Live Migration and Bridge Settings

Live Migration and Bridge Settings

[Christopher Boumenot]

Greetings,

I'm having a problem with live migration, and the default network 
settings used by Xen.  I was hoping someone could give me advise on what 
I'm doing wrong.

I currently have three Linux installations, two of the PCs are running 
SuSe 9.1/Xen 2.0.5 with the stock xenU and xen0.  The third PC is 
running coLinux (debian) on a WindowsXP machine that exports the Xen's 
test ttylinux distribution in a flat file.

For this setup I have both Xen hosts NFS mount a directory containing 
the image.  I start a ping process on one of the Xen hosts to monitor 
network connectivity.  I then live migrate the VM from one to host to 
the other.  The ping process will usually fail for approximatley 15-20 
seconds (sometimes upto ~50 seconds), and then eventually restart.  The 
VM migrates properly.

I've looked at several tcpdump traces to see if I could find the 
unsolicited arp being sent, but I don't.  In fact I see the Xen host 
with the ping process re-issue the arp to find the IP address, and then 
the pings succeed.  I added some debug code to the netfront driver to 
verify that the arp is in fact being queued, and everything looks fine.

I started to suspect that something is wrong with my bridge setup.  I 
started out using the default Xen vif-bridge scripts, but I found that 
the VM only had connectivity to dom0.  The VM wasn't able to ping other 
hosts on the subnet.  From my googling, I discovered that people had 
similar problems, and based on their success I changed my init scripts. 
  Instead of Xen creating the bridge, I create my own, and label it 
eth0.  I do this before the init scripts configure eth0.

/sbin/ip link set name hw-eth0 dev eth0
/sbin/brctl addbr eth0
/sbin/brctl addif eth0 hw-eth0
/sbin/ip link set up hw-eth0

This works well for me.

I've also tried the routing setup, but I still had problems reaching 
other hosts from the VM.

While I'm able to migrate domains, I am unable to match the times I had 
read in the literature (<1 second).  Any ideas on where I went wrong, or 
areas that I should be looking at?


Thanks,
Christopher



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click

RE: Live Migration and Bridge Settings

[Ian Pratt]

> For this setup I have both Xen hosts NFS mount a directory containing 
> the image.  I start a ping process on one of the Xen hosts to monitor 
> network connectivity.  I then live migrate the VM from one to host to 
> the other.  The ping process will usually fail for 
> approximatley 15-20 
> seconds (sometimes upto ~50 seconds), and then eventually 
> restart.  The 
> VM migrates properly.

How are your various machines interconnected? Are they all on the same
subnet (via an Ethernet switch), or are they routed?

> I started to suspect that something is wrong with my bridge setup.  I 
> started out using the default Xen vif-bridge scripts, but I 
> found that 
> the VM only had connectivity to dom0.  The VM wasn't able to 
> ping other 
> hosts on the subnet.  From my googling, I discovered that people had 
> similar problems, and based on their success I changed my 
> init scripts. 
>   Instead of Xen creating the bridge, I create my own, and label it 
> eth0.  I do this before the init scripts configure eth0.

Calling it eth0 seems a very odd thing to do, but I guess it would work.

Ian


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_ide95&alloc_id\x14396&op=click

Re: Live Migration and Bridge Settings

[Christopher Boumenot]

Ian Pratt wrote:
>>For this setup I have both Xen hosts NFS mount a directory containing 
>>the image.  I start a ping process on one of the Xen hosts to monitor 
>>network connectivity.  I then live migrate the VM from one to host to 
>>the other.  The ping process will usually fail for 
>>approximatley 15-20 
>>seconds (sometimes upto ~50 seconds), and then eventually 
>>restart.  The 
>>VM migrates properly.
> 
> 
> How are your various machines interconnected? Are they all on the same
> subnet (via an Ethernet switch), or are they routed?
> 

All of the hosts are connected through a switch, and they are all on the 
same subnet.


Chris



-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click