Ask Your Question
0

Duplicate and lost packets with dvr between instances both with floating ips

asked 2015-06-03 00:52:54 -0500

anonymous user

Anonymous

Set up is OpenStack Kilo with Centos 7 using DVR with vxlan:

# cat /etc/redhat-release 
CentOS Linux release 7.1.1503 (Core) 

# uname -r
3.10.0-229.4.2.el7.x86_64

# rpm -qa|grep neutron
python-neutron-2015.1.0-1.el7.noarch
openstack-neutron-ml2-2015.1.0-1.el7.noarch
python-neutronclient-2.3.11-1.el7.noarch
openstack-neutron-2015.1.0-1.el7.noarch
openstack-neutron-common-2015.1.0-1.el7.noarch
openstack-neutron-openvswitch-2015.1.0-1.el7.noarch

When pinging between 2 instances that both have floating ips using their floating ips and are on different compute nodes there is high packet loss as well as duplicate packets.

[root@testy-test-3 ~]# ping -D -O 10.0.12.37
PING 10.0.12.37 (10.0.12.37) 56(84) bytes of data.
[1433306467.426536] 64 bytes from 10.0.12.37: icmp_seq=1 ttl=60 time=1.71 ms
[1433306468.427374] 64 bytes from 10.0.12.37: icmp_seq=2 ttl=60 time=0.682 ms
[1433306469.427393] 64 bytes from 10.0.12.37: icmp_seq=3 ttl=60 time=0.632 ms
[1433306470.427428] 64 bytes from 10.0.12.37: icmp_seq=4 ttl=60 time=0.669 ms
[1433306471.427428] 64 bytes from 10.0.12.37: icmp_seq=5 ttl=60 time=0.669 ms
[1433306472.427368] 64 bytes from 10.0.12.37: icmp_seq=6 ttl=60 time=0.631 ms
[1433306473.427391] 64 bytes from 10.0.12.37: icmp_seq=7 ttl=60 time=0.666 ms
[1433306475.426710] no answer yet for icmp_seq=8
[1433306475.427644] 64 bytes from 10.0.12.37: icmp_seq=9 ttl=60 time=0.865 ms
[1433306476.428339] 64 bytes from 10.0.12.37: icmp_seq=10 ttl=60 time=0.614 ms
[1433306477.428342] 64 bytes from 10.0.12.37: icmp_seq=11 ttl=60 time=0.630 ms
[1433306478.428254] 64 bytes from 10.0.12.37: icmp_seq=12 ttl=60 time=0.510 ms
[1433306479.428308] 64 bytes from 10.0.12.37: icmp_seq=13 ttl=60 time=0.571 ms
[1433306480.428366] 64 bytes from 10.0.12.37: icmp_seq=14 ttl=60 time=0.645 ms
[1433306481.428444] 64 bytes from 10.0.12.37: icmp_seq=15 ttl=60 time=0.697 ms
[1433306482.428349] 64 bytes from 10.0.12.37: icmp_seq=16 ttl=60 time=0.620 ms
[1433306483.428365] 64 bytes from 10.0.12.37: icmp_seq=17 ttl=60 time=0.623 ms
[1433306484.428282] 64 bytes from 10.0.12.37: icmp_seq=18 ttl=60 time=0.551 ms
[1433306484.428307] 64 bytes from 10.0.12.37: icmp_seq=18 ttl=60 time=0.564 ms (DUP!)
[1433306485.428417] 64 bytes from 10.0.12.37: icmp_seq=19 ttl=60 time=0.664 ms
[1433306486.428307] 64 bytes from 10.0.12.37: icmp_seq=20 ttl=60 time=0.584 ms
[1433306486.428334] 64 bytes from 10.0.12.37: icmp_seq=20 ttl=60 time=0.598 ms (DUP!)
[1433306487.428529] 64 bytes from 10.0.12.37: icmp_seq=21 ttl=60 time=0 ...
(more)
edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
0

answered 2015-06-05 23:39:24 -0500

ross-annetts gravatar image

The problem turned out to be an issue with the bonded ethernet interface being used for br-ex on both the compute nodes. Linux bonding round-robin mode was being used to connect the compute nodes to a pair of stacked switches. Changing the bonding mode to LACP and enabling LACP on the switch side resolved the issue.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2015-06-03 00:52:15 -0500

Seen: 530 times

Last updated: Jun 03 '15