Description of problem: After following steps for Fedora test day on https://fedoraproject.org/wiki/Test_Day:2012-09-18_OpenStack nova boot was failing with: 2012-09-18 12:54:59 WARNING nova.scheduler.manager [req-97f0d033-6a3e-4e89-8cc1-38d 2a5235d55 c695bad39d17436eb96c5f748ed2dd47 b670bf3efaee479194f284861a363b5d] Failed to schedule_run_instance: No valid host was found. 2012-09-18 12:54:59 WARNING nova.scheduler.manager [req-97f0d033-6a3e-4e89-8cc1-38d 2a5235d55 c695bad39d17436eb96c5f748ed2dd47 b670bf3efaee479194f284861a363b5d] [insta nce: 6fbac148-9fd2-469d-8c18-e0de4646faf7] Setting instance to ERROR state. Restarting the scheduler solved the problem. Version-Release number of selected component (if applicable): F-17 How reproducible: Steps to Reproduce: Follow steps on https://fedoraproject.org/wiki/Test_Day:2012-09-18_OpenStack until the 'launch an instance' step. Actual results: Expected results: Additional info:
Set version to 18 because this is the f18 version of openstack
Hmm, we need to get to the bottom of what's causing the scheduler to not consider the host valid ... and why that changes after a restart
After some more poking around - this seems to have beed a RAM issue in my VM (although needs to be confirmed). However logging should be more helpfull than this.
Ok, sounds like we need more details to figure out exactly what was going on here. If the only improvement we need to make is just to improve logging, that can be a bug we file directly upstream
I will try to reproduce and update the bug accordingly.
(In reply to comment #4) > Ok, sounds like we need more details to figure out exactly what was going on > here. > I tried to reproduce this - but cannot get this behavior with the latest packages (2012.2-1.fc18). The scheduler acts as expected and reports that there isn't enugh memory when trying to start a large enough instance to warrant it. > If the only improvement we need to make is just to improve logging, that can > be a bug we file directly upstream Loging can be made more helpful with debug mode turned on, so we don't really need to report a bug upstream.