The Big Picture

The first full working week of 2026 was quiet and operator-flavored, with the community easing back from the holidays — the TC explicitly skipped its January 6 meeting because members were still on PTO. No single thread dominated; the most active discussion was a Ceilometer metrics question. The most consequential items were structural and forward-looking: a real proposal for a NetApp-backed Nova image backend, a renewed debate over what comes after uWSGI, a concerning memory-leak diagnosis in the openstacksdk metrics path, and the 2026.2 "H" release-name poll closing.

Releases & Announcements

The 2026.2 release-name poll (names starting with "H") ran by Allison Price and closed Thursday, January 8 at 17:00 UTC. Élod Illés posted the release countdowns for weeks R-12 and R-11 as the 2026.1/Gazpacho cycle progresses toward its planned early-April release. The TC confirmed there was no weekly IRC meeting on January 6 (members on PTO), with the next meeting set for January 13.

Development & Technical Decisions

A real openstacksdk/keystoneauth1 memory leak. Sam Clippinger posted a detailed diagnosis worth attention from anyone running at scale: every client request through Proxy.request() calls _report_stats(), and the Prometheus path records per-URL counters and histograms keyed on the full request URL — with no config switch to disable it. Because nova-scheduler queries Placement using each VM's UUID, every unique URL spawns its own counter. In their build farm (over 500K VMs/week, Caracal), each of six nova-scheduler copies climbs to ~8 GB over a few weeks until workers get OOM-killed and VMs land in error state. He notes prometheus_client can't simply be uninstalled (it's pulled in transitively via oslo.metrics → oslo.messaging) and asks whether keystoneauth1's Adapter/CloudRegion should honor a collect-timing-style off switch. This is a genuine, reproducible production scaling bug, not a configuration mistake. (A related "Memory leak in nova-scheduler, possibly all services?" thread from Mike Lowe drew several replies.)

What replaces uWSGI? Picking up a thread that has resurfaced periodically since 2023, a contributor noted uWSGI is still in maintenance mode three years on, and that the OpenStack-Ansible plan from the Epoxy PTG to find a replacement hasn't materialized — OSA still relies on uWSGI. The open question to the community: what should the next "standard" WSGI runtime be — gunicorn, uvicorn, granian, or something else — given operators benefit most from a single well-documented, well-supported option. No decision yet, but a real strategic gap.

NetApp FlexClone for Nova ephemeral storage. Jayaanand Borra proposed a new NetAppNfs image backend in nova/virt/libvirt/imagebackend.py that calls NetApp's ONTAP REST /storage/file/clone endpoint to offload cloning to the array — aiming for Ceph-like, sub-second, metadata-only instance creation on NFS instead of the slower QCOW2-over-NFS copy path. He's soliciting design feedback (driver red flags, credential handling via oslo.privsep, community interest). A plausible performance win for NFS-backed deployments if it gains traction.

Smaller technical threads. Sylvain followed up (from December) on Designate's "Updating a root zone NS record is not allowed" restriction blocking a multi-region PRIMARY/SECONDARY setup across three independent deployments, asking whether the ~9–10-year-old restriction could be made configurable. A Nova resize bug with swap vs. no-swap flavor combos (duplicate vdb) drew confirmation. And an OSA operator on Epoxy asked how to back up the upgraded MariaDB Galera. Sean Mooney pointed a user toward correct flavor/placement usage for NVIDIA T4 resources rather than custom resource classes.

Community & Events

Takashi Kajinami thanked Alan Bishop, who resigned from the puppet-cinder, puppet-glance, and puppet-manila core teams due to a change in work status — a notable loss of review capacity for the Puppet modules. On the metrics front, Jay Faulkner opened his critique of the new (LFX Insights) dashboards, finding them far thinner than the existing ones for his core use cases: per-reviewer Gerrit stats, company comparisons beyond the top 5, and per-project filtering. Lajos Katona continued a Keystone/LDAP thread on whether interpreting LDAP password-expiry could be added (and what process — spec, RFE, or bug — would be required). Kostya Velychkovsky pointed an operator toward OpenStack Freezer for backup needs. Tacker cancelled its weekly IRC meeting again.