[Xen-users] tap interface shuts down for unknown reasons

[James Roman <james_roman@xxxxxxxxxx>]

One of our guest servers has recently begun to disappear from the network after working properly for various periods of time. It seems that for some reason the tap interface for the guest domain disables itself. I have recently changed our network setup, but I am not sure that it is related.

Our current network configuration assigns a eth0 to Dom0. The guest domains use two bonded NICs (eth2 and eth3) which are bonded together (mode 4 across a Cisco channel group) into bond0. The guest domains bridge to bond0. eth0 and bond0 connect to two different VLANs. I have NOT configured trunking across any interface.

When the guests are started they are accessible by the network. After a period of time one of the guest domains (win2k3) disappears. I have two guest domains running (1 win2k3 and 1 Centos Linux), and it seems to only affect the windows guest. The event log on the guest server shows nothing unusual (in fact it appears to be functioning normally based on CPU and memory usage via xm list.)  The message log on Dom0 shows the following:

Aug 11 10:08:00 vs-001 kernel: xenbr0: port 3(tap1) entering disabled state
Aug 11 10:08:00 vs-001 kernel: device tap1 left promiscuous mode
Aug 11 10:08:00 vs-001 kernel: xenbr0: port 3(tap1) entering disabled state
Aug 11 10:08:00 vs-001 avahi-daemon[4269]: Interface tap1.IPv6 no longer relevant for mDNS.
Aug 11 10:08:00 vs-001 avahi-daemon[4269]: Leaving mDNS multicast group on interface tap1.IPv6 with address fe80::60a5:51ff:feb5:bf10.
Aug 11 10:08:00 vs-001 avahi-daemon[4269]: Withdrawing address record for fe80::60a5:51ff:feb5:bf10 on tap1.

ifconfig on Dom0 no longer lists tap1 (which is associated with the windows guest). Not knowing how to manually rebuild the interface, the only method I can find to restore contact to the guest domain is to destroy it and bring it back up. Dom0 message log shows the following and the guest domain is once again accessible via network and via VNC console port.

Aug 11 10:32:26 vs-001 kernel: xenbr0: port 6(vif5.0) entering disabled state
Aug 11 10:32:26 vs-001 kernel: device vif5.0 left promiscuous mode
Aug 11 10:32:26 vs-001 kernel: xenbr0: port 6(vif5.0) entering disabled state
Aug 11 10:32:29 vs-001 kernel: device tap1 entered promiscuous mode
Aug 11 10:32:29 vs-001 kernel: xenbr0: port 3(tap1) entering learning state
Aug 11 10:32:29 vs-001 kernel: xenbr0: topology change detected, propagating
Aug 11 10:32:29 vs-001 kernel: xenbr0: port 3(tap1) entering forwarding state
Aug 11 10:32:30 vs-001 kernel: device vif6.0 entered promiscuous mode
Aug 11 10:32:30 vs-001 kernel: ADDRCONF(NETDEV_UP): vif6.0: link is not ready
Aug 11 10:32:31 vs-001 avahi-daemon[4269]: New relevant interface tap1.IPv6 for mDNS.
Aug 11 10:32:31 vs-001 avahi-daemon[4269]: Joining mDNS multicast group on interface tap1.IPv6 with address fe80::20e8:6ff:fe68:45ca.
Aug 11 10:32:31 vs-001 avahi-daemon[4269]: Registering new address record for fe80::20e8:6ff:fe68:45ca on tap1.

First, is there any way to manually rebuild the tap interface when it goes down, so that I don't have to destroy the guest domain?

Second, is there any way to determine why the tap interface for that guest keeps going down? Is there a fundamental problem with the current network configuartion that is causing this anomaly?

Vitals:
Dom0 Centos 5.2
xen-3.0.3-64.el5_2.1

-- 

James D. Roman
Sr. Network Administrator
Science Systems and Application, Inc.
Phone: 301-867-2101

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users