[ovs-discuss] internal error parsing flow key

[Ben Pfaff]

On Tue, May 21, 2019 at 11:30:05AM +0200, Jakub Jursa wrote:
> we're running openvswitch as Neutron ml2 agent in OpenStack Rocky
> environment and we're using ovs offloading (
> https://docs.openstack.org/neutron/rocky/admin/config-ovs-offload.html
> ), the traffic mainly consists of multicast. Problem is that we're
> seeing thousands of logs like this
> 
> 2019-05-21T09:20:04.405Z|00048|odp_util(handler13)|ERR|Dropped 13 log
> messages in last 57 seconds (most recently, 1 seconds ago) due to
> excessive rate
> 2019-05-21T09:20:04.405Z|00049|odp_util(handler13)|ERR|internal error
> parsing flow key
> recirc_id(0),dp_hash(0),skb_priority(0),in_port(2),skb_mark(0),ct_state(0),ct_zone(0),ct_mark(0),ct_label(0),eth(src=84:b5:9c:f6:1c:00,dst=01:00:5e:0a:0a:ee),eth_type(0x8100),vlan(vid=1017,pcp=6),encap(eth_type(0x0800),ipv4(src=10.1.36.65,dst=227.10.10.238,proto=2,tos=0xc0,ttl=1,frag=no))

This is for IGMP traffic (proto=2), so it is undoubtedly related to the
multicast traffic you mentioned.

OVS 2.10, which you report using, did not have very good logging
messages for debugging this kind of problem.  The latest OVS on master
has much better log messages, that might help track the problem down.
Regrettably, these are not part of any released version of OVS yet,
although they will be in 2.12.

Do you see bad behavior or performance, or just nasty log messages?