On Fri, May 04, 2012 at 09:34:27PM +0200, Oliver Francke wrote:
> … showed the following:
>
> root@fcmsnode10:~# ovs-dpctl show
> system@vmbr1:
> lookups: hit:263209087 missed:904392 lost:0
> flows: 5
> port 0: vmbr1 (internal)
> port 1: eth1
> port 4: vlan10 (internal)
> port 7: tap410i1d0
> port 13: tap433i1d0
> port 15: tap377i1d0
> port 16: tap416i1d0
> port 18: tap287i1d0
> port 19: tap451i1d0
> port 21: tap822i1d0
> port 23: tap160i1d0
> port 24: tap376i1d0
> port 27: tap1084i1d0
> port 28: tap1085i1d0
> port 30: tap1113i1d0
> port 31: tap339i1d0
> port 38: tap760i1d0
> system@vmbr0:
> lookups: hit:11883603451 missed:6262740342 lost:114647219
> flows: 1295
> port 0: vmbr0 (internal)
> port 1: vlan146 (internal)
> port 2: eth0
> port 4: tap266i0d0
> port 8: tap323i0d0
> port 13: tap283i0d0
> port 31: tap410i0d0
> port 41: tap134i0d0
>
> and some more ~140 ports
Hmm, vmbr0 has a pretty high flow count and far too many lost packets.
I suggest, first, upgrading to OVS 1.4.1, which should reduce the lost
packet count, and then setting vmbr0's flow eviction threshold
significantly higher (which should reduce CPU usage) with:
ovs-vsctl set bridge vmbr0 other-config:flow-eviction-threshold=10000
The latter will probably become unnecessary with OVS 1.7, but that's
not released yet.
_______________________________________________
discuss mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/discuss
