Enis, John and all; I spotted ansible-cloudman on GitHub today which reminded me I've been meaning to write about the approach we setup late last year to run bcbio on AWS. It uses elasticluster (https://github.com/gc3-uzh-ch/elasticluster) which has the advantage of being all Ansible scripts and bootstrapping from standard images -- so no more making AMIs. It also uses SLURM instead of SGE, which is a nice change.
We wrote an interface that automates all of the stuff you need to setup on AWS: IAM users, VPCs and what not. It is a pretty streamlined process from the command line, including specifying the cluster size and stopping/starting it:
https://bcbio-nextgen.readthedocs.org/en/latest/contents/cloud.html#aws-setu...
I also wrote up some benchmarking work using to give an idea of using it in practice:
http://bcb.io/2014/12/19/awsbench/
All of the code is here:
https://github.com/chapmanb/bcbio-nextgen-vm
As always, happy to overlap/share with whatever y'all decide to do. We could make bcbio specific stuff optional as needed, although it is pretty lightweight -- just the driver scripts and a Docker image of bcbio. It's basically a ready to use cluster with this little extra added so hopefully could be useful for future plans with CloudMan.
Hope this is useful, Brad
galaxy-dev@lists.galaxyproject.org