Open Stack

Ken D'Ambrosio <ken at ...> writes:

> 
> So, I've got an Icehouse cloud that had been working fine; then we had 
> to bring everything down for an extended power outage, and stuff still 
> isn't quite right.  The nutshell problem is that only one of my two 
> neutron gateways seems to be able to route to the metadata server.  The 
> "working" neutron gateway, however, has switched after a reboot.
> 
> Symptoms:
> * Initially, I set eth0 to point to my first neutron gateway 
> (hereinafter "NG")
> * It failed to be able to access the metadata server (verified by 
> telnetting to 169.254.169.254:80)
> * Associating eth0 with the secondary NG worked fine
> * All this held true through multiple VM reboots, and a cloud controller 
> reboot
> Thanks,
> 
> -Ken
> 
> 

Hi!
I might have a tip. Yesterday, I ran into an issue where metadata was
unavailable on one compute node for one tenant in a DVR setup. I checked and
neutron-ns-metadata-proxy was not running in the qr namespace of the
tenant's router. The fix was to kill it and remove the pid file from
/var/lib/neutron/external/pids/. Then restarting the neutron-l3-agent
restarts the metadata proxies.
I'm running the configuration with
/etc/nova/nova.conf:service_metadata_proxy=True
Tomas