Red Hat Bugzilla – Bug 997141
Multinode packstack instance won't launch insteance unless all in one run first
Last modified: 2013-11-15 06:53:43 EST
Description of problem:
When you don't first run an all-in-one installation (instead skipping directly to a multi-node installation), launching an instance will fail on "no valid host found" (this was encountered when using neutron, but I haven't had a chance to test if this occurs when not using neutron)
Steps to Reproduce:
1. Install a fresh copy of RHEL 6.4
2. Install Packstack and the OpenStack repos
3. `yum -y update` and reboot to install the new kernel
4. `packstack --gen-answer-file`, and edit the answer file to point COMPUTE_HOSTS to a different node
5. `packstack --answer-file=answers.txt` to run packstack
6. Launch an instance
Instance fails to launch, giving error 'noValidHost'
Instance launches fine
P.S. After a bit of digging, I would guess that this has something to do with the default Packstack Neutron installation type being 'local'. Perhaps Packstack should warn if you have set the Neutron type to be 'local' but have a multi-node configuration?
Having separate compute hosts should work via Packstack installation. Do you still see the problem?
It's been a month with needinfo, so I'm closing it now.
@sross please, reopen it if you can still reproduce the problem.