[Openstack] PLEASE HELP multi_host not working

Mark Russell mark.russell at canonical.com
Wed Jul 4 13:53:35 UTC 2012


Hi Marnus,

I experienced the exact same symptom (no fixed IP given to instance).
Is it possible that you previously deleted and recreated your network
with networks that overlap or match the old network?  In my case the
problem was related to this bug:

https://bugs.launchpad.net/nova/+bug/754900

In this bug, instances are getting fixed IPs but from a previously
deleted pool.

Another bug (marked as a dupe of the above) describes this situation a
little better (no IP given to instance):

https://bugs.launchpad.net/nova/+bug/812573

One way to confirm if deleting and recreating the network is the problem
is to tail -f /var/log/nova/nova-network.log while starting and stopping
an instance on that node.  When I was seeing this problem, nothing would
show up in nova-network.log during launch, but when I terminated the
instance, nova-network.log would show an error saying essentially that
"network 2 could not be found".  However we were on network #5 by then.
 So clearly something was off in nova db.

There is likely a better way to fix this than what I did (drop and
recreate nova db) but it was the simplest fix given that this was a
brand new installation.  There is almost certainly a "better" way to fix
it by manually hacking nova db.

Hope this helps,
Mark


On 07/03/2012 11:22 PM, Marnus van Niekerk wrote:
> I installed a minimal cloud image (cirros) and found that it had no
> network configures after booting.  Even running udhcpc manually does not
> work - it never gets an answer from dhcp.
> 
> I checked and dnsmasq is running on all of the the compute nodes:
> /usr/sbin/dnsmasq --strict-order --bind-interfaces --conf-file=
> --domain=novalocal --pid-file=/var/lib/nova/networks/nova-br100.pid
> --listen-address=10.10.11.135 --except-interface=lo
> --dhcp-range=10.10.11.130,static,120s --dhcp-lease-max=64
> --dhcp-hostsfile=/var/lib/nova/networks/nova-br100.conf
> --dhcp-script=/usr/bin/nova-dhcpbridge --leasefile-ro
> 
> Why would the DHCP not be working?  I am running out of ideas here. 
> Have tried everything I can read from searching google and still
> multi_host does not come close to working.  I am now even considering
> abandoning OpenStack and going back to Eucalyptus.  I have even tried
> changing to FlatNetwork with flat_injected=True and then IPs are set in
> the VMs but the metadata service still fails (no route to host).
> 
> I have checked and each host has the 169.254.269.254 addresses natted to
> itself.
> I have also deleted and recreated the network with the --multi_host=T flag.
> 
> I have this on the "controller":
> #--routing_source_ip=10.10.20.114
> --ec2_dmz_host=10.10.20.114
> --multi_host=True
> --enabled_apis=ec2,osapi_compute,osapi_volume,metadata
> --dmz_cidr=10.10.20.114/32
> 
> And this on the compute nodes:
> #--routing_source_ip=10.10.20.114
> --ec2_dmz_host=10.10.20.114
> --multi_host=True
> --enabled_apis=metadata
> --dmz_cidr=10.10.20.114/32
> 
> 
> On 03/07/2012 10:03, Marnus van Niekerk wrote:
>> On 03/07/2012 09:53, Marnus van Niekerk wrote:
>>> I can now see the bridge created and assigned an address on each
>>> compute node, but all of the VMs get stuck after the bootloader -
>>> they never boot any further.
>>
>> Sorry, they do actually boot after a while but without any networking..
>>
>> cloud-init-nonet waiting 120 seconds for a network device.
>> cloud-init-nonet gave up waiting for a network device.
>> ci-info: lo    : 1 127.0.0.1       255.0.0.0       .
>> ci-info: eth0  : 1 .               . fa:16:3e:39:5f:02
>> route_info failed
>> Waiting for network configuration...
>> Waiting up to 60 more seconds for network configuration...
>> Booting system without full network configuration...
>>
>>
>>
>>
> 
> 
> 
> 
> 
> _______________________________________________
> Mailing list: https://launchpad.net/~openstack
> Post to     : openstack at lists.launchpad.net
> Unsubscribe : https://launchpad.net/~openstack
> More help   : https://help.launchpad.net/ListHelp
> 


-- 
Mark Russell
Premium Service Architect | Canonical, Ltd.
<mark.russell at canonical.com> | GPG: 4096R/B3BBA7D1

www.ubuntu.com | www.canonical.com






More information about the Openstack mailing list