[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-API] XCP 1.1 OVS and active-backup mode seeing traffic on both interfaces

  • To: xen-api@xxxxxxxxxxxxxxxxxxx
  • From: brooks@xxxxxxxxxxx
  • Date: Sat, 21 Jan 2012 16:39:10 -0800 (PST)
  • Delivery-date: Sun, 22 Jan 2012 00:39:36 +0000
  • List-id: Discussion of API issues surrounding Xen <xen-api.lists.xensource.com>

I'm seeing traffic on both interfaces of an active-backup bond. This is a bond of eth0 and eth1 with both management and vm traffic. Given the requirement that a bonded management interface be in active-backup mode and since balance-slb doesn't work unless the ports are connected to the same switch (or stack) I have the bond in active-backup mode:

# xe pif-list uuid=c5145148-5a9d-fe94-8129-d66fb26cb0f4 params=other-config
other-config (MRW)    : bond-mode: active-backup

#ovs-vsctl list port
_uuid               : a3320c0c-5511-4705-9355-fadbe8e06b90
bond_downdelay      : 200
bond_fake_iface     : true
bond_updelay        : 31000
external_ids        : {}
fake_bridge         : false
interfaces : [59378258-2394-4c53-8708-e92106488a30, c2c11a5e-b6d5-449c-b05f-b4d5ed2e53f1]
mac                 : "00:1b:24:3e:09:35"
name                : "bond0"
other_config : {bond-miimon="100", bond-mode=active-backup, bond-use_carrier="1"}
qos                 : []
tag                 : []
trunks              : []

# ovs-appctl bond/show bond0
updelay: 31000 ms
downdelay: 200 ms
next rebalance: 9144 ms
slave eth1: enabled
        active slave
        hash 213: 5 kB load
slave eth0: enabled
        hash 0: 1 kB load
        hash 51: 0 kB load
        hash 251: 0 kB load
        hash 255: 0 kB load

It would seem like it's setup and working correctly with eth1 being the active interface. So, why am I seeing VM traffic on eth0?

# tcpdump -vv -i eth0

16:25:46.653749 IP (tos 0x0, ttl 245, id 40076, offset 0, flags [none], proto: ICMP (1), length: 84) > vpc: ICMP echo request, id 27167, seq 18452, length 64

16:25:46.653805 IP (tos 0x0, ttl 64, id 31595, offset 0, flags [none], proto: ICMP (1), length: 84) vpc > ICMP echo reply, id 27167, seq 18452, length 64

The problems I'm seeing are intermittent packet loss to both the VM and to dom0 as well as the interfaces being dead after a live migration until I send traffic from the VM. This same environment works fine under XenServer 5.5 (running in linux bridge mode). There's very little knowledge available in relation to OVS so any guidance would be appreciated.

xen-api mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.