Created attachment 739354 [details] The keystonerc_admin file Description of problem: If packstack fails to complete the first time, and is re-run, the admin password in keystonerc_admin is incorrect. Version-Release number of selected component (if applicable): 2013-1.1-0.5.dev538.el6.noarch How reproducible: Not sure Steps to Reproduce: 1. run packstack --allinone 2. interrupt it so the install fails 3. re-run packstack Actual results: the password in keystonerc_admin is wrong Expected results: the password in keystonerc_admin is right Additional info:
Created attachment 739355 [details] answerfile 1 this is the first answerfile
Created attachment 739356 [details] Answerfile 2 This is the second answerfile
Proposed fixes to puppet modules https://review.openstack.org/#/c/27171/ https://review.openstack.org/#/c/27169/ https://review.openstack.org/#/c/27168/ https://review.openstack.org/#/c/27117/ once all of these are merged I can pull them into packstack so randomly generated passwords/tokens are correclty picked up if changed between packstack runs.
Not to be pushy, but all of the above patches are merged now. I encountered this after having to rerun packstack several times due to fedora's mirrors being flakey earlier today.
This can happen when packstack stops before OpenStack client manifest is applied. It's always safest to run --allinone only first time and then use --answer-file=~/packstack-answers-<timestamp>.txt I guess that we will also have to implement workaround, so that packstack will look for the generated answer file and use that instead of generating new one. That way this issue will be fixed completely.
Martin: is this a feature we plan to implement? I would probably consider this a documentation bug ("running packstack multiple times without specifying the same answer file will result in a variety of problems so don't do that").
Yes I agree that this is also documentation bug. Supporting use case I mentioned in comment #5 would need some changes in Packstack's backend, which I'm not sure if it's worth now.
Documentation change submitted upstream: https://review.openstack.org/#/c/165454/