On Thu, Apr 28, 2005 at 03:40:10PM +0300, Martti Kuparinen wrote:
> Hi!
> 
> I'm having bridge and/or vlan problems with my Xen server. Here's
> how things are configured on this NetBSD/xen 3.99.3 server:
> 
> wm0 is used only by the Xen domain0.
> 
> wm1 is connected to a switch and this port transports all VLANs
> to /from the Xen server in tagged mode. I create the VLANs and
> bridges withn these kind of /etc/ifconfig.vlanNNNN files:
> 
> 
> create
> vlan 1208 vlanif wm1
> !ifconfig bridge1208 create
> !brconfig bridge1208 add vlan1208 stp vlan1208 up
> 
> 
> When I start the virtual machines each xvifN.M is connected to
> one of the bridgeNNNN interfaces with
> 
> # brconfig bridgeNNNN add xvifN.M stp xvifN.M up
> # ifconfig xvifN.M up
> 
> All is fine when only 3 VMs are running (in other words 4 entities
> are connected to the bridge, i.e. vlanNNNN, xvif1.M, xvif2.M, xvif3.M).
> 
> When I start the 4th (or stop later any of the 4 the running VMs)
> I see a 30 second stop in all network activity. Same thing happens
> when I start more VMs, each start introduces 30 sec network stop.
> 
> To make things harder to analyze the 4th VM isn't always the
> trouble maker, sometimes it's the 10th started VM which makes the
> network to halt for a while.
> 
> Any ideas what's going on? Below is my kernel config diff against
> the generic XEN0 kernel.

30s, this really could be caused by STP. Did you try disabling stp
on the bridge ?

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--