This is an archived copy of the Xen.org mailing list, which we have
preserved to ensure that existing links to archives are not broken.
The live archive, which contains the latest emails, can be found at
http://lists.xen.org/
As far as I know, open vswitch not 'smart', it simply allow to
control flow of frames (not packets!). I don't know if xapi somehow
transmit new/old packets to old/new VM location, but anyway, this is
not OVS works. You can see what happens to OVS in
/etc/xensource/scripts/vif file (it's shell script to process
plug/unplug).
The magic is the grace arp, guest sending from newer location to
reconfigure all switches on net to send new frames to new location.
Some kernels (f.e. current centos 6) are broken and have
difficulties with migration. Others (f.e. 3.0 and most -xen kernels)
works fine with this.
The migration process (for PV) is very cooperative process, I think
you can easily detects migration process and reacts to it (by
issuing dhclient command f.e.). I don't know details, but I think
they can be easily digged out from xapi source. My assumption is the
xenstore usage, you can create application for domU which will
subscribe for changes for this key (key shall be digged out from
xapi sources) and reacts to it.
On 23.08.2011 19:07, Payal Shah wrote:
Hi,
I guess I didn't rephrase the question properly. I'm aware
that we could do live VM migration with different susbnets. I
wanted to know that with the help of Open vSwitch, can we
migrate to VM to a new subnet and the Open vSwitch is smart
enough to route the packets to the new subnet. Will the Open
vSwitch be aware that the VM is in a new subnet??? Also is there
any way that the VM could get an ip from the new subnet if there
is a dhcp configured, and can Open vSwitch route packets to the
new VM???
I hope this makes sense.. Thank you for the previous
responses
On Tue, Aug 23, 2011 at 5:11 AM, George
Shuklin-4 [via Xen] <[hidden email]>
wrote:
I do hear about guys who change route
tables on router (one /32 route
per VM) according to VM migration. I think this can be
used, but it
require very tight integration between router
reconfiguration and vm
migration.
Technically OVS is just switch, not a router, so you can
migrate VM
between different ip networks.
XCP do require to have same 'network' object on source and
destination
hosts but no one says they need to have same gateway.
В Пн., 22/08/2011 в 10:50 -0500, Payal Shah пишет:
> Hello ,
> Thank you for the reply. I got the results I was
expecting.
>
>
> I had one more question regarding open vSwitch and
live VM migration.
> Live VM migration requires that both the hosts
should be in the same
> subnet. If the hosts are in different subnets then
how can we do live
> VM migration? Can open vSwitch be used in this
scenario?? Or is there
> any other virtual switch that can route the packets
as per the new
> subnet?
>
>
> Thank you in advance
>
>
> On Wed, Aug 3, 2011 at 8:39 PM, George Shuklin
> <[hidden
email]> wrote:
> On XCP (1.1b, may be in XCP 1.0) there is
a note in
> xensource.log about %% of migration.
>
> And you can do xentrace in dom0 to see
all hypercalls (but you
> must understand what they means to get
needed information).
>
> On 04.08.2011 00:42, Payal wrote:
>
>
> I have two hosts as Xen Server
named host A and host B
> in the same network. A
> VM is created on Host A. Host A
and B have a shared
> storage using NFS. While
> doing migration from host A to
host B, how can I see
> the memory pages being
> migrated.
> If I run a small continuous
program on the VM residing
> on host A, will the
> migration complete first or the
sript running will
> complete first? How can I
> see that data is going in clear
text? Is there a tool
> to see the packets
> going through? I tried using
wireshark and spanning
> the ports, but I
> couldn't infer much. ANy help on
this appreciated.
> Thank you,
>
> --
> View this message in context:
> http://xen.1045712.n5.nabble.com/Live-VM-migration-tp4664019p4664019.html
> Sent from the Xen - User mailing
list archive at
> Nabble.com.
>
>
_______________________________________________
> Xen-users mailing list
Home