Community pursues tighter Kubernetes integration in OpenStack Stein
16 April 2019 | 0
The latest release of open source infrastructure platform OpenStack, called ‘Stein’, was released today with updates to container functionality, edge computing and networking upgrades, as well as improved bare metal provisioning and tighter integration with popular container orchestration platform Kubernetes, led by super-user science facility CERN.
It also marks roughly a year since the OpenStack Foundation pivoted towards creating a more all-encompassing brand that covers under-the-bonnet open source in general, with a new umbrella organisation called the Open Infrastructure Foundation. OpenStack itself had more than 65,000 code commits in 2018, with an average of 155 per day during the Stein cycle.
Magnum, the certified Kubernetes installer, can now launch clusters significantly quicker than previous iterations, down from roughly 10-12 minutes per node to five minutes. Integration has been improved due to increased collaboration between the two open source heavyweight platforms, both from the OpenStack development side and the Kubernetes side, such as with the ‘OpenStack Cloud Provider’ service found within Kubernetes.
“That’s happened over the last year through involvement upstream in the Kubernetes community,” said Jonathan Bryce executive director, OpenStack Foundation. “On the Magnum side, it’s taking that work and also parallelising the way the provisioning actions are actually run — so that instead of saying you want a five-node Kubernetes cluster, you want a 50-node Kubernetes cluster, and it can basically spin up those machines in parallel and bring your environment up much more quickly.”
CERN, which has long been running an OpenStack environment at massive scale, had become a top contributor to Magnum, including leading the project. According to Bryce this user-led involvement means that the tighter integrations are in line with the needs of major users of the service.
Another user-led improvement for the Stein release is the clustering service Senlin, which has been driven by gaming giant Blizzard. APIs in this service now issue synchronous failures in case of problems with clusters or nodes, cool-downs, or conflicts at the service level.
“They’ve had the project lead on that project for a couple of cycles and obviously have been a big contributor to Senlin over the years too,” says Bryce. “They had an environment with multiple data centres and multiple OpenStack clouds and they wanted to have the ability to spread out and automate their operations across those.”
Improvements have also been included to the resource reservation service Blazar, with a new resource allocation API that provides operators with visibility into their reserved cloud resources.
And a new project, Placement, which originates from the Nova compute service, promises to simplify specifying a host for workload migration. The project, according to the foundation, increases API performance “by 50%” for common scheduling operations.
Identity service Kolla has been refined for better multi-factor authentication receipts, while Kolla has added support for full and incremental backups of MariaDB.
In a press statement, CEO of OpenStack cloud business Vexxhost, Mohammed Naser, said that all these improvements together speak to the open infrastructure platform being much simpler to deploy and run, especially from the operator perspective.
Naser said: “The level of stability has significantly increased… This has been really helpful because as we try to continue to run the latest release as it comes out, obviously one of the bigger concerns is making sure we don’t break our customers that are relying on those APIs.
“I think that what’s happening is that the operator community has been much more involved. The indirect result of that, is there isn’t this weird gap between the developers and the operators, and so a lot of the additions being made are very conscious, to make sure that we don’t break anything down the line.”
Bare metal improvements
Bryce draws attention to the improvements in the Ironic bare metal, which have been incremental over the last few releases and is now getting to a state where users can manage physical servers in the same way they have managed virtual servers, he says. In other words, they’re much faster and more user friendly than might be expected for physical server management, with more automation and scalability capabilities too.
“Physical servers aren’t the right answer for every workload, and virtual servers aren’t the right answer for every workload, so for OpenStack to be able to provide an elegant way to provision and manage both types of resources is a super valuable feature,” he says.
The “Open Infrastructure Summit” takes place this April in Denver, Colorado.
Commenting, Thierry Carrez, VP of engineering at OpenStack says that the Open Infrastructure branding is already encouraging a better environment of collaboration with other users and projects outside of its umbrella.
“We have run a few Open Infra days already, the new format for what used to be called OpenStack days, and as a result we’ve seen a lot of involvement from adjacent communities that have joined the discussion,” he says.
IDG News Service