4.3 years ago by
United States
I should let someone actively working on CloudMan respond instead - but since I did spend a lot of time on this problem a while ago and I definitely shared your frustration - I will give you my entirely dated and unofficial answer.
Is it REALLY possible? It is certainly possible - I believe several institutions have gotten CloudMan up and running on OpenStack. For instance, https://usegalaxyp.org/ runs on an OpenStack instance at the Minnesota Supercomputing Institute (or at least it did when I worked there last fall). But that said - there wasn't much documentation available when I started that and the effort took a couple months and I landed up doing a lot of novel development to get it working. On the other hand - I did try to document things as a result and I had a pretty particular set of requirements (like not having volumes/cider available) that made it more difficult.
So possible? Yes. I still think it is an uphill battle though, if you stick with it I would expect to spend a lot of time looking at source code and figuring out processes that work for you. The lack of documentation is frustrating - but it stems in part from the fact that the process is constantly evolving even for AWS deployments.
More specifically, this work - https://github.com/jmchilton/cloudman_openstack_bootstrap - was largely super-ceded by my work on a CloudBioLinux Deployer (https://github.com/chapmanb/cloudbiolinux/blob/master/deploy/cloudman.md). I was definitely using the deployer to deploy fresh instances of CloudMan to OpenStack on a weekly basis for a while - but they were pretty specialized not the generic instructions for AWS found in https://github.com/chapmanb/cloudbiolinux/blob/master/deploy/cloudman.md. To compound that lack of documentation, CloudMan and CloudBioLinux have evolved a lot since then so even what remains probably doesn't work.
You mentioned no documentation for userdata-default.yaml - that can be found in part here https://wiki.galaxyproject.org/CloudMan/UserData.
