The Big Picture

The 2025.2 "Flamingo" cycle crossed its midpoint this week, and the most time-sensitive item on the list had nothing to do with code: a hard July 19 deadline to renew OpenInfra individual memberships, after which the 180-day eligibility rule locks out anyone hoping to vote in the 2026 board election. Beyond that, the week was a balanced mix of governance signposting (PTL/TC nomination season approaching, a looming PBR/setuptools bug) and a crop of practical Nova, Neutron, and Cinder questions that point at real gaps in default behavior — most notably an OVN QoS migration headache and a Neutron default that's now exposing external subnets to non-admin users.

Heads Up / Action Needed

• Renew your OpenInfra individual membership by July 19 (Wes Wilson's "Action Required"). Following the integration with the Linux Foundation, all members must re-confirm. The 180-day eligibility requirement is firm and will not be extended, so missing it means no vote in the 2026 Individual Governing Board election. The process is free and takes a few minutes.
• PTL/TC nominations for 2026.1 open August 6 — start thinking now if you intend to lead a project or join the TC.
• The OpenInfra Summit CFP (project updates and Forum sessions, Paris-Saclay, 17–19 October 2025) was closing July 8.
• A critical, impending PBR bug driven by upstream setuptools changes was flagged by the TC — worth watching if you maintain release tooling or packaging.
• The OpenStack user survey is open until August — the TC is asking for broader participation.

Releases & Announcements

Goutham Pacha Ravi's TC weekly summary (R-12) confirmed the community is officially in the second half of the Flamingo cycle. The TC reported a smooth rollout of Developer Certificate of Origin (DCO) enforcement with minimal issues, progress on Cyborg's potential revitalization through new and returning maintainers, and continued attention to inactive projects and the Eventlet deprecation. Cyril Roelandt opened a useful cross-project thread on deliverables defined in governance but not yet released for Cinder, Glance, Telemetry, and Barbican — the kind of mid-cycle audit that catches release gaps before they become end-of-cycle scrambles.

Development & Technical Decisions

Two Neutron threads stand out as default-behavior issues operators should know about. First, Maximilian Stinsky-Damke flagged that since Neutron 2024.2 (25.0.0) the subnet-external-network extension is loaded by default, making subnets of external networks visible to non-admin users. Those subnets are typically unusable for spawning VMs or load balancers and just clutter the Networks and Octavia dashboards — and there appears to be no straightforward way to disable the extension or revert to the old behavior. If you've upgraded to 2024.2 or later and noticed dashboard clutter, this is why.

Second, Simon Stephan described a thorny OVN QoS migration problem: moving from libvirt-based per-flavor bandwidth limits to OVN, where that mechanism no longer works, and finding that Neutron's native QoS has no system-wide or per-domain default and no built-in way to auto-attach a default policy to newly created projects. His proposed workaround — listening for project.create events on RabbitMQ and using openstacksdk to assign a shared default policy — is a reasonable pattern, but the thread is really a feature-gap report worth the Neutron team's attention.

On the Nova side, Masahito Muroi raised a flavor-visibility approval workflow limitation: the is_public flag controls whether a flavor is visible but can't gate use, creating a chicken-and-egg problem for "request approval before using a custom flavor" use cases, since users can't see flavors they haven't been approved for. Separately, Jiatong Shen asked about enabling compression for live migration of memory-heavy, high-dirty-rate workloads (databases) that struggle to converge over 10 GbE without resorting to force-complete.

Storage drew a substantive Cinder thread from Yuta Kambe (Fujitsu) on retype between Ceph backends causing zero-filling, which wastes time and space and forces admins to run rbd sparsify repeatedly — he's asking whether the community is considering improvements to the retype implementation rather than treating sparsify as the permanent answer.

On hardware enablement, the Redfish VirtualMedia path discussion matured: HPE Superdome exposes VirtualMedia under /Systems/{id}/VirtualMedia/ rather than the older /Managers/{id}/VirtualMedia/, and the standard has moved that way — sushy/Ironic already support both paths in recent versions, so this is largely resolved for those on current code.

Oleg Bondarev posted the Neutron bug deputy report for the week of June 30. Skyline saw two items: a request to extend the core reviewers team to speed up slow reviews, and a config quirk where Skyline appends /v3 to the keystone URL in the generated RC file when deployed via OpenStack-Helm (workaround: deploy keystone without /v3).

Community & Events

Ian Choi reported that the Zanata translation job failures triggered by the Python 3.8 deprecation are resolved (Java 8 turns out to be available on Ubuntu Noble), and that the broader Weblate migration is being accelerated through the i18n mentoring program. The Neutron drivers meeting on July 11 was canceled for lack of agenda, with a standing reminder that open neutron-specs still need review.

Operator Troubleshooting

The recurring Kolla-Ansible Ceilometer "unhealthy" container problem surfaced again (this week in German, from Mai-Anh), with the operator wondering whether deprecated or unmaintained Ceilometer components — the API or certain agents — are the cause. It's a question the community keeps fielding, and a sign that the state of Ceilometer in current Kolla images deserves a clearer documented answer. Jon Bernard also opened a Cinder thread about volumes not rescanning during VM attach, one to watch for anyone hitting attach-time path discovery issues.