Savanna with Quantum Network

Unfortunately we are not test Savanna with Neutron now. Here is a bp for integration with Neutron - https://blueprints.launchpad.net/savanna/+spec/add-neutron-support

BTW you can try to use flag use-floating-ips to make savanna use only internal ips, theoretically if savanna will able to access instances using internal network than cluster should be created ok.

Dear Sergey,

Thanks for you reply. But in my case I have set "use_floating_ips=False" in my config file. And the interesting part is even when use_floating_ips is set to true i am getting the same error.

Any help please.

Thanks.

quantum does not support auto assignment of floating ip.
link: https://blueprints.launchpad.net/neutron/+spec/auto-associate-floating-ip

Savanna could work w/o floating ip if internal network is accessible from host where savanna installed.

I read code one more time and I think savanna will not work with neutron at all before this bp will be implemented.

Okie, but I dont want it to have floating IP, is it possible to patch the code, to work without neutron and to work only with internal IP's so that we need not wait till the bp to be implemented for people who want to try out savanna.

also use_floating_ips doesn't seem to have any effect i believe. So cant it be considered as a bug.

Thanks & Regards.

when I am launching cluster after some floating ip available, I can see the instace, but when the launching finished and the cluster turned error ,the instaces were deleted. Does that mean, if I use quantum, I can not run savanna? What can I do if I want to run savanna?
Thanks.

WangDong, in order to use Savanna you can change Quantum to Nova-network in your OpenStack environment.
Quantum (Neutron) support targeted for Savanna 0.3 which is tentatively expected to release in mid-October:
https://wiki.openstack.org/wiki/Savanna/Roadmap

Dear DongWang,

Kindly replace your network.py with the modifications i made after that restart the savanna-api.

The network.py will be at <savanna install dir>/lib/python2.7/site-packages/savanna/service/network.py

After replacing the file and restarting savanna api, launch cluster, all the vm's in your cluster will come up, and you can verify this from the instances page.

Once you find the instances are up, assign floating ip for all nodes.

If you are running savanna api with --debug flag, you will see some informative logs.

I have added some logs, for my understanding, and the code is not neat and clean. (i am not a programmer basically) :)

Kindly let me know if this works or not.

Thanks

This is my modifications to the actual code. Didnt find a method to share file here, so pasting the code.

http://paste.openstack.org/show/42740/

Regards

I also have this problem, and modify the source code in service/networks.py and instances.py

I have the same problem, and updated the network.py code per the comments above. However, when I launch the cluster it sits in the 'waiting' state indefinitely. The debug output keeps looping throught the servers it created for the clustuer. The following information is what keeps getting displayed. Looks like the management_ip is None, and not sure if that is the root of my problem.

2013-08-29 11:00:58.793 14371 WARNING savanna.service.networks [-] This is the server networks : {u'External': [u'10.10.10.95']}
2013-08-29 11:00:58.793 14371 WARNING savanna.service.networks [-] Length of IP List : 1
2013-08-29 11:00:58.794 14371 WARNING savanna.service.networks [-] This is the network_label External
2013-08-29 11:00:58.794 14371 WARNING savanna.service.networks [-] This is the management IP :None
2013-08-29 11:00:58.794 14371 WARNING savanna.service.networks [-] This is the IP :10.10.10.95
2013-08-29 11:00:59.877 14371 WARNING savanna.service.networks [-] This is the intance : <savanna.db.models.Instance[object at 383a4
d0] {created=datetime.datetime(2013, 8, 29, 17, 38, 11, 76763), updated=datetime.datetime(2013, 8, 29, 17, 38, 11, 76771), extra=Non
e, node_group_id=u'8841ab56-23a7-46d6-a65c-95222cb9102c', instance_id=u'c29b8ced-6382-4f89-8bf9-96a662275a37', instance_name=u'TestCluster-NameNode_JobTacker_Medium-001', internal_ip=u'10.10.10.98', management_ip=None, volumes=[]}>
2013-08-29 11:00:59.877 14371 WARNING savanna.service.networks [-] This is the server : <Server: TestCluster-NameNode_JobTack
er_Medium-001>

Work on basic support of neutron has been started:

https://review.openstack.org/#/c/44001/
https://review.openstack.org/#/c/44054/

Dear Bill,

The changes will work only with some conditions.

1. Each VM in cluster should have only 1 Network Interface attached and a floating IP.
2. There should be exactly 1 floating IP for all the instances spun up in the cluster.

Based on your description the management IP is null, which means you didnt apply floating IP to instances. Try to add floating IP for each node, and monitor the logs. You can see the floating IP coming up as the management IP and after that the cluster configuration will start and succeed.

I was able to edit the network.py file to allow the internal IPs to be used. This corrected my issue. Change was to:

...
  cluster_node_opts = [
      cfg.BoolOpt('use_floating_ips',
+ default=False, # True - default setting; uses floating IP. False allows for external/internal IPs to be used instead
...

Thx Thej, our current infrastructure won't allow for floating IPs. The change above will get us working until the Neutron bp is completed.