On Wednesday 14 September 2005 00:45, Robert Hulme wrote:
> I'm having some problems on xen 2.0.7
>
> I've setup 5 domains and linked them together such that you have A <->
> B <-> C <-> D <-> E. I have two problems:
Did you use several bridges inside dom0 (one for A+B, one for B+C and so on)
to connect them, or did you directly connect vifs from different domUs
without involving dom0?
> 1. When pinging from A to E (or E to A) the first ping takes ~ 34ms,
> following pings take ~0.8ms then after about 10 packets it goes up to
> ~30ms, then it goes back to normal again and back up at a random time
> later, and so on. So I get these random spikes.
> 2. After a random number of packets (usually about 30) the pings stop
> being delivered to E (when sending to A) this happens for a long time
> then randomly a batch (usually 10-20) arrive then it goes again, and
> so on. This is so bad that the packet loss over ~1500 packets is 85%
>
> I am at a loss as to how to fix this - but it's a serious problem
> (obviously) as was intending to use Xen for a network QoS simulation
Networking problems (lost pings, NFS issues, SSH hanging, ...) with xen-2.0.x
were reported quite often in the last few weeks, search for those threads,
some of them contained some "magic" you might try, like disabling checksum
offloading, etc...
Otherwise: try xen 3, or don't run services requiring reliable networking
inside domU on xen 2 :(
I'm running several database applications inside domUs, they're more demanding
in terms of HDD IO and CPU, so the networking issues don't affect me that
much, but if you find a solution, let me know :)
/Ernst
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home