParavirtOps -- Pasi
Jos Vos writes ("Re: [Fedora-xen] Goodbye Xen on
RH/Fedora?"):> On Wed, Jan 21, 2009 at 01:53:19PM +0200, Pasi Kärkkäinen wrote:
> > I think Xensource is putting a lot of effort into opensource Xen.
>
> I''ve heard totally different stories from others. For almost
every
> problem you report, they''ll tell you that "it is fixed in the
> commercial version".
I don''t know where you''re asking. As far as I know Citrix
(who bought
Xensource) don''t offer paid end-user support as a service for the
upstream version of Xen. On the other hand they do pay the salaries
of various people who work on it (including me and of course Keir).
If you find bugs in upstream Xen we''re certainly interested; we spend
most of our time trying to keep the quality up and manage the incoming
patchstream. It''s true that lists like xen-users aren''t that
well-read by developers like me but I think you''ll find that''s
often
true with a Free Software project; many developers prefer to try to
help improve the code than to help individual users one at a time.
As you can see at least some of us do read this list. I keep an eye
on things here as well as on other distro lists - despite my Debian
background :-) - to see if there''s anything we can do to improve
things. Most of the issues reported here are about Fedora-specific
packaging of course, which I''m no expert on.
It''s true that the situation with the kernel is very disappointing.
I don''t think I can really explain what I see as the causes.
It''s too
much of a political hot potato and I don''t want to offend anyone,
particularly my hosts here.
Fortunately we now have Jeremy Fitzhardinge in charge of getting Xen
support into Linux upstream and that seems to be making reasonable
progress - although of course we would all prefer it to be faster.
Regards,
Ian.
NB this is not the official position of Citrix about anything,
just my personal opinion.
I tried the Qumranet drivers before I went with Xen. I don''t think there is necessarily a problem with the Qumranet drivers, in fact, they could potentially have better inbound speeds than the GPLPV ones (though it seems unlikely as much as people test and James works on them on the xen-users list). The reason the Qumranet drivers don''t cut it is because they are only network drivers. This means your data access (and possibly other stuff GPLPV hits) is still fully virtualized. Another reason I went with Xen is the PHY: option. I use a physical data source, as opposed to a file, for my guests. Each one has its own HD, actually, though partitions or RAID arrays would obviously work as well. If I remember correctly, when I tried this (some time ago), KVM had no such option. Dustin -----Original Message----- From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] On Behalf Of Evan Lavelle Sent: Wednesday, January 21, 2009 07:01 To: fedora-xen@redhat.com Subject: Re: [Fedora-xen] Goodbye Xen on RH/Fedora? Veli-Pekka Kestilä wrote:> Also I don''t think kvm will be that different or hard to learn if it > becomes to that. It actually has paravirtual network drivers for windows > from Qumranet which you can get without extra fee so I think if you are > having windows clients it could be way to go in future. For xen you need > to pay to Novell for that priviledge.Has anyone tried the Qumranet drivers? My XP clients on Xen are very slow. There are free Windows drivers at http://wiki.xensource.com/xenwiki/XenWindowsGplPv, but I think they still need some development. -Evan -- Fedora-xen mailing list Fedora-xen@redhat.com https://www.redhat.com/mailman/listinfo/fedora-xen
Also, while I''m on the topic, PCI-passthrough was a factor in my decision as well. Unfortunately, I''m stuck on F8 as well, because I can''t get networking to work on CentOS Xen with my hardware. However, I understand what is going on, so I''m just holding my breath. Dustin -----Original Message----- From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] On Behalf Of Dustin Henning Sent: Wednesday, January 21, 2009 13:35 To: ''Evan Lavelle''; fedora-xen@redhat.com Subject: RE: [Fedora-xen] Goodbye Xen on RH/Fedora? I tried the Qumranet drivers before I went with Xen. I don''t think there is necessarily a problem with the Qumranet drivers, in fact, they could potentially have better inbound speeds than the GPLPV ones (though it seems unlikely as much as people test and James works on them on the xen-users list). The reason the Qumranet drivers don''t cut it is because they are only network drivers. This means your data access (and possibly other stuff GPLPV hits) is still fully virtualized. Another reason I went with Xen is the PHY: option. I use a physical data source, as opposed to a file, for my guests. Each one has its own HD, actually, though partitions or RAID arrays would obviously work as well. If I remember correctly, when I tried this (some time ago), KVM had no such option. Dustin -----Original Message----- From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] On Behalf Of Evan Lavelle Sent: Wednesday, January 21, 2009 07:01 To: fedora-xen@redhat.com Subject: Re: [Fedora-xen] Goodbye Xen on RH/Fedora? Veli-Pekka Kestilä wrote:> Also I don''t think kvm will be that different or hard to learn if it > becomes to that. It actually has paravirtual network drivers for windows > from Qumranet which you can get without extra fee so I think if you are > having windows clients it could be way to go in future. For xen you need > to pay to Novell for that priviledge.Has anyone tried the Qumranet drivers? My XP clients on Xen are very slow. There are free Windows drivers at http://wiki.xensource.com/xenwiki/XenWindowsGplPv, but I think they still need some development. -Evan -- Fedora-xen mailing list Fedora-xen@redhat.com https://www.redhat.com/mailman/listinfo/fedora-xen -- Fedora-xen mailing list Fedora-xen@redhat.com https://www.redhat.com/mailman/listinfo/fedora-xen
Dustin Henning wrote:> I tried the Qumranet drivers before I went with Xen. I don''t think > there is necessarily a problem with the Qumranet drivers, in fact, they > could potentially have better inbound speeds than the GPLPV ones (though it > seems unlikely as much as people test and James works on them on the > xen-users list). The reason the Qumranet drivers don''t cut it is because > they are only network drivers. This means your data access (and possibly > other stuff GPLPV hits) is still fully virtualized.Storage drivers are in the works, hopefully out soon.> Another reason I went > with Xen is the PHY: option. I use a physical data source, as opposed to a > file, for my guests. Each one has its own HD, actually, though partitions > or RAID arrays would obviously work as well. If I remember correctly, when > I tried this (some time ago), KVM had no such optionkvm has had this from day 1; ''qemu /dev/volgroup/logvol'' will start a guest from the specified logical volume. For good performance I recommend ''qemu -drive file=/dev/volgroup/logvol,cache=off''. Of course, libvirt will handle all that for you. -- I have a truly marvellous patch that fixes the bug which this signature is too narrow to contain.
Dustin Henning
2009-Jan-21 18:47 UTC
RE: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
It is probably worth noting that if someone were to run a kernel update with a package manager after getting this working (using information from later posts in this thread), they would need to do a manual mkinitrd. At least that was my experience some time ago when switching from hda to sda in CentOS, future kernel-xen versions installed via yum tried to boot to hda and failed. However, I think I actually recompiled originally, though, as opposed to adding the new drivers to initrd, so I could be wrong. Dustin -----Original Message----- From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] On Behalf Of Mark McLoughlin Sent: Wednesday, January 21, 2009 08:49 To: Jan ONDREJ (SAL) Cc: Fedora Xen Subject: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?] (This is all getting offtopic for fedora-xen, we should really move to fedora-virt) On Wed, 2009-01-21 at 14:28 +0100, Jan ONDREJ (SAL) wrote:> virtio_net works well, but I have trouble to boot from virtio_blk. > > I can add second disk as virto block device, but I can''t boot from > first disk.When switching from IDE to virtio, you need to first build a new initrd in the guest with e.g.: $> mkinitrd --with virtio_pci --with virtio_blk -f /boot/initrd-$(kernelversion) $(kernelversion) You only need to do this once. After that, if a new kernel is installed while you''re booted off a virtio disk, then mkinitrd will include the modules automatically. Cheers, Mark. -- Fedora-xen mailing list Fedora-xen@redhat.com https://www.redhat.com/mailman/listinfo/fedora-xen
''Jan ONDREJ (SAL)''
2009-Jan-21 19:05 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
You missed, that I am migrating not only my packages, but my whole virtual machine from paravirtualized Xen to paravirt. driver KVM. This mkinitrd magic is only required for disk driver change (xenblk_front to virtio_blk). So it''s something different, like your upgrade of CentOS. SAL On Wed, Jan 21, 2009 at 01:47:34PM -0500, Dustin Henning wrote:> It is probably worth noting that if someone were to run a kernel > update with a package manager after getting this working (using information > from later posts in this thread), they would need to do a manual mkinitrd. > At least that was my experience some time ago when switching from hda to sda > in CentOS, future kernel-xen versions installed via yum tried to boot to hda > and failed. However, I think I actually recompiled originally, though, as > opposed to adding the new drivers to initrd, so I could be wrong. > Dustin > > -----Original Message----- > From: fedora-xen-bounces@redhat.com [mailto:fedora-xen-bounces@redhat.com] > On Behalf Of Mark McLoughlin > Sent: Wednesday, January 21, 2009 08:49 > To: Jan ONDREJ (SAL) > Cc: Fedora Xen > Subject: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen > on RH/Fedora?] > > (This is all getting offtopic for fedora-xen, we should really move to > fedora-virt) > > On Wed, 2009-01-21 at 14:28 +0100, Jan ONDREJ (SAL) wrote: > > virtio_net works well, but I have trouble to boot from virtio_blk. > > > > I can add second disk as virto block device, but I can''t boot from > > first disk. > > When switching from IDE to virtio, you need to first build a new initrd > in the guest with e.g.: > > $> mkinitrd --with virtio_pci --with virtio_blk -f > /boot/initrd-$(kernelversion) $(kernelversion) > > You only need to do this once. After that, if a new kernel is installed > while you''re booted off a virtio disk, then mkinitrd will include the > modules automatically. > > Cheers, > Mark. > > -- > Fedora-xen mailing list > Fedora-xen@redhat.com > https://www.redhat.com/mailman/listinfo/fedora-xen >
Wow, maybe the F8 documentation doesn''t cover that, or maybe I''m remembering wrong, but good to know for future testing. Thanks. -----Original Message----- From: Avi Kivity [mailto:avi@redhat.com] Sent: Wednesday, January 21, 2009 13:43 To: Dustin.Henning@prd-inc.com Cc: ''Evan Lavelle''; fedora-xen@redhat.com Subject: Re: [Fedora-xen] Goodbye Xen on RH/Fedora? Dustin Henning wrote:> I tried the Qumranet drivers before I went with Xen. I don''t think > there is necessarily a problem with the Qumranet drivers, in fact, they > could potentially have better inbound speeds than the GPLPV ones (thoughit> seems unlikely as much as people test and James works on them on the > xen-users list). The reason the Qumranet drivers don''t cut it is because > they are only network drivers. This means your data access (and possibly > other stuff GPLPV hits) is still fully virtualized.Storage drivers are in the works, hopefully out soon.> Another reason I went > with Xen is the PHY: option. I use a physical data source, as opposed toa> file, for my guests. Each one has its own HD, actually, though partitions > or RAID arrays would obviously work as well. If I remember correctly,when> I tried this (some time ago), KVM had no such optionkvm has had this from day 1; ''qemu /dev/volgroup/logvol'' will start a guest from the specified logical volume. For good performance I recommend ''qemu -drive file=/dev/volgroup/logvol,cache=off''. Of course, libvirt will handle all that for you. -- I have a truly marvellous patch that fixes the bug which this signature is too narrow to contain.
Mark McLoughlin
2009-Jan-22 09:20 UTC
RE: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Wed, 2009-01-21 at 13:47 -0500, Dustin Henning wrote:> It is probably worth noting that if someone were to run a kernel > update with a package manager after getting this working (using > information > from later posts in this thread), they would need to do a manual > mkinitrd.No. If you boot off /dev/vda and update the kernel, mkinitrd will build an initrd containing the virtio modules. Cheers, Mark.
Emre Erenoglu
2009-Jan-22 10:06 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Thu, Jan 22, 2009 at 10:20 AM, Mark McLoughlin <markmc@redhat.com> wrote:> On Wed, 2009-01-21 at 13:47 -0500, Dustin Henning wrote: > > It is probably worth noting that if someone were to run a kernel > > update with a package manager after getting this working (using > > information > > from later posts in this thread), they would need to do a manual > > mkinitrd. > > No. If you boot off /dev/vda and update the kernel, mkinitrd will build > an initrd containing the virtio modules.Well, that may depend if you use mkinitrd or mkinitramfs, and your distribution mkinitramfs or mkinitrd may not be checking which device type he''s working on. Emre
Mark McLoughlin
2009-Jan-22 10:21 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Thu, 2009-01-22 at 11:06 +0100, Emre Erenoglu wrote:> > On Thu, Jan 22, 2009 at 10:20 AM, Mark McLoughlin <markmc@redhat.com> > wrote: > On Wed, 2009-01-21 at 13:47 -0500, Dustin Henning wrote: > > It is probably worth noting that if someone were to run a > kernel > > update with a package manager after getting this working > (using > > information > > from later posts in this thread), they would need to do a > manual > > mkinitrd. > > > No. If you boot off /dev/vda and update the kernel, mkinitrd > will build > an initrd containing the virtio modules. > > > Well, that may depend if you use mkinitrd or mkinitramfs, and your > distribution mkinitramfs or mkinitrd may not be checking which device > type he''s working on.Yes, since we''re on a Fedora mailing list, I''m assuming Fedora mkinitrd :-) Cheers, Mark.
Jan ONDREJ (SAL)
2009-Jan-22 10:36 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Thu, Jan 22, 2009 at 10:21:50AM +0000, Mark McLoughlin wrote:> On Thu, 2009-01-22 at 11:06 +0100, Emre Erenoglu wrote: > > > > On Thu, Jan 22, 2009 at 10:20 AM, Mark McLoughlin <markmc@redhat.com> > > wrote: > > On Wed, 2009-01-21 at 13:47 -0500, Dustin Henning wrote: > > > It is probably worth noting that if someone were to run a > > kernel > > > update with a package manager after getting this working > > (using > > > information > > > from later posts in this thread), they would need to do a > > manual > > > mkinitrd. > > > > No. If you boot off /dev/vda and update the kernel, mkinitrd > > will build > > an initrd containing the virtio modules. > > > > Well, that may depend if you use mkinitrd or mkinitramfs, and your > > distribution mkinitramfs or mkinitrd may not be checking which device > > type he''s working on.May be adding virtio disk drivers to initrd files always may help all users. I don''t know, if there is anything, why initrd have to be small. SAL
Mark McLoughlin
2009-Jan-22 10:39 UTC
Re: Switching from IDE to virtio_blk [qas Re: [Fedora-xen] Goodbye Xen on RH/Fedora?]
On Wed, 2009-01-21 at 14:45 +0000, Richard W.M. Jones wrote:> On Wed, Jan 21, 2009 at 03:06:43PM +0100, Emre Erenoglu wrote: > > On Wed, Jan 21, 2009 at 2:48 PM, Mark McLoughlin <markmc@redhat.com> wrote: > > > > > (This is all getting offtopic for fedora-xen, we should really move to > > > fedora-virt) > > > > > > On Wed, 2009-01-21 at 14:28 +0100, Jan ONDREJ (SAL) wrote: > > > > virtio_net works well, but I have trouble to boot from virtio_blk. > > > > > > > > I can add second disk as virto block device, but I can''t boot from > > > > first disk. > > > > > > When switching from IDE to virtio, you need to first build a new initrd > > > in the guest with e.g.: > > > > > > $> mkinitrd --with virtio_pci --with virtio_blk -f > > > /boot/initrd-$(kernelversion) $(kernelversion) > > > > > > You only need to do this once. After that, if a new kernel is installed > > > while you''re booted off a virtio disk, then mkinitrd will include the > > > modules automatically. > > > > > > You will also need to specify /dev/vdX on the kernel root= line and make > > sure your init script inside your initrd triggers the virtio drivers at boot > > so that the /dev/vdX are created. > > Yes I have to agree with Emre here - I don''t think it''s as simple as > just rebuilding mkinitrd. I got that far but gave up later on. > > /me checks notes ... > > Yup, I got as far as working out that you would have to edit fstab and > possibly /boot/grub/device.map and /boot/grub/menu.lst, before giving > up.Could this have been an x86_64 Fedora 9 xen guest? If so, you probably hit a nasty special case - the F9 x86_64 xen kernel didn''t have support for running 32 bit binaries like grub, so the bootloader would never have been installed into the MBR. That works fine for pygrub, but not with KVM''s real BIOS.> If anyone would like to fill in the wiki page here on the subject: > > http://wiki.libvirt.org/page/VirtioOkay, added some bits. Cheers, Mark.