[Xen-users] Packets dropped on vif's

[Daniel Nielsen <djn@xxxxxxxxxx>]

Hi.

We've been running Xen in production for a couple of months now, and it has
been a fun and learning experience. And now we have seen some odd behaviour.

We boot our dom0's via PXE and use NFS for root-fs. Our domU's also use NFS
for everything and no swap. This allows for nice live migration. However, we
are experiencing low I/O performance in the domU's, and very often the
domU's loose the connection to the NFS server (not virtualized). The exact
message is: 

nfs: server X.X.X.X not responding, timed out

Using IOzone from a domU I'm able to provoke this to happen. IOzone in a
dom0 performs much better and doesn't result in errors in the syslog.

Also, I've noticed, with ifconfig that the vif's attached to the domU's are
dropping packets. monitoring this number, fx. with a 'watch ifconfig
vif24.0' indicates that is rising, fx when IOzone is running.

Googling and reading xen-users list archive have not yet pointed me in the
right direction. And now I'm out of ideas. Does anyone on this list have any
suggestion so we can improve IO in the domU's?

We're running Xen 3.0.3 on a debian stable with a custom kernel config. The
dom0's use eth0 for their networking, and the domU's all bridge the eth1
interface on the physical server.

/Daniel
Portalen
Support? Skriv til support@xxxxxxxxxxxxx 


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