[timeu]

We (scientific research institute) are currently operating a relatively small commercially supported Openstack (Liberty release) installation next to our HPC infrastructure and so far it works quite well. Of course, because we are using Openstack from a commercial vendor, the installation/provisioning is quite straightforward (although there is quite a bit of complexity, if you want to do a lot of adjustments). So far we didn't have any big troubles during operations and our users (researchers) have started to use it (creating VMs for specific applications).

We will hopefully upgrade soon to the Newton release which will enable us to provide SaaS (Murano) and CaaS (Magnum) to the users. We will probably also evaluate Openshift on top of Openstack.

The biggest pain points is the integration with our existing HPC infrastructure (specifically the parallel filesystem/storage). We haven't really looked much at the code except of debugging some authentication issues, when we tried to integrate keystone with our core IT's AD, so we can't comment on that.

Additionally, the official openstack documentation (the commercially supported one is quite superficial) is often very confusing and it's very hard to find proper CHANGELOGS for each of the openstack components. Also we really don't like (or rather despise) openstack's issue tracker (Launchpad). It is very hard to find any existing issues, when we encounter some problems.