The Big Picture

With the 2026.1 "Gazpacho" cycle ~16 weeks from release and spec freezes in effect, this week's center of gravity was the community's own health: the OpenInfra Foundation published a two-part deep dive on the contributor and maintainer experience, and the TC held a frank discussion about governance "bad bureaucracy" and "bit rot." The recurring theme across both — getting review attention is still the single biggest pain point — ties directly to the Foundation's 2026 goal of increasing reviewer bandwidth. On the technical side, Thomas Goirand put forward a concrete heartbeat-reliability proposal, and Takashi Kajinami warned of a breaking dependency change in the whereto library.

Development & Technical Decisions

Heartbeat that actually checks the main thread. Thomas Goirand's proposal is the week's most actionable technical item. After a RabbitMQ outage (disk full), his team found that cinder-volume and cinder-backup main threads were dead while the heartbeat thread kept reporting the services as "up" — a dangerous false-healthy state. His fix: instantiate a manager attached to the service whose is_working() method pings the main thread over RabbitMQ and expects a pong within a timeout, returning False otherwise. He's prototyping in Cinder (WIP at review 970796) and intends to extend it to Nova and Neutron. Reviewers should weigh the trade-off he acknowledges — more RabbitMQ messages per service — against the monitoring reliability gain.

Breaking change incoming: whereto → PCRE2. Takashi Kajinami flagged that the whereto library (used to generate Apache redirect rules in docs builds) is migrating from python-pcre to PCRE2, because newer OSes (Debian; RHEL 10) have dropped the old PCRE3 dev packages. Projects using whereto will need to update their bindep.txt to replace libpcre3-dev/pcre-devel with the PCRE2 equivalents once the new release hits upper-constraints. Kajinami plans to handle much of the transition but asked teams (oslo, cinder, glance, nova, sdk, trove and others) to pay attention.

Nova scheduler and other operator threads. The most active discussion was Andy Speagle working through how to use $aggregates in Nova's JsonFilter for scheduler hints — unclear how the aggregates list is structured in HostState. Other technical threads: an Octavia user wanting an easier way to map a pool member to its compute/server ID; Eduardo Morais hitting a duplicate-entry crash in providerresourceassociations after restarting a freshly OVS→OVN-migrated environment; a Neutron L3 HA report of packet loss / route flapping after restarting openvswitch-vswitchd on the active controller; and a Kolla-Ansible deploy failing at nova-compute registration due to a missing ceph.client.cinder.keyring.

Heads Up / Action Needed

• FIPS compliance goal moved back to "proposed." The TC reported the cross-project FIPS goal is being returned to proposed status, seeking feedback on testing and a goal champion — volunteers needed.
• whereto/PCRE2 bindep updates (above) will be required of affected projects soon.
• Neutron bug deputy (week of Dec 1) was quiet: one Critical (tap-as-a-service py313 jobs timing out, fix proposed) plus two High items, one of which — ovs_agent raising ValueError on an os_ken parser call without kwargs — is unassigned.
• docs.openstack.org link instability: Florian Haas reported intermittent connection failures (suspected rate limiting) breaking external link checkers — relevant if you cross-reference the docs in CI.

Community & Events

The Foundation's two-part contributor/maintainer experience analysis (Ildiko Vancsa) is the standout community read. The metrics round covered 23 project teams across five cycles (Bobcat→Flamingo), tracking a Review Efficiency Index, time-to-review, time-to-merge, patchsets-per-review, and active-maintainer counts; average incoming changes per team dropped ~23% (partly from including less-active teams). The survey round found Flamingo maintainers rated their experience more positively than Epoxy, with review attention still the biggest challenge for both maintainers and contributors — and new pain points emerging around broken CI and review-feedback quality (shallow or overly nit-picky reviews).

That fed directly into the TC's R-16 discussion, which dug into the friction around making big changes: a sentiment that governance has become "bad bureaucracy" suffering "bit rot" (e.g., a Golang PTI that persists despite no real use), and a plan to rewrite the obsolete Technical Vision document as a foundation for pruning outdated policy. The TC is also seeking volunteers to shadow election officials ahead of the February PTL/TC elections. Routine notes: Nova, Octavia, Neutron Drivers, and QA office hours were cancelled or paused for the holidays, and Brian Haley continued the unmaintained-branches-to-EOL work.