=========================== Error: Failed to launch instance "test": Please try again later [Error: No valid host was found. ].
Please look at the nova-api.log and nova-scheduler.log. The NoValidHost error indicates that the scheduler was unable to locate a compute node with enough available vCPUs, RAM, or HDD space to host an instance of the selected flavour. Ensure that the nova-compute service is active on the compute node and that the compute node has enough available resources to host the instance being launched. Note that the baremetal-* commands are only relevant to users of the baremetal driver. Please look at the nova-api.log and nova-scheduler.log. Check whether rabbitmq is running on controller. Appropriate password set for rabbitmq in nova.conf on controller and compute. Compute should be connecting to rabitmq on controller itself. You can set the flag scheduler_default_filters=AllHostsFilter in nova.conf. Restart the all service on controller and compute.
I have the same problem and I saw in my nova-scheduler.log that there are several odd messages:
Host has more disk space than database expected (96gb > 62gb)
No hosts matched due to not matching 'force_hosts' value of 'x217.boku.ac.at'
Filter RetryFilter returned 0 hosts
I realize this is an old post, but I came here via Google as I was having similar troubles trying to intentionally launch an instance on a specific hypervisor and getting the 'force_hosts' error. The problem I ran into was that the "host" name can be different than the "hypervisor" name. The nova boot command is looking for the "host" name and not the hypervisor name (or hypervisor id).
Asked: 2014-06-24 05:48:39 -0500
Seen: 18,663 times
Last updated: Jun 24 '14
check whether the compute node is functional and accessible from controller. cross check the host entries.