[Pkg-xen-devel] xen: Issues with Networking on virtual hosts with Xen.3.0.3-1

[Niels Thykier]

Control: reassign -1 src:xen

Hi Stephen,

You filed this bug against xen-utils-3.0.3-1, but such a package does
not exist in Debian.  My guess is that it was once built from xen and I
have (tried to) reassign it accordingly.

That said, at first glance if your problem is related to a package from
Debian etch.  Unless this problem can be reproduced in a newer version
of the package, the maintainers of xen may close it as-is since Debian
no longer support etch (and haven't done so for quite a while).  Even if
the problem can be reproduced, the fix will probably not be distributed
via Debian's mirror.

For the benefits of the maintainers, Stephen's report was:

"""
We have a major issue at present which i'm finding difficult to resolve.
I have 3 x physical hosts all running Debian Etch and Xen 3.0.3-1
hypervisor on each running 4 x Debian Etch Virtual Machines. The
hypervisor is configured with network-nat and vif-nat. These hosts have
been inherited by me to look after from a previous colleague. They have
been working fine with no traffic issues for the last 2 years up until
the last week whereby I noticed virtual machines just randomly dropping
their connection to the host and vif generated interface. It drops the
connection for approx 30-40 seconds (sometimes longer) and the all of a
sudden decides to kick in and work again. I can still connect to the
virtual machine via xm console machinename.domainname.com and logon
whilst it looses it's network connection. On the virtual machine i run a
ifconfig and it shows the correct IP / Mac, however when i run route -n
it struggles and hangs to get the gateway address (ie; the VIF for the
host). I can ping the virtual machines static IP address, but can not
ping the default gateway (ie; the VIF). When i run a tcpdump on both the
virtual machine eth0 and the physical host vif interface i can see a
load of arp requests on both. ie; who-is at xx.xx.xx.xx tell
xx.xx.xx.xx. After 30-40 seconds it decides to clear itself and continue
the traffic. It must have been a long standing issue over the 2yrs
however as we are generating more and more traffic to the virtual
machines i think it's becoming overloaded and stopping the VM from
accepting any incoming/outgoing connections until it clears itself
somehow. Problem is i don't know how to troubleshoot this within Xen. I
tried using the xen-bugtool however i get a python error stating
"ImportError: No module named xen.util". What steps do i need to carry
out to troubleshoot and resolve my issue?
"""

~Niels