▤ Stackers Network

January 31, 2026

Stackers Network Digest — January 31, 2026

27 threads · 53 messages · openstack-discuss

The Big Picture

Election season opened in earnest this week. With Gazpacho at R-9 (nine weeks to coordinated release), the Technical Committee set the stage for the combined 2026.2 PTL and TC elections and began resetting leadership models for several at-risk teams under the Distributed Project Leadership (DPL) model. But the standout technical conversation was strategic rather than release-bound: a serious community discussion about building an attestation service for confidential VMs, which several contributors framed as the missing keystone for OpenStack's confidential-computing story. Alongside it, Sean Mooney's effort to revitalize Cyborg showed the community's recurring pattern of one or two contributors stepping up to rescue a neglected-but-important project.

Development & Technical Decisions

Attestation is the missing piece for confidential computing. Jean-Philippe Jung opened the week's most-discussed thread (9 messages, 5 participants) asking whether any existing OpenStack project covers attestation — the cryptographic proof that a workload is genuinely running inside a real Trusted Execution Environment (AMD SEV-SNP, Intel TDX, Arm CCA). His argument: without attestation, confidential computing is meaningless because you're trusting the provider's word rather than silicon-anchored evidence. The attestation service would gate VM decryption-key release after verifying that the hardware stack is unmodified. The thread surfaced past efforts including the OpenAttestation work and the trusted-computing-pools blueprint; this is shaping up as a major 2026.2/PTG design topic, dovetailing with the Intel TDX work raised the prior week.

Cyborg revitalization. Sean Mooney announced his intent to ramp up Cyborg maintenance, explicitly modeling the approach on the Watcher revival he led roughly 18 months earlier. His concrete cleanup plan covers oslo.db/oslo.service compatibility, microversion-parse naming, the long-overdue eventlet removal, and a backlog of bot-proposed patches; stabilizing CI (the cyborg-tempest-plugin lacks stable-branch jobs post-2024.2 and carries EOL branch definitions, with no grenade/SLURP upgrade testing); and cleaning up release metadata and Launchpad tracking. Coordination is happening on this etherpad. This is the kind of focused effort that determines whether a project stays alive.

Ironic weighs a first-class physical location API. Jay Faulkner continued floating the idea of structured physical-world metadata in Ironic — datacenter, room, row, rack, slot, cabling — primarily to make nodes searchable in a performant way (today's freeform JSON fields can't do this). He tied it to in-flight bulk-node-action work ("power down every machine in DC 17, row 5, rack 9") but cautioned that the thread already sprawled from location into serial numbers and purchase dates, and that targeting specific racks undercuts the failure-isolation HA model of nova-fronted Ironic. He plans to bring it to the PTG.

Scalability and stability pain points. Alfredo Moralejo Alonso reported a sharp taskflow scalability problem surfacing in Watcher: with parallelization ≥2, action plans of ~35 tasks degrade badly, and 50-task plans can take up to 45 minutes to schedule the next task (LP 2139228, with a minimal reproducer). A Blazar operator hit repeated SQLAlchemy QueuePool limit timeouts (LP 2139111). And a Keystone/Kolla/OpenStack-Helm thread asked the Keystone team to confirm whether it's safe to run the Keystone WSGI app with multiple threads per process — a question deployment tools need answered as they tune concurrency.

Heads Up / Action Needed

  • DPL liaison deadline (Feb 4): Goutham Pacha Ravi notified the Freezer, Watcher, Requirements, Release, Oslo, and Ironic teams that the TC opened dpl-reset reviews. Teams have until February 4 to oppose and retain the DPL model; otherwise PTL nominations will be sought for them in the 2026.2 elections.
  • Election officials wanted. The TC is recruiting volunteers to shadow the 2026.2 PTL and TC elections — neutral parties not contesting the elections themselves.
  • PTL/TC nomination window confirmed. The R-9 TC summary confirms nominations open at 23:45 UTC on February 4, 2026 and close at 23:45 UTC on February 18, 2026.
  • OpenStack-Ansible revived the os_watcher and os_freezer Ansible roles.
  • uWSGI replacement and a MariaDB/MySQL character-set goal remain under TC discussion; the character-set work is being shaped as a possible 2026.2 community goal.

Community & Events

Several team mid-cycles cluster around early February: Manila (Feb 3 and 5, 15:00–16:30 UTC, recorded to the Manila YouTube channel) and Cinder (Feb 11, 1400–1600 UTC). The Public Cloud SIG held its bi-weekly meeting; the Digital Sovereignty WG pushed its meeting-time poll. The new Ops Radio Hour survey launched (open through February 25) to gather input on cadence, time slots, and discovery channels ahead of the next call on February 27.

A small but handy contribution: Piotr Milewski shared a Tampermonkey script that color-codes Zuul job statuses in Gerrit (green = success, red = voting failure, orange = non-voting failure), making patch health readable at a glance. The InfoScale team (Alokedip Choudhuri) opened a cross-project driver contribution thread spanning Cinder, os-brick, and Nova. Two newcomers introduced themselves on the list — a welcome sign of fresh operator interest — and a Jay Faulkner Ironic review jam started running weekly on Tuesdays 10–11 PT.