Introduced at the Tokyo Summit, Superuser TV offers community and industry insights, plus educational topics to support the OpenStack community. With content ranging from deployments to diversity, from emerging technologies to cloud strategy, Superuser TV is aiming to provide the community with access to a variety of perspectives and knowledge
60 percent of deployments in production and high rates of adoption of OpenStack’s core services are key findings from the report released by the User Committee. The full report can be downloaded here: openstack.org/user-survey
Why embracing this notion is the key to unlock an open data center infrastructure, according to Boris Renski, the co-founder and CMO of Mirantis.
Community feedback
OpenStack is always interested in feedback and community contributions, if you would like to see a new section in the OpenStack Weekly Community Newsletter or have ideas on how to present content please get in touch: community@openstack.org.
Reports from Previous Events
Deadlines and Contributors Notifications
Security Advisories and Notices
Tips ‘n Tricks
What you need to know from the developer’s list
- calebb: Shade now supports volume snapshots
- pleia2: Launched code search [1].
- sdague: grenade-multinode live upgrade tests now running on nova non voting
- AJaeger: Contributors guide is published [2].
- Tell us yours via IRC with a message “#success [insert success]”
- November 9th 1700UTC
- Requires a cluster restart, in which people won’t be able to do searches during that time.
- New features from upgrade:
- Aggregations
- Rolling upgrades within a major release
- Should improve performance
- IRC channel change from #openstack-relmgr-office to #openstack-release
- “Office hours” are being dropped.
- Just drop by the channel or on the dev list with subject containing “[release]” anytime you need something.
- Ironic tries to keep master backwards compatible. There are deployers doing continuous deployments of Ironic off of master.
- Based on the deprecation tag policy [3], it only covers released and tagged code, but not unreleased code or features introduced in an intermediate release.
- A proposal [4] by Jim:
- Three month deprecation period is needed for features that were never released.
- A feature that was introduced in a intermediate release needs to be deprecated in the next intermediate release or coordinated release, and supported until the next release and 3 months.
- There was a two part session at the summit [5]
- Previously, there was an unwritten policy that DLMs should be optional, which led to writing poor DLM-like things backed by databases.
- Continuing our existing pattern of usage like databases and message-queues, we’ll use an oslo abstraction layer: tooz [6].
- Current OpenStack project requirements that surfaced in discussion for DLMs, it’s likely that Consul, Etcd and Zoo Keeper will be fine to use via Tooz. No project required a fair locking implementation in the DLM.
- We want to avoid the situation of unmaintained drivers. We adopted a similar requirement from oslo.messaging driver requirements [7]:
- Two developers responsible for it
- Gating functional tests that use dsvm
- Test drivers in-tree need to be clearly referenced as a test driver in the module name.
- Davanum brings in Devstack ZooKeeper support [8].
- An etcd driver is in review for Tooz [9].
- A Consul driver in Tooz is also planned [10].
- Concerns raised about the default DLM driver being ZooKeeper:
- It’s a new platform for operators to understand
- We don’t know how well ZooKeeper will work with openjvm as oppose to Oracle’s JVM.
- Evolve the current cross-project meeting to an “as needed” rotation, held at any time on Tuesdays or Wednesdays.
- Based on feedback [11] it’s difficult to have meetings at different times on Tuesday and Wednesdays.
- There was consensus that the meeting can be “as needed” on Tuesday, and that most announcements will happen in the mailing list, and sometimes show up in this weekly Dev List Summary.
- Projects like Heat,Tempest, Rally, and other projects that work with resources are polling for updates on asynchronous operations.
- Boris proposes having API’s expose the ability to just get the status by UUID, instead of fetching all data on a resource.
- Clint suggests instead of optimizing for polling, we should revisit the proposal for a pub/sub model, so users can subscribe to updates for resources.
- Sean suggests near term work around is to actually use Searchlight, which today monitors the notification bus for Nova.
- Searchlight is hitting the Nova API more than ideal, but at least it’s one service.
- Longer term we need a dedicated event service in OpenStack. Everyone wants web sockets, but anticipating 10,000+ open web sockets, this isn’t just a bit of python code, but a highly optimized server underneath.