Ask Your Question

OVS Intermittent ARP reply blocking

asked 2017-08-03 15:20:47 -0500

Sylain gravatar image

I have a the latest Openstack Newton installed on CentOS7.

I have a intermittent problem where one of my VM does not get ARP replies, for some time. The ARP replies are coming from an external host (

I do see the ARP replies reaching the NIC using tcpdump (tcpdump -i p1p1 arp), but they are being dropped in the OVS bridge.

This is intermittent, after minutes, the ARP will finally make its way up to the VM.

When the problem is on, I see the following flows: (ovs-ofctl dump-flows br-voip)

recirc_id(0),in_port(7),eth(src=00:50:56:8d:21:de,dst=fa:16:3e:cf:82:05),eth_type(0x8100),vlan(vid=262),encap(eth_type(0x0806)), packets:1, bytes:64, used:0.333s, actions:drop

The packets count may go as high as a few hundreds, and after minutes, I see the the previous flow being replaced by:

recirc_id(0),in_port(7),eth(src=00:50:56:8d:21:de,dst=fa:16:3e:cf:82:05),eth_type(0x8100),vlan(vid=262,pcp=0),encap(eth_type(0x0806)), packets:1, bytes:64, used:3.882s, actions:pop_vlan,10│

And the ARP reply finally makes it through...

What is the problem ?

Thanks for your help.

edit retag flag offensive close merge delete

1 answer

Sort by » oldest newest most voted

answered 2017-08-07 16:12:23 -0500

Sylain gravatar image

Nailed it. An Ethernet driver bug was causing transmitted (some) ARP request to be copied back in rx on the same port.

OVS was learning that mac fa:16:3e:cf:82:05 source was p1p1, so the ARP reply (coming from p1p1) was dropped by OVS.

OVS was doing the right thing from the start.

edit flag offensive delete link more


Hi, I think am running into same issue. How do you know it was driver bug? How did you patch it?

Sudhindra gravatar imageSudhindra ( 2018-08-16 04:06:51 -0500 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2017-08-03 15:20:47 -0500

Seen: 1,168 times

Last updated: Aug 07 '17