[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] Bad network performances

On Thu, Jun 30, 2011 at 12:14 AM, Lorenzo Milesi
<lorenzo.milesi@xxxxxxxxxx> wrote:
> Hi.
> I've a Debian6 server, running kernel 2.6.32-5-xen-amd64, Xen hypervisor 
> 4.0.1-2.
> It runs 2VM, an HVM firewall and a PV Ubuntu 10.04.
> I'm experiencing terrible network performances on the main interface, shared 
> on both VMs and used as br0 on the dom0.
> At first, I'm seeing the issue that network is stuck for some 20/30 seconds 
> when the HVM machine boots. But that's not a big deal, since hopefully it 
> won't boot often.
> Now I'm having this trouble that causes the VM not to answer to network 
> requests, i.e. samba locked, other web services not responding... After some 
> time everything comes back to normal.
> The network card is an onboard Broadcom Corporation NetXtreme II BCM5716 
> Gigabit Ethernet.
> I see no error on syslog, dmesg, xend.log, xend-debug.log, and I don't know 
> where else to search!
> Any adivce on where to look at?

The first place I'd look would be the switch. Some switches are set so
that if there's arp change on a port, it would re-learn the topology
to prevent loop, and possibly causing all traffic to stop being
forwarded temporarily.

The EASIEST way to check this:
- install libvirt-bin, which would provide virbr0 and NAT
- set your HVM guest to use virbr0

The point is to make sure that you do NOT add a new MAC address to the
port (in this case by using NAT provided by virbr0). If that works,
than you can decide either to use virbr0, or setup such bridge
manually (e.g. through manual use of brctl and iptables), or try to
make vif-nat work (which, in some case, can be a PITA).


Xen-users mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.