From bb30ea841b31e535a6c2d9698265d4e4149052df Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Efe=20G=C3=B6kdemir?= Date: Tue, 22 Sep 2026 04:12:44 +0300 Subject: [PATCH] docs: document GPU Operator network prerequisites MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Efe Gökdemir --- gpu-operator/getting-started.rst | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/gpu-operator/getting-started.rst b/gpu-operator/getting-started.rst index cd0428c9f..7a93323d0 100644 --- a/gpu-operator/getting-started.rst +++ b/gpu-operator/getting-started.rst @@ -60,6 +60,12 @@ Prerequisites #. Nodes must be configured with a container engine such as CRI-O or containerd. +#. The client machine and GPU worker nodes must have outbound HTTPS access to + ``nvcr.io`` and ``ngc.download.nvidia.com``. The Operator pulls container + images from NGC, and driver installation may download driver packages from + the NGC download service. If your cluster is behind a firewall or proxy, + allow TCP port 443 to both endpoints before installing the Operator. + #. If your cluster uses Pod Security Admission (PSA) to restrict the behavior of pods, label the namespace for the Operator to set the enforcement policy to privileged: