Expanding Openstack

asked 2013-12-16 21:58:31 -0600

Leonard gravatar image

Hello all, Is my understanding of Openstack correct in that you can setup on one node initially and then add more nodes one by one to create a central storage pool?

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2013-12-16 22:51:25 -0600

larsks gravatar image

It's hard to answer your question because it's not clear exactly what you're asking. Let me know if the following helps:

You can certainly start with a single node installation and then grow that to cover additional nodes. The thing that complicates your question is that each service can be expanded separately: for example, you can start with a single node installation, and then add a bunch of compute nodes. Images will continue to be hosted by your single glance instance on your original node, and if you're using cinder, volumes will continue to be allocated from your single original node, too...but instances will be scheduled and booted on your new compute nodes (and will use local storage on those nodes for hosting the disk images from which they are booting).

Similarly, instead of adding compute nodes, you could add a bunch of nodes to support cinder. These may all have independent storage.

You may also elect to implement a shared storage pool using glusterfs, sheepdog, or some other shared storage mechanism, and use this to back the instances or the volume service.

You could do both of the above, scaling our your block storage and compute: but you may also want to scale out your image service. And your network service. And possibly your API services as you start having more traffic. And then you're going to want to look into things like HA and load balancing.

The point here is that most of the services that comprise OpenStack are designed to scale horizontally. There is some information about scaling OpenStack on the openstack wiki.

edit flag offensive delete link more


Thanks for your reply. I think what I'm asking is it possible for to add more "compute" nodes so that RAM, CPU, network and disk are all shared. I believe I saw in one tutorial that it was as simple as running one command on the new node. Would VPSes automatically have HA across these compute nodes?

Leonard gravatar imageLeonard ( 2013-12-16 23:02:38 -0600 )edit

It sounds as if you're still conflating several things. You can absolutely add additional compute nodes to provide additional RAM and CPU resources to your OpenStack deployment (although typically that is more than a single command). You can separately add additional disk and network resources, too.

larsks gravatar imagelarsks ( 2013-12-18 14:36:30 -0600 )edit

Regarding HA, that's not something you get "out of the box". Much like AWS, if an instance dies, it's not going to magically come back. There are load balancing and autoscaling services that can help (through the Neutron and Heat projects, respectively).

larsks gravatar imagelarsks ( 2013-12-18 14:38:33 -0600 )edit

Thanks again. I'm just worried about the expansion ability of a single node. I know the recommended setup is two or three servers.If I start with a single server running all the services, can I then convert it into a cloud controller node?

Leonard gravatar imageLeonard ( 2013-12-19 11:02:28 -0600 )edit

Could you also please advise me on this https://ask.openstack.org/en/question/9154/will-a-weak-cloud-controller-node-slow-down-the-compute-node/

Leonard gravatar imageLeonard ( 2013-12-19 11:07:16 -0600 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2013-12-16 21:58:31 -0600

Seen: 306 times

Last updated: Dec 16 '13