[Gllug] Fun with Xen (not!)

[Henrik Morsing]

Hi,

I'm trying to help my dad out with a weird Xen problem... Set-up has worked for six months but one of the Xen bridges (it seems) has gone into a frequent but intermittent fault.

IBM server with Xen 3.2.1 has two NICs bonded to a switch. On top of bind0 is five VLANs, only one of this has now gone funny... The VLANs are linked to bridges to the VIFs, which all go into one DomU.o

So, the faulty one is eth0+eth1 -> bond0 -> bond0.16 -> xenbr16 -> vif2.2 -> eth2.

This morning we decided to create a new VLAN and chain into the domU, as the other ones worked fine, we assumed a new one would also work, so we created bond0 -> bond0.20 -> xenbr20 -> vif3.2 -> eth2. It doesn't work either...

I've been trying to trace traffic on the devices ping from a PC into the domU. Basically, it will work from a number of seconds (anything from a couple to in rare ocaasions, thousands) but then incoming ICMP seems to stop at bond0.20, not arriving at xenbr20, and the domU starts sending arp replies (!) out which only shows at xenbr20 (not at bind0.20 or vif3.2!).

We've trying manually entering arp entries but the arp calls are a red herring really, it's just a symptom of whatever it is that's wrong.

I'm not that experienced with Xen or bridges... I've tried setting STP on an off but without success. bridge-nf are all set to 1, which I read somewhere they should be.

Any ideas? He works at a school and all the classroom PCs the teachers use for lesson plans and presentations are out of action, so it's quite critical :(


Thanks
--
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug