From 9a5fb59ddcd9b360259216050587dec71b87e31c Mon Sep 17 00:00:00 2001 From: Abigail McCarthy <20771501+a-mccarthy@users.noreply.github.com> Date: Tue, 14 Apr 2026 15:03:57 -0400 Subject: [PATCH] Update docs for 26.3.1 release Signed-off-by: Abigail McCarthy <20771501+a-mccarthy@users.noreply.github.com> Co-authored-by: Tariq --- gpu-operator/life-cycle-policy.rst | 14 +++++++++-- gpu-operator/release-notes.rst | 38 ++++++++++++++++++++++++++++++ repo.toml | 2 +- 3 files changed, 51 insertions(+), 3 deletions(-) diff --git a/gpu-operator/life-cycle-policy.rst b/gpu-operator/life-cycle-policy.rst index dcc09cb7d..f36c6646d 100644 --- a/gpu-operator/life-cycle-policy.rst +++ b/gpu-operator/life-cycle-policy.rst @@ -80,7 +80,7 @@ When post-release testing confirms support for newer versions of operands, these Refer to :ref:`Upgrading the NVIDIA GPU Operator` for more information. .. note:: - All the following components are supported as :ref:`government-ready ` in the NVIDIA GPU Operator v26.3.0, except for NVIDIA GDS Driver, NVIDIA Confidential Computing Manager, and NVIDIA GDRCopy Driver. + All the following components are supported as :ref:`government-ready ` in the NVIDIA GPU Operator v26.3, except for NVIDIA GDS Driver, NVIDIA Confidential Computing Manager, and NVIDIA GDRCopy Driver. **D** = Default driver, **R** = Recommended driver @@ -88,9 +88,10 @@ Refer to :ref:`Upgrading the NVIDIA GPU Operator` for more information. :header-rows: 2 * - :rspan:`1` Component - - GPU Operator Version + - :cspan:`1` GPU Operator Version * - v26.3.0 + - v26.3.1 * - NVIDIA GPU Driver |ki|_ - | `595.58.03 `_ @@ -98,6 +99,11 @@ Refer to :ref:`Upgrading the NVIDIA GPU Operator` for more information. | `580.126.20 `_ (**D**, **R**) | `570.211.01 `_ | `535.288.01 `_ + - | `595.58.03 `_ + | `590.48.01 `_ + | `580.126.20 `_ (**D**, **R**) + | `570.211.01 `_ + | `535.288.01 `_ * - NVIDIA Driver Manager for Kubernetes - `v0.10.0 `__ @@ -126,6 +132,7 @@ Refer to :ref:`Upgrading the NVIDIA GPU Operator` for more information. * - Validator for NVIDIA GPU Operator - v26.3.0 + - v26.3.1 * - NVIDIA KubeVirt GPU Device Plugin - `v1.5.0 `__ @@ -139,12 +146,15 @@ Refer to :ref:`Upgrading the NVIDIA GPU Operator` for more information. * - | NVIDIA Confidential Computing | Manager for Kubernetes - `v0.3.0 `__ + - `v0.4.0 `__ * - NVIDIA GDRCopy Driver - `v2.5.1 `__ + - `v2.5.2 `__ * - NVIDIA Kata Sandbox Device Plugin - `v0.0.2 `__ + - `v0.0.3 `__ .. _known-issue: diff --git a/gpu-operator/release-notes.rst b/gpu-operator/release-notes.rst index 13ede10ba..038c9f3c9 100644 --- a/gpu-operator/release-notes.rst +++ b/gpu-operator/release-notes.rst @@ -33,6 +33,44 @@ Refer to the :ref:`GPU Operator Component Matrix` for a list of software compone ---- +.. _v26.3.1: + +26.3.1 +======= + +New Features +------------ + +* Updated software component versions: + + - NVIDIA GDRCopy Driver v2.5.2 + - NVIDIA Kata Sandbox Device Plugin v0.0.3 + - NVIDIA Confidential Computing Manager for Kubernetes v0.4.0 + +* The ClusterPolicy and NVIDIADriver custom resources now support ``hostNetwork`` for all GPU Operator operands. + Previously, only DCGM Exporter supported the ``hostNetwork`` field. + Setting ``hostNetwork: true`` for a component causes its pods to share the host's network namespace, + binding directly to the host's network interfaces and IP address rather than using the cluster's virtual network. + This is useful in environments where GPU Operator component pods need to expose ports directly on the host network, + such as when a Prometheus instance scrapes metrics from the host network namespace, + or in bare-metal and HPC environments where cluster network overhead or non-standard network configuration + makes host networking preferable. (`PR #2246 `_) + +* Added support for mounting ``/lib/modules`` from the host when using precompiled drivers. + This is required for precompiled driver containers on SUSE Linux Enterprise Server (SLES) 15 SP7 and SLES 16, + which use host kernel modules without requiring the full kernel to be bundled in the driver container. (`PR #2252 `_) + + +Fixed Issues +------------ + +* Fixed an issue in the OLM bundle where the NVIDIA KubeVirt GPU Device Plugin referenced an amd64-only image digest instead of a multi-arch digest. + On ARM servers, this caused pods to fail with an ``Exec format error``. (`PR #2265 `_) + +* Fixed an issue where the operating system release name was recalculated from the node label tag rather than being stored when it was first retrieved. + This could cause errors when the tag format was not recognized. (`PR #2244 `_) + + .. _v26.3.0: 26.3.0 diff --git a/repo.toml b/repo.toml index b96407585..2c098d86d 100644 --- a/repo.toml +++ b/repo.toml @@ -172,7 +172,7 @@ docs_root = "${root}/gpu-operator" project = "gpu-operator" name = "NVIDIA GPU Operator" version = "26.3" # Update repo_docs.projects.openshift.version to match latest patch version maj.min.patch -source_substitutions = { minor_version = "26.3", version = "v26.3.0", recommended = "580.126.20", dra_version = "25.12.0" } +source_substitutions = { minor_version = "26.3", version = "v26.3.1", recommended = "580.126.20", dra_version = "25.12.0" } copyright_start = 2020 sphinx_exclude_patterns = [ "life-cycle-policy.rst",