What is the purpose of the "network_size" configuration parameter?

In flatdhcp mode network_size should be number of ips in the entire fixed range. If you are doing a /12 then this number would be 2^20. That said, it will take a very long time for you to create your initial network, as an entry for each ip will be created in the database.

In vlan mode, network_size is the number of ips per project. The fixed_range space is divided up into subnets of network_size.

Vish

On May 17, 2011, at 3:41 PM, Davor Cubranic wrote:

> New question #157962 on OpenStack Compute (nova):
> https://answers.launchpad.net/nova/+question/157962
>
> I am installing a two-server OpenStack cluster (controller + compute node), using the scripted installation as described in the online Admin manual [1].
>
> In the section on "Project Network Configuration", I was asked for "Network range for ALL projects (normally x.x.x.x/12)". This, I specified as 172.16.0.0/12, and from a quick read of the installer's source, it appears that this will become the "fixed_range" parameter in nova.conf config file.
>
> Next, I was asked for "Total amount of usable IPs for ALL projects", which will become the "network_size" config parameter. The documentation [2] describes this parameter as "Number of IP addresses to use for VM guests across all projects."
>
> I do not understand the purpose of this parameter, or what a reasonable value would be. The reason I selected a /12 network range is to have lots of available addresses for my VMs. (2^20, +/- a handful of IPs reserved for subnet gateway, broadcast, etc.) So should I enter 1,000,000 for "network_size"? What are the consequences of entering such a large number, as opposed to, say, 100, or 1000?
>
> The documentation is not very helpful on this issue. Notice its description of the parameter just restates the question from the installer, without further explanation. To make things more confusing, the example nova.conf file given in [2] uses a /12 fixed_range, and 8 for the network size. Does this mean that there can be no more than 8 running VMs in the entire cluster?
>
> [1] http://docs.openstack.org/cactus/openstack-compute/admin/content/scripted-ubuntu-installation.html
> [2] http://docs.openstack.org/cactus/openstack-compute/admin/content/setting-flags-in-nova-conf-file.html
>
> --
> You received this question notification because you are a member of Nova
> Core, which is an answer contact for OpenStack Compute (nova).

Thank you. It would be great if this explanation could be part of the documentation.

Adding this to the documentation, thanks for the information.

Hi Anna,

network_size is still explained as "Number of IP addresses to use for VM guests across all projects." at [1]. But the complete flags reference [2] has a better explanation: "Number of addresses in each private subnet" -- I assume that's the one that you changed.

Could the two be brought in sync? And also, I'm not sure where is the right place for it, but a more extended explanation of the differences&consequences for Vlan vs FlatDhcp managers would be really useful, as given by Vish above.

[1] http://docs.openstack.org/cactus/openstack-compute/admin/content/setting-flags-in-nova-conf-file.html
[2] http://docs.openstack.org/cactus/openstack-compute/admin/content/reference-for-flags-in-nova-conf.html

Hi Davor - Thanks for pointing out the additional location to update the network_size flag.

While I'll be able to put this on my backlog, if you are so inclined, you could edit the documents yourself. They are located in the openstack-manuals project at https://code.launchpad.net/openstack-manuals. I have updated the doc/source/docbkx/openstack-compute-admin/computeinstall.xml file for reference [1], but you could also edit doc/source/docbkx/openstack-compute-admin/computenetworking.xml with additional information as you see fit.

Thanks for asking!
Anne