[ovs-discuss] [Q]packet loss in ovs and bonding configuration issue

Jesse Gross jesse at nicira.com
Fri Nov 22 09:00:04 UTC 2013


On Thu, Nov 21, 2013 at 5:34 PM, Soyoung Jeong <jeong.soyoung at gmail.com> wrote:
> Hi,
>
> I'm using OVS in xenserver 6.0.2  and the default OVS is 1.0.99 (very old
> version!)
>
> I have a bonded port in the  OVS and in that port I found some packets lost
> in the OVS.
>
>
> [root at xcnc133 ~]# ovs-dpctl show xapi12
>
> lookups: frags:0, hit:12808235384, missed:2091274971, lost:15849
>
> port 0: xapi12 (internal)
>
> port 1: bond0 (internal)
>
> port 2: eth6
>
> port 3: eth4
>
> port 4: xapi14 (internal)
>
> port 5: xapi13 (internal)
>
> port 6: vif6.0
>
> port 7: vif169.0
>
> port 8: vif171.0
>
> port 9: vif12.0
>
> ...
>
>
> There is no traffic congestion and no resource starvation in dom0.
>
> So I investigated the bonding configuration and I think something's wrong
> about my bonding type.
>
>
> [root at xcnc137 log]# ovs-vsctl list port bond0
>
> _uuid                        : 14ac8377-fd93-48a6-95c6-b5884d6374cd
>
> bond_downdelay      : 200
>
> bond_fake_iface       : true
>
> bond_mode             : balance-slb
>
> bond_updelay          : 31000
>
> external_ids             : {}
>
> fake_bridge             : false
>
> interfaces                : [cbfa82ff-fa2b-46ba-9f6c-1fada08cb3f1,
> de9e7b9e-4fd0-429d-bc14-ba0cbd57ad00]
>
> lacp                         : off
>
> mac                         : "a0:36:9f:14:18:a0"
>
> name                       : "bond0"
>
> other_config        : {bond-detect-mode=carrier, bond-miimon-interval="100"}
>
> qos                       : []
>
> tag                      : []
>
> trunks                 : []
>
>
> I read the OVS manual for bonding type and it says:
>
> balance-slb: Balances flows among slaves based on source MAC address and
> output VLAN, with periodic rebalancing as traffic patterns change.
> active-backup:  Assigns all flows to one slave, failing over to a backup
> slave when the active slave is disabled. This is the only bonding mode in
> which interfaces may be plugged into different upstream switches.
>
> According to the explanation, I think I have a wrong configuration in my OVS
> because I have different upstream switches for bonded port,,
>
> Do you think it could be the reason of some packet loss in OVS? Are those
> rules still effective in OVS 1.0.99?

Yes, you need to use active-backup in this case.



More information about the discuss mailing list