Description of problem: Nova has built-in default quotas of 51200MB of RAM and 20 cores per tenant. When installing RDO via TripleO, if the correct number of cores per baremetal machine and correct amount of RAM is used, the deployment fails. Tested with 12-core systems with 64GB of RAM. Version-Release number of selected component (if applicable): 2014.1.1-3.fc20 How reproducible: Every time, following these steps: http://openstack.redhat.com/Deploying_RDO_on_a_Baremetal_Environment_using_Instack Steps to Reproduce: 1. Run 'sudo yum -y install instack-undercloud' 2. Run 'sudo yum -y install instack-undercloud' 3. Prepare instack.answers file 4. Run 'instack-install-undercloud-packages' 5. Run 'command $(sudo cat /root/stackrc | xargs)' as user stack 6. Run 'instack-prepare-for-overcloud' 7. Deploy overcloud using Heat, Tuskar, or the Horizon web UI. Actual results: The deployment fails with an error reporting that the requested ram (65536) exceeds the available ram (51200) Expected results: Deployment succeeds Additional info: The workaround I used for this was to modify nova quotas: $ source stackrc $ keystone tenant-list copied id of tenant 'admin' $ nova quota-update --ram 500000 <admin id> $ nova quota-update --cores 120 <admin id> This should probably be set to reasonable limits for a baremetal installation for user admin in the 'instack-prepare-for-overcloud' script, or at a minimum the install guide (http://openstack.redhat.com/Deploying_RDO_on_a_Baremetal_Environment_using_Instack) should include instructions for increasing the quota. If possible, maybe the limits can be removed entirely by using -1 for the quotas, but I didn't test that.
James, this was previously assigned to openstack-nova. It sounds like an instack-undercloud issue to me, and seems like a reasonable idea. I've reassigned the bz; let me know if you think it should go somewhere else. Thanks!
cleaning up this old bz, we disabled nova quotas in instack-undercloud a while a go