[ovs-discuss] DPDK OVS with vhostuser problem

Loftus, Ciara ciara.loftus at intel.com
Thu Mar 3 15:28:54 UTC 2016


> Hi
> 
> I followed the instructions in INSTALL.md and INSTALL.DPDK.md.
> I have created two ports in a bridge, and set the flows 1<->2.
> Doing ovs-ofctl dump-flows br0 shows it right.
> The purpose of this setup is to measure latency of those ports (using a simple
> ping).
> 
> I invoke a VM, with those 2 ports. It gives those messages:
> 2016-03-02T17:08:17.410601Z qemu-system-x86_64: -netdev type=vhost-
> user,id=mynet1,chardev=char1,vhostforce: chardev "char1" went up 2016-
> 03-02T17:08:17.410851Z qemu-system-x86_64: -netdev type=vhost-
> user,id=mynet2,chardev=char2,vhostforce: chardev "char2" went up
> 
> From within the VM (qemu), I have eth0 and eth1.
> I configure them as:
> ifconfig eth0 1.1.1.1
> ifconfig eth1 1.1.1.2
> 
> Ping will fail though:
> ping 1.1.1.2 -I eth1 obviously works.
> ping 1.1.1.2 -I eth0 fails.
> 
> In the host, I wanted to see some statistics, but ovs-ofctl dump-flows br0
> fails, with this message:
> 
> ovs-ofctl: /usr/local/var/run/openvswitch/br0.mgmt: failed to open socket
> (Connection refused)
> 
> I see that this issue is right after invoking the VM.
> 
> 1. What do I do wrong?
> 2. How can I track the packets to see why they are not forwarded correctly?

Hi Eli,

Can you give more information about your set up please - OVS / QEMU / DPDK tags / commit ids?
When ovs-ofctl dump-flows fails with 'connection refused', are the ovs-vswitchd/ovsdb-server processes still running? Or has something crashed?

Thanks,
Ciara

> 
> Thanks,
> Eli
> 




More information about the discuss mailing list