Re: [Qemu-devel] live migration between different qemu versions

[Dr. David Alan Gilbert]

* Alexey (address@hidden) wrote:
> 
> 
> On 2016-01-12 13:04, Dr. David Alan Gilbert wrote:
> >* Alexey (address@hidden) wrote:
> >>Hello,
> >>
> >>On 2016-01-12 12:19, Dr. David Alan Gilbert wrote:
> >>>* Alexey (address@hidden) wrote:
> >>>>Hi David.
> >>>>
> >>>>On 2016-01-11 22:51, Dr. David Alan Gilbert wrote:
> >>>>>* Alexey (address@hidden) wrote:
> >>>>>>Hello.
> >>>>>
> >>>>>Hi,
> >>>>>
> >>>>>>I have two servers between which I need have live migration.
> >>>>>>
> >>>>>>First server have QEMU emulator version 2.3.0
> >>>>>>Second server have QEMU emulator version 2.5.0
> >>>>>>
> >>>>>>Migration command look like this:
> >>>>>>/usr/bin/virsh migrate --live DOMAIN_NAME --migrateuri
> >>>>>>tcp://second.server
> >>>>>>qemu+ssh://second.server/system --verbose
> >>>>>>
> >>>>>>I got next error on console:
> >>>>>>error: operation failed: migration job: unexpectedly failed
> >>>>>>
> >>>>>>And next error in log on second server:
> >>>>>>2016-01-11T15:07:12.888647Z qemu-system-x86_64: Length too large:
> >>>>>>/address@hidden/efi-virtio.rom: 0x32000 > 0x2d000: Invalid argument
> >>>>>>2016-01-11T15:07:12.888685Z qemu-system-x86_64: error while loading
> >>>>>>state
> >>>>>>for instance 0x0 of device 'ram'
> >>>>>>2016-01-11T15:07:12.889733Z qemu-system-x86_64: load of migration
> >>>>>>failed:
> >>>>>>Invalid argument
> >>>>>>
> >>>>>>In my libvirt domain I have option <rom bar="off"/>.
> >>>>>>
> >>>>>>I need to do a live migration in any possible way. How can I do this
> >>>>>>one?
> >>>>
> >>>>>Is the efi-virtio ROM that you're using on the two versions the same
> >>>>>identical size?
> >>>
> >>>>No, I have different efi-virtio.rom files, and I try to find a way which
> >>>>allows me not to copy the same efi-virtio.rom to all servers.
> >>>
> >>>Right, the error does look like it's just complaining about the ROM size:
> >>>
> >>>  2016-01-11T15:07:12.888647Z qemu-system-x86_64: Length too large:
> >>>  /address@hidden/efi-virtio.rom: 0x32000 > 0x2d000: Invalid argument
> >>>
> >>>It copes with small variations, but as soon as it flips over a 4K (?)
> >>>boundary then it has nowhere to put the extra data coming from the
> >>>larger ROM.
> >>old server with qemu 2.3.0:
> >>-rw-r--r-- 1 root root 203264 Nov 21  2014
> >>/usr/lib/ipxe/qemu/efi-virtio.rom
> >>
> >>new server with qemu 2.5.0:
> >>-rw-r--r-- 1 root root 184320 Oct 19  2014
> >>/usr/lib/ipxe/qemu/efi-virtio.rom
> >>
> >>See, that destination server has smaller rom than source.
> >>I interested in way which I could do live migration between different
> >>ROMs.
> >
> >You can't do it if the ROMs dont approximately match in size;  what
> >happens here
> >is that the 2.3.0 machine allocates just over 203k for the ROM
> >(203264=0x31A00
> >and it's rounded that upto 0x32000).
> >On the destination it's size is based on the 184320 (=0x2d000), so it
> >hasn't
> >got room to receive the larger ROM.
> >One way to fix this is to pad the ROMs; so just adding blank space to the
> >end
> >of the newer efi-virtio.rom to make it the same size as the old one.
> Ok, thanks. I'll try to do this.
> Tell me please right way to append zeros to "BIOS (ia32) ROM Ext. (137*512)"
> file?

I'd use dd; something like:
  dd if=/dev/zero bs=1 count=18944 >> theromfile
I think that should do it (203264-184320=18944)

> It will be grate, if I could specify ROM size as option to qemu.

Yeh.

Dave

> 
> >>Also I interested in migration in both side: from old to new and from
> >>new to
> >>old.
> >
> >OK, but note that backwards migration isn't tested much in
> >upstream qemu.
> >
> >>>>>Also, can you post the commandline that you're using please.
> >>>>I already done this in my first email: virsh migrate...
> >>>>>>/usr/bin/virsh migrate --live DOMAIN_NAME --migrateuri
> >>>>>>tcp://second.server
> >>>>>>qemu+ssh://second.server/system --verbose
> >>>Do you mean running domain command line?
> >>>I meant for the QEMU itself; but if we've already found the ROMs are
> >>>different sizes, lets fix that first and see how you go.
> >>Do you mean command line of running domain?
> >>
> >>qemu-system-x86_64 -enable-kvm -name 208664-73 -S -machine
> >>pc-i440fx-1.7,accel=kvm,usb=off -cpu
> >>kvm64,+lahf_lm,+popcnt,+sse4.2,+sse4.1,+ssse3,+ss,-mca,-mce -m 1024
> >>-realtime mlock=off -smp 4,sockets=4,cores=1,threads=1 -uuid
> >>08cf6bf2-a701-4a00-a0cb-4ecbe514c3d7 -no-user-config -nodefaults
> >>-chardev 
> >>socket,id=charmonitor,path=/var/lib/libvirt/qemu/208664-73.monitor,server,nowait
> >>-mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc
> >>-no-shutdown
> >>-boot strict=on -device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2
> >>-device
> >>virtio-scsi-pci,id=scsi0,num_queues=4,bus=pci.0,addr=0x4 -device
> >>virtio-serial-pci,id=virtio-serial0,bus=pci.0,addr=0x6 -drive 
> >>file=/dev/disk/vbd/208664-73,if=none,id=drive-scsi0-0-0-0,format=raw,cache=none,discard=unmap,aio=native
> >>-device 
> >>scsi-hd,bus=scsi0.0,channel=0,scsi-id=0,lun=0,drive=drive-scsi0-0-0-0,id=scsi0-0-0-0,bootindex=2
> >>-netdev tap,fd=53,id=hostnet0,vhost=on,vhostfd=54 -device 
> >>virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:00:40:b9,bus=pci.0,addr=0x3,rombar=0
> >>-chardev pty,id=charserial0 -device
> >>isa-serial,chardev=charserial0,id=serial0 -chardev 
> >>socket,id=charchannel0,path=/var/lib/libvirt/qemu/208664-73.agent,server,nowait
> >>-device 
> >>virtserialport,bus=virtio-serial0.0,nr=1,chardev=charchannel0,id=channel0,name=org.qemu.guest_agent.0
> >>-device usb-mouse,id=input0 -device usb-kbd,id=input1 -vnc
> >>[::]:12,password
> >>-device VGA,id=video0,vgamem_mb=16,bus=pci.0,addr=0x2 -device
> >>virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5 -object
> >>rng-random,id=objrng0,filename=/dev/random -device 
> >>virtio-rng-pci,rng=objrng0,id=rng0,max-bytes=1024,period=2000,bus=pci.0,addr=0x7
> >>-msg timestamp=on
> >
> >OK, nothing too unusual there.
> >
> >Dave
> >
> >>>
> >>>Dave
> >>>
> >>>>
> >>>>Regards,
> >>>>Alexey Mochkin.
> >>>--
> >>>Dr. David Alan Gilbert / address@hidden / Manchester, UK
> >--
> >Dr. David Alan Gilbert / address@hidden / Manchester, UK
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK