<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>openstack/nova — Stackers Network</title>
    <link>https://stackers.network/projects/openstack-nova/</link>
    <description>Weekly code-activity digest for openstack/nova from Stackers Network.</description>
    <item>
      <title>openstack/nova — Week of 2026-07-18</title>
      <link>https://stackers.network/projects/openstack-nova/2026-07-18</link>
      <guid>https://stackers.network/projects/openstack-nova/2026-07-18</guid>
      <pubDate>Sat, 18 Jul 2026 00:00:00 +0000</pubDate>
      <description>AMD SEV-SNP support. The libvirt driver gained end-to-end detection and configuration for AMD SEV-SNP, spanning roughly 1,600 lines across capability probing, host validation, and domain XML generation. The launchSecurity element now understands SEV-SNP, SEV-ES is ignored when…</description>
      <content:encoded><![CDATA[<p><strong>AMD SEV-SNP support.</strong> The libvirt driver gained end-to-end detection and configuration for AMD SEV-SNP, spanning roughly 1,600 lines across capability probing, host validation, and domain XML generation. The launchSecurity element now understands SEV-SNP, SEV-ES is ignored when SEV-SNP is detected, and startup-time validation of SEV/SEV-ES support replaces the old per-launch capability check. Related cleanups drop iommu='on' for virtio devices under SEV and omit optional launchSecurity fields for the plain SEV/SEV-ES cases. See the detection groundwork at https://review.opendev.org/c/openstack/nova/+/994764.</p>
<p><strong>Compute startup and driver interface.</strong> Instance startup processing on compute restart is now delegated to the virt drivers themselves, restructuring ~840 lines in the compute manager and libvirt driver. A companion reproducer for vTPM startup validation lands alongside it. Image metadata also learned to accept <code translate="no">hw_firmware_stateless=True</code> for the BIOS firmware type.</p>
<p><strong>Bug fixes.</strong> The websocket proxy no longer mutates global config (fixed on both master and a stable branch), the API stops returning empty <code translate="no">attachment_id</code> values on volume attachments, and <code translate="no">QEMU_IMG_LIMITS</code> process limits become configurable across three branches. Smaller fixes cover a service_ref destroy check plus functional-test stabilization and speedups.</p>
<p><strong>Maintenance.</strong> Dead code from the old image v1 API path and <code translate="no">UpdateGlanceImage</code> was removed, along with stale policy and cursive doc references.</p>]]></content:encoded>
    </item>
    <item>
      <title>openstack/nova — Week of 2026-07-11</title>
      <link>https://stackers.network/projects/openstack-nova/2026-07-11</link>
      <guid>https://stackers.network/projects/openstack-nova/2026-07-11</guid>
      <pubDate>Sat, 11 Jul 2026 00:00:00 +0000</pubDate>
      <description>Concurrency model. The largest change introduces a concurrency_backend config option for Nova services (review.opendev.org/c/openstack/nova/+/993760), letting operators select the underlying concurrency implementation, with a follow-up correcting the release name referenced by…</description>
      <content:encoded><![CDATA[<p><strong>Concurrency model.</strong> The largest change introduces a <code translate="no">concurrency_backend</code> config option for Nova services (<a href="https://review.opendev.org/c/openstack/nova/+/993760">review.opendev.org/c/openstack/nova/+/993760</a>), letting operators select the underlying concurrency implementation, with a follow-up correcting the release name referenced by the new option. Related test-infrastructure work moves the Database and CellDatabases fixtures to file-backed SQLite with WAL in threading mode, and extends the same file-backed Placement SQLite approach to functional threading tests — laying groundwork for exercising Nova under the new backend.</p>
<p><strong>Bug fixes.</strong> Two matching patches fix mutation of global config in the websocket proxy, landing on master and a stable branch. The API now returns 400 rather than a server error for invalid locked-memory properties, and <code translate="no">QEMU_IMG_LIMITS</code> process limits become configurable rather than hardcoded. A test-only fix addresses intermittent 409s in <code translate="no">TestGracefulShutdown</code>.</p>
<p><strong>Elsewhere.</strong> The <code translate="no">cyborg-tempest</code> job is renamed to <code translate="no">cyborg-tempest-py3</code>.</p>]]></content:encoded>
    </item>
    <item>
      <title>openstack/nova — Week of 2026-07-04</title>
      <link>https://stackers.network/projects/openstack-nova/2026-07-04</link>
      <guid>https://stackers.network/projects/openstack-nova/2026-07-04</guid>
      <pubDate>Sat, 04 Jul 2026 00:00:00 +0000</pubDate>
      <description>Policy and test scaffolding. The largest change by far is a rework of policy scope enforcement testing, trimming roughly 2,400 net lines from the test suite (986946). Alongside it, mtty support landed for exercising fake mdevs, and the vTPM CI job was renamed and extended to…</description>
      <content:encoded><![CDATA[<p><strong>Policy and test scaffolding.</strong> The largest change by far is a rework of policy scope enforcement testing, trimming roughly 2,400 net lines from the test suite (<a href="https://review.opendev.org/c/openstack/nova/+/986946">986946</a>). Alongside it, mtty support landed for exercising fake mdevs, and the vTPM CI job was renamed and extended to cover vGPU testing via the same mtty path — giving the mediated-device code paths a functional gate without real hardware.</p>
<p><strong>Feature and config cleanup.</strong> The long-unused <code translate="no">shutdown_terminate</code> instance flag was removed, and the libvirt driver dropped its dependency on <code translate="no">[libvirt] num_memory_encrypted_guests</code>, deriving the limit from libvirt itself instead. Several rounds of dead-code removal and type fixes in <code translate="no">privsep</code>, <code translate="no">utils</code>, and libvirt logging round out the maintenance sweep, along with a <code translate="no">pyproject</code> packaging update.</p>
<p><strong>Bug fixes.</strong> The websockify proxy no longer mutates global config on each connection, and noVNC traffic logging is now gated on <code translate="no">debug=True</code>. On the compute side, <code translate="no">init_host</code> pre-loads flavor and <code translate="no">system_metadata</code> to avoid repeated lazy loads, live-migration ensures resource limits are set before <code translate="no">_call_livem_checks_on_host</code>, and the security-group port query now restricts returned fields. Smaller fixes address a CLI-exit stacktrace, a mis-scoped mock, and a regression test for bug #2153425.</p>]]></content:encoded>
    </item>
    <item>
      <title>openstack/nova — Week of 2026-06-20</title>
      <link>https://stackers.network/projects/openstack-nova/2026-06-20</link>
      <guid>https://stackers.network/projects/openstack-nova/2026-06-20</guid>
      <pubDate>Sat, 20 Jun 2026 00:00:00 +0000</pubDate>
      <description>Availability zones. A new unpin-AZ capability landed, allowing instances pinned to an availability zone to be detached from that pinning, together with the accompanying documentation. The feature accounts for roughly 400 lines of behavior change plus user-facing docs. Scheduler…</description>
      <content:encoded><![CDATA[<p><strong>Availability zones.</strong> A new unpin-AZ capability landed, allowing instances pinned to an availability zone to be detached from that pinning, together with the accompanying documentation. The feature accounts for roughly <a href="https://review.opendev.org/c/openstack/nova/+/990682">400 lines of behavior change</a> plus user-facing docs.</p>
<p><strong>Scheduler hints hardening.</strong> Internal <code translate="no">_nova</code>-prefixed scheduler hints are now stripped on server create, with the same fix landed four times across active branches as backports. Elsewhere on the API side, a bug was fixed where multiple security groups sharing the same name within a project caused errors, with a dedicated reproducer test added for bug 2105896.</p>
<p><strong>Testing and stability.</strong> A substantial (~480 line) functional test suite now exercises graceful shutdown of the compute service. An intermittent <code translate="no">test_refresh_nfs_swap_cache_stale</code> failure was fixed, and the <code translate="no">nova-alt-configurations</code> job was reworked. On the runtime side, daemon mode now restores the <code translate="no">fork</code> start method on Python 3.14, restoring compatibility with the upcoming interpreter.</p>
<p><strong>Maintenance.</strong> Release notes builds were sped up via a small docs configuration change.</p>]]></content:encoded>
    </item>
    <item>
      <title>openstack/nova — Week of 2026-06-13</title>
      <link>https://stackers.network/projects/openstack-nova/2026-06-13</link>
      <guid>https://stackers.network/projects/openstack-nova/2026-06-13</guid>
      <pubDate>Sat, 13 Jun 2026 00:00:00 +0000</pubDate>
      <description>Release notes housekeeping. Most of the week's ~18k-line churn comes from converting EOL and EOM release notes into static pages (989210 and a companion change for EOM), freezing historical notes out of the live build. Bug fixes. Several long-standing edge cases got cleaned up…</description>
      <content:encoded><![CDATA[<p><strong>Release notes housekeeping.</strong> Most of the week's ~18k-line churn comes from converting EOL and EOM release notes into static pages (<a href="https://review.opendev.org/c/openstack/nova/+/989210">989210</a> and a companion change for EOM), freezing historical notes out of the live build.</p>
<p><strong>Bug fixes.</strong> Several long-standing edge cases got cleaned up. Resize was tightened to reject unpermitted flavors, port handling was fixed so that an error while updating a port on a multi-network instance no longer leaves the port hanging around, and an AttributeError in <code translate="no">Service.stop()</code> during early-startup shutdown was patched. A separate fix addresses an error that surfaced when multiple security groups in the same project shared a name. Each fix landed alongside a regression or reproducer test.</p>
<p><strong>Libvirt and config surface.</strong> The libvirt driver gained a new config option for the live migration monitor interval, giving operators a knob over how frequently migrations are polled. A small libvirt utility that no longer had callers was dropped.</p>
<p><strong>Maintenance.</strong> Cell URL parsing and Host header parsing were both reworked to use shared oslo.utils helpers instead of in-tree code. On the CI side, the graceful-shutdown jobs picked up <code translate="no">irrelevant-files</code> rules, and a flaky resize-reschedule functional test got more generous wait retries.</p>]]></content:encoded>
    </item>
    <item>
      <title>openstack/nova — Week of 2026-06-06</title>
      <link>https://stackers.network/projects/openstack-nova/2026-06-06</link>
      <guid>https://stackers.network/projects/openstack-nova/2026-06-06</guid>
      <pubDate>Sat, 06 Jun 2026 00:00:00 +0000</pubDate>
      <description>Quota and limits. The largest functional change reworks unified limits to cover all resource types, closing gaps where certain resources weren't being enforced through the Keystone limits path (975872). Alongside that, resize to flavors a project isn't permitted to use is now…</description>
      <content:encoded><![CDATA[<p><strong>Quota and limits.</strong> The largest functional change reworks unified limits to cover all resource types, closing gaps where certain resources weren't being enforced through the Keystone limits path (<a href="https://review.opendev.org/c/openstack/nova/+/975872">975872</a>). Alongside that, resize to flavors a project isn't permitted to use is now properly rejected.</p>
<p><strong>Threading migration.</strong> Eventlet imports are now poisoned when running in native threading mode, hardening the ongoing move off eventlet by making accidental imports fail loudly rather than silently re-introducing greenlet behavior.</p>
<p><strong>Bug fixes.</strong> A handful of guest- and storage-path bugs landed: swap disk creation is no longer skipped on NFS during cold migration; the LVM image creation path no longer hangs on an interactive prompt; the RFB/VNC code now refuses to read up to 4 GiB of reason text from a remote endpoint; and an error triggered by duplicate security group names within a project was resolved. A reproducer test for bug 2105896 was also added.</p>
<p><strong>Docs and CI.</strong> A substantial new document lands agentic-coding guidance for contributors working with AI assistants on Nova. On the CI side, the nova-vtpm job drops to <code translate="no">tempest_concurrency=1</code>, the coverage job gets another timeout bump, a stable-only grenade-multinode fix lands, and a stray log-message typo is corrected.</p>]]></content:encoded>
    </item>
    <item>
      <title>openstack/nova — Week of 2026-05-23</title>
      <link>https://stackers.network/projects/openstack-nova/2026-05-23</link>
      <guid>https://stackers.network/projects/openstack-nova/2026-05-23</guid>
      <pubDate>Sat, 23 May 2026 00:00:00 +0000</pubDate>
      <description>Libvirt and virt drivers. The week's headline change retires Linux bridge VIF support from Nova, pulling roughly 160 lines of legacy plumbing out of the libvirt driver and closing out a long-deprecated networking path in favor of OVS and friends…</description>
      <content:encoded><![CDATA[<p><strong>Libvirt and virt drivers.</strong> The week's headline change retires Linux bridge VIF support from Nova, pulling roughly 160 lines of legacy plumbing out of the libvirt driver and closing out a long-deprecated networking path in favor of OVS and friends (<a href="https://review.opendev.org/c/openstack/nova/+/957040">review.opendev.org/c/openstack/nova/+/957040</a>). Alongside that, stale comments were swept out of the ironic virt driver, and a lingering type error in the console code path was corrected.</p>
<p><strong>Quotas, security groups, and ports.</strong> Several user-facing correctness fixes landed in the API and resource-tracker neighborhood. User-scoped quota usage counts are now computed correctly, security-group lookups no longer fall over when two groups in the same project share a name, and a regression where ports remained allocated after a failed multi-network update was repaired. <code translate="no">get_flavor_by_flavor_id()</code> was also refactored to require an explicit context, tightening an internal contract used across the scheduler and compute managers.</p>
<p><strong>Stability and hygiene.</strong> A metadata-proxy change stops the shared secret from being written to logs, and <code translate="no">exc_info</code> logging in image deletion was fixed. Two reproducer tests for outstanding bugs (2152697 and 2105896) merged, and CI noise from eventlet deprecation warnings was further muted.</p>]]></content:encoded>
    </item>
  </channel>
</rss>
