[CentOS-virt] KVM - Network bridge issue - lots of overruns

AemNet sysadmin-aemnet at aemnet.it
Mon Sep 2 14:49:39 UTC 2013


Hi everybody

  I've a very big network problem with some server running CentOS with 
KVM and some linux-like VMs on the top of it.
The vm's network is bridged with a physical dedicate NIC (an Intel 
PRO/1000) with 2.4.14-NAPI driver (but the problem persists with various 
versions of the driver) and the CentOS version is 6.4 (same problem with 
a 5.7 sever) with all the last updates.
Randomly I get a lots of overruns on the vnet device on the physical 
server ad obviously some packet were lost in the VM.

brctl show br1
bridge name     bridge id               STP enabled     interfaces
br1             8000.002655e7590d       no              eth3
                                                         vnet1

ifconfig vnet1
vnet1     Link encap:Ethernet  HWaddr FE:54:00:BF:27:12
           inet6 addr: fe80::fc54:ff:febf:2712/64 Scope:Link
           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
           RX packets:13358349859 errors:0 dropped:0 overruns:0 frame:0
           TX packets:21831525660 errors:0 dropped:0 overruns:1746068 
carrier:0
           collisions:0 txqueuelen:500
           RX bytes:5080769358430 (4.6 TiB)  TX bytes:16548002754111 
(15.0 TiB)


As you can see there are a huge numbers of overruns also if the physical 
NIC is dedicate solely to this VM and the network traffic is less than 
150Mbps.
The server is a quite new machine (HP DL120 G6) but I have the same 
problem on an old DL360 G5 (but this is running about 10 vms).
What can I do to solve the problem?
My idea was to use the KVM passthrough functionality for the network 
device, but this method isn't good for a server that runs more than 2 
vms as my 360G5...

Thanks
Beppe

PS sorry for my bad english


More information about the CentOS-virt mailing list