Thu, Aug 04, 2016 at 05:45:17PM CEST, mariusz.g.mazur(a)gmail.com wrote:
I'm running kernel 4.7.0 and teamd 1.25. The following config
should set
all of the configured links to the same MAC address. It doesn't:
[root@vc1n3 ~]# teamdctl sw2 config dump
{
"device": "sw2",
"link_watch": {
"delay_up": 5000,
"name": "ethtool"
},
"ports": {
"eno4": {},
"eno5": {},
"enp17s0f1": {}
},
"runner": {
"active": true,
"fast_rate": true,
"name": "lacp",
"tx_hash": [
"eth",
"ipv4",
"ipv6"
]
}
}
And the links are:
3: enp17s0f1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master
sw2 state UP mode DEFAULT group default qlen 1000
link/ether 00:0a:f7:06:1e:6a brd ff:ff:ff:ff:ff:ff
6: eno4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master sw2
state UP mode DEFAULT group default qlen 1000
link/ether 00:0a:f7:06:1e:6a brd ff:ff:ff:ff:ff:ff
7: eno5: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master sw2
state UP mode DEFAULT group default qlen 1000
link/ether f2:8a:c5:1a:67:47 brd ff:ff:ff:ff:ff:ff
As you can see, eno5 was changed (from the NIC's default
'98:be:94:45:31:f5'), but to something seemingly random. Interestingly
enough if I explicitly add "hwaddr": "00:0a:f7:06:1e:6a" to the
config,
all three NICs are set up properly and I have no issues.
But as it stands, the default behavior is buggy for me.
Indeed looks like a but. Could you please try libteam/teamd from our
git? There's one lacp fix there. But I think it won't help.
Could you please send how exactly are you bringing the team device and
ports up? This looks like a possible race condition in teamd. I don't
think this is kernel related.
Thanks!