[Openstack-operators] Metadata proxy failures

J-P Methot jpmethot at gtcomm.net
Mon Aug 17 15:43:26 UTC 2015


I have an HA openstack setup with haproxy and neutron networking.
Sometimes, when we need to work on the haproxy nodes and need to
shutdown the haproxy (we are still testing things), metadata breaks.
Actually metadata breaks a lot, easily and quite often. It's probably
the most fragile openstack service in our setup.

Anyhow, after much searching, I believe the issue is that the proxy
service listening inside the router namespace on the network just dies.
Because of that, if I do the following command :

ip netns exec qrouter-7dcd02ba-bb54-4f4e-9792-6b2f0bafe3e0 netstat -antp

I get an empty output. If I then reboot the network node and do the same
command, I get the following :

Active Internet connections (servers and established)
Proto Recv-Q Send-Q Local Address           Foreign Address
State       PID/Program name
tcp        0      0  *
LISTEN      18401/python

So, my question is, what exactly controls this proxy? More importantly,
if it's neutron-metadata-agent, then how come it fails silently? Not
only is it not logged, but restarting that service does not fix the
issue until I reboot the net node, which leads me to believe it doesn't
start properly because of another service.

Jean-Philippe Méthot
Administrateur système / System administrator
GloboTech Communications
Phone: 1-514-907-0050
Toll Free: 1-(888)-GTCOMM1
Fax: 1-(514)-907-0750
jpmethot at gtcomm.net

More information about the OpenStack-operators mailing list