nova-compute on cloud controller??

No, you don't need to install nova-compute on the controller node in such a setup.

Ok. Thanks Max.

I'll try this configuration.

Regards!

Hi Max:

I've launched one instance in the following scenario:

    - cloud controller (all services except nova-compute)
    - nova-compute 1 (only nova-compute service)
    - nova-compute 2 (only nova-compute service)
    - nova-compute 3 (only nova-compute service)

I asked if nova-compute was necessary on cloud controller because of my instance is not deployed correctly when nova-compute service is not installer on cloud controller. The instance log is:

cloud-init start running: Thu, 24 May 2012 18:16:21 +0000. up 9.85 seconds

2012-05-24 18:17:47,705 - util.py[WARNING]: 'http://169.254.169.254/2009-04-04/meta-data/instance-id' failed [6/120s]: url error [[Errno 113] No route to host]

2012-05-24 18:17:50,704 - util.py[WARNING]: 'http://169.254.169.254/2009-04-04/meta-data/instance-id' failed [9/120s]: url error [[Errno 113] No route to host]

2012-05-24 18:17:53,704 - util.py[WARNING]: 'http://169.254.169.254/2009-04-04/meta-data/instance-id' failed [12/120s]: url error [[Errno 113] No route to host]

2012-05-24 18:19:44,782 - DataSourceEc2.py[CRITICAL]: giving up on md after 123 seconds

no instance data found in start

Skipping profile in /etc/apparmor.d/disable: usr.sbin.rsyslogd

 * Starting AppArmor profiles [80G
[74G[ OK ]

landscape-client is not configured, please run landscape-config.

acpid: starting up with proc fs

 * Stopping System V initialisation compatibility[74G[ OK ]

acpid: 1 rule loaded

acpid: waiting for events: event logging is off

 * Starting System V runlevel compatibility[74G[ OK ]

 * Starting ACPI daemon[74G[ OK ]

 * Starting save kernel messages[74G[ OK ]

 * Starting automatic crash report generation[74G[ OK ]

 * Starting regular background program processing daemon[74G[ OK ]

 * Starting deferred execution scheduler[74G[ OK ]

 * Starting CPU interrupts balancing daemon[74G[ OK ]

 * Stopping save kernel messages[74G[ OK ]

 * Starting crash report submission daemon[74G[ OK ]

 * Stopping System V runlevel compatibility[74G[ OK ]

Hi Juan,

first things first: I've got the same setup up and running, which is why I said that you don't need nova-compute on the cloud controller.

I suppose you checked "nova-manage service list" and you have smileys on your compute services?

There was an issue when images need to fetch meta data, so you might have a look here: https://answers.launchpad.net/nova/+question/159317

Hi Max,

Yes, of course. All services have smiles.
What kind of instances have you launched? uec-images.ubuntu.com or customized ones?

Anyway, I'll try to launch a customized instance and I'll report about anything.

Thanks!

PD: I had already issued 'iptables -t nat -A PREROUTING -d 169.254.169.254/32 -p tcp -m tcp --dport 80 -j DNAT --to-destination $CONTROLLER:8773' in all of compute nodes.

Actually I've got both types running, uec-images and customized CentOs images.

Ok.

So, did you add any iptables rule to fetch instance meta-data? or you are using multi-host option (nova-network in each node)?

I'm not using the multi-host option for nova-network. The controller node is the only one having nova-network running.

Though I had really a hard time getting the instances to be able to connect to the outside world, incoming with floating ips and outgoing from the instances themselves, I never ran into trouble with the meta-data. But I worked most of the time with the custom built images and added the use-images later on.

Hi again:

How many network interfaces do you have in each node?
could you paste your nova.conf file, please?

I don't know what can be wrong...

Thanks Max.

I only have eth0, but it's trunked so i've got 3 eth0.X and a bridge for each of it.

--root_helper=sudo nova-rootwrap
--auth_strategy=keystone
--allow_admin_api=true
############
# nova.flags
--api_paste_config=/etc/nova/api-paste.ini
--connection_type=libvirt
--ec2_host=10.0.2.21
--glance_api_servers=10.0.2.21:9292
--image_service=nova.image.glance.GlanceImageService
--lock_path=/var/lock/nova
#--log-config=<None>
--logdir=/var/log/nova
--my_ip=10.0.2.21
--rabbit_host=10.0.2.21
--rabbit_password=guest
--resume_guests_state_on_host_boot=true
--s3_host=10.0.2.21
--start_guests_on_host_boot=true
--state_path=/var/lib/nova
--zombie_instance_updated_at_window=172800
############
# nova.log
# publish_errors=True
############
# nova.network.linux_net
--dhcpbridge_flagfile=/etc/nova/nova.conf
--dhcpbridge=/usr/bin/nova-dhcpbridge
--routing_source_ip=10.0.2.21
--network_manager=nova.network.manager.FlatDHCPManager
--network_host=10.0.2.21
--flat_network_dhcp_start=192.168.2.2
--flat_network_bridge=br232
--flat_interface=eth0.232
--flat_injected=false
--force_dhcp_release
--public_interface=eth0.235
--floating_range=10.0.5.128/25
--fixed_range=192.168.2.0/24
--network_size=255
############
# nova.service
# osapi_compute_listen="0.0.0.0"
# osapi_compute_listen_port=8774
# osapi_volume_listen="0.0.0.0"
# osapi_volume_listen_port=8776
############
# nova.test
--use_deprecated_auth=false
############
# nova.api.ec2
--keystone_ec2_url=http://10.0.2.21:5000/v2.0/ec2tokens
############
# nova.vnc
--vnc_enabled=true
--vncserver_proxyclient_address=10.0.2.22
--novncproxy_base_url=http://10.0.2.21:6080/vnc_auto.html
--vncserver_listen=10.0.2.22
###########
# nova.virt.libvirt.connection
--libvirt_type=kvm
--libvirt_use_virtio_for_bridges=true
###########
# nova.scheduler.manager
--scheduler_driver=nova.scheduler.simple.SimpleScheduler
--sql_connection=mysql://nova:nova@10.0.2.21/nova

Great Max!

Finally my insfrastruce is working. I got to deploy multi_host scenario (nova-network in all compute nodes).
I had serveral problems with 'fixed_ips' and 'floating_ips' tables. There were some bugs associated with this tables (https://lists.launchpad.net/openstack/msg10269.html)

I'd like to thank you for your attention and accurate answers.

Regards,
Juan F.