GRE packets flooding, taking network down

asked 2014-11-06 04:22:26 -0500

StephenKing gravatar image

I've been testing Mirantis OpenStack 5.1 on a 3-controller HA setup with 3 compute nodes using GRE tunnels for networking.

After the initial deployment, everything went fine. Now I came to the incredible weird idea that I could just see what happens, when I reboot all nodes.

It now looks like the GRE tunnels do not come up. I see zillions of the following packets:

node-19 > node-16: GREv0, key=0x0, length 78: IP6 fe80::64a0:eeff:fe9c:5193 ip6-allrouters: ICMP6, router solicitation, length 16

These packets are sent from all to all nodes and the nodes also receives packets by the others (on the eth2 interface, which has the Storage + Management network). The interface load caused by these packets causes a 95% packet loss for my pings between nodes. This goes so far that pacemaker thinks it's the only node in the cluster being alive.

Any suggestion, how to fix that?

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2014-11-13 10:31:15 -0500

Charles Benon gravatar image

updated 2014-11-20 04:51:08 -0500

It seems you have some issues on the virtual/physical diagram.

For what I saw with Mirantis 5.1, you will first to identify your management network:
Open vSwitch (OVS) GRE tunnels are provided through Management Network (so, I guess br-mgmt on the diagram).

After, I would make sure the bridge as its own layer 2 segmentation and all your Controllers and Computes have a common interface for this (I would ping between all of them to make sure it is the case).

Feel free to share a diagram with the bridges, physical ports between your hosts (help yourself with ip addr, ovs-vsctl show and /etc/neutron/plugins/ml2/ml2_conf.ini for the GRE tunnels).

edit flag offensive delete link more

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2014-11-06 04:22:26 -0500

Seen: 1,003 times

Last updated: Nov 20 '14