OpenStack’s Ocata release improves container integration and networking and High Availability are simplified
Here at OCF we have integrated a number of installations that include OpenStack. There have been some exciting developments recently that I would like to share with you.
OpenStack Ocata Release
OpenStack’s 15th release arrived after a shorter 4-month development cycle rather than the usual biannual release. This is to allow for a change in the way development happens on OpenStack and to allow a better feedback loop from Developer to Mid-Cycle ops to the Main Summit. A side-effect of this change is that major features are in short supply so Ocata is mainly a stability release. As part of my work engineering deployments for OCF, I was again able to contribute some small changes.
Despite the shorter development window, the continued push to offer containers as a service on OpenStack is finally starting to bear fruit. I evaluated OpenStack’s container support during the Mitaka release and found it wanting. We therefore provided a container requirement at the time by deploying an OpenShift cluster on top of OpenStack and I would still recommend this method. However, I continue to watch developments with interest. I feel that containers are a more natural fit running as a platform on top of infrastructure but will be delighted to be proved wrong!
In TripleO, a project aimed at installing, upgrading and operating OpenStack clouds using OpenStack’s own cloud facilities as the foundation, we finally see the arrival of scaling of the control plane and the ability to deploy more than three controllers for any one service. This is great news for larger deployments where load balancing becomes important.
In the block storage world, Cinder is finally able to run in Active/Active mode which greatly improves the resiliency of the final deployment.
Finally and although we have enjoyed the progress of TripleO and the improvements it offers with each release, we note that the OpenStack-Ansible project has gained support for CentOS 7 as well as Ceph. Ansible is an increasingly popular configuration management tool and I am excited about the potential of a unifying deployment and configuration tool and expect to see further integration in future TripleO releases as the SSH-based utility is a natural fit here.
Red Hat OpenStack Platform 10
We recently saw the release of Red Hat’s OSP 10 product and were able to rapidly upgrade projects underway to take advantage of some great new features. From a deployment point of view, we saw the arrival of composable roles in TripleO which allows us to be more flexible with the configuration of the control plane and compute environments. It also allows for automated deployment to mixed hardware, allowing easier deployment of heterogeneous hardware into an OpenStack environment.
We were also able to leverage new NFV capabilities to automatically deploy nodes with InfiniBand directly into the virtual machine. This also allows PCI pass-through of nodes with GPU capability.
Another advantage is the simplification of High Availability (HA) in the control plane which has long been overly complex in OpenStack. Core services are now managed by Pacemaker whereas the remaining services are simply load-balanced using HA Proxy.
Another simplification in an area which has long been overly complex is OpenStack networking. With RHEL 7.3, OpenvSwitch can use the conntrack backend and a much simpler – and therefore more efficient – path for packets to transit across the infrastructure.
I am particularly excited about Manila and FileShare-As-A-Service. It seems a logical step to allow tenants to collaborate and easily share research results and data with each other within the cloud and I look forward to hearing from users who are currently starting to take advantage of this in our deployments.
Red Hat has changed their support matrix from this release to something more akin to Ubuntu’s Long Term Support. OSP 10 will be supported for three years but OSP 11 and 12 will only be supported for one year. In theory, this should allow users to choose between features and stability and makes sense to me. Red Hat has also introduced enhanced support for OpenStack which will be of interest to customers migrating mission-critical applications to their private cloud.
OCF recently became Red Hat’s only UK Premier Cloud Partner and have a number of years’ experience in Cloud technology. We are seeing interest in other cloud products such as CloudForms, allowing single pane of glass management for containers, private and public cloud. Combined with our experience in HPC, 2017 is shaping up to be a very interesting year for High Performance Clouds. Recent high-profile failures in public cloud demonstrate that these providers are not infallible and the costs both financial and technical associated with vendor lock-in should not be ignored.
What do you think of the latest developments? Please feel free to leave a comment below.