We recently upgraded a number of hypervisors from 2.4.x to OVS 2.5.x
(mainly to resolve the vswitchd 100% cpu issue fixed in
3e8cc1b44f295fe2e01e52d3f567ba9e8f9d0aa2 ).
However we seem to be running into a new issue. LACP bonded interfaces
created with balance-tcp and rebalancing disabled seem to use only a
single interface for all flows. Setting bond-rebalance-interval to 10000
and back to 0 seems to make flow balancing work again.
Sample output from a host with 2 10gbit interfaces below:
[root@mccppod013-hv09 ~]# iperf3 -c mccppod013-hv07 -P 10 | grep SUM | tail -n1
[SUM] 0.00-10.00 sec 11.4 GBytes 9.78 Gbits/sec receiver
[root@mccppod013-hv09 ~]# ovs-vsctl set port bond0
other_config:bond-rebalance-interval=10000
[root@mccppod013-hv09 ~]# ovs-vsctl set port bond0
other_config:bond-rebalance-interval=0
[root@mccppod013-hv09 ~]# iperf3 -c mccppod013-hv07 -P 10 | grep SUM | tail -n1
[SUM] 0.00-10.00 sec 21.5 GBytes 18.4 Gbits/sec receiver
Bond config:
---- bond0 ----
bond_mode: balance-tcp
bond may use recirculation: yes, Recirc-ID : 1
bond-hash-basis: 0
updelay: 0 ms
downdelay: 0 ms
lacp_status: negotiated
active slave mac: 94:57:a5:65:fd:f0(eno49)
slave eno49: enabled
active slave
may_enable: true
slave eno50: enabled
may_enable: true
[root@mccppod013-hv09 ~]# ovs-vsctl list port bond0
_uuid : 86d5f865-04a1-43cd-8b20-94e40a5e3aca
bond_active_slave : "94:57:a5:65:fd:f0"
bond_downdelay : 0
bond_fake_iface : false
bond_mode : balance-tcp
bond_updelay : 0
external_ids : {}
fake_bridge : false
interfaces : [63d0543d-7f10-4ae0-a0dc-c69b23f65442,
b8b4d6ab-e810-4356-81a9-f75e17dcea9a]
lacp : active
mac : []
name : "bond0"
other_config : {bond-rebalance-interval="0"}
--
Sten Spans
"There is a crack in everything, that's how the light gets in."
Leonard Cohen - Anthem
_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss
