[Build] Install the NVIDIA driver and CUDA toolkit from the distribution package manager using NVIDIA local repo packages instead of the run file installers.

Author: gmarciani

.kiro/specs/diagnostics-tool/.config.kiro

@@ -0,0 +1 @@
+{"specId": "01a0b120-5e6f-44b0-9897-0a810b545649", "workflowType": "requirements-first", "specType": "feature"}

.kiro/specs/diagnostics-tool/design.md

@@ -15,6 +15,7 @@ This file is used to list changes made in each version of the AWS ParallelCluste
 
 **CHANGES**
 - Enforce NFSv4-only on the ParallelCluster-managed NFS server (head node). The NFSv3 client stack (rpcbind, rpc-statd, lockd) are unchanged, so cluster nodes can still mount external NFSv3 servers.
+- Install the NVIDIA driver and CUDA toolkit from the distribution package manager using NVIDIA local repo packages instead of the run file installers.
 - In GPU Health Check, skip DCGM diagnostics when NVIDIA MIG is enabled because dcgmi diag does not support MIG.
 - Upgrade Slurm to version 25.11.6 (from 25.11.4).
 - Upgrade EFA installer to 1.49.0 (from 1.47.0).
@@ -37,7 +38,7 @@ via custom Slurm settings or the cluster name contains upper-case letters.
 - Fix compute node bootstrap hanging without a clear error when the compute subnet cannot reach DynamoDB.
 - Fix login nodes not mounting `/opt/parallelcluster/shared` when EFS is used as the internal shared storage type.
 - Fix SELinux not actually being disabled on RHEL-family OSes (kernels >= 6.4) due to a [deprecated mechanism](https://github.com/SELinuxProject/selinux-kernel/wiki/DEPRECATE-runtime-disable) being silently ignored by newer kernels.
-- Fix `build-image` failure by skipping installation of `gdrcopy` and `dcgm` if parent image has an existing installed version.
+- Fix `build-image` failure by skipping installation of `fabric-manager`, `gdrcopy` and `dcgm` if parent image has an existing installed version.
 
 **DEPRECATIONS**
 - Amazon Linux 2 is no longer supported.

.kiro/specs/diagnostics-tool/requirements.md

@@ -28,6 +28,11 @@
 default['cluster']['nvidia']['driver_version'] = '580.105.08'
 default['cluster']['nvidia']['dcgm_version'] = '4.5.1-1'
 
+# Extra packages installed from the NVIDIA local repo alongside the driver
+# meta-package, as a comma-separated list.
+# * nvidia-xconfig: required by DCV to generate the X configuration
+default['cluster']['nvidia']['driver_extra_packages'] = 'nvidia-xconfig'
+
 default['cluster']['nvidia']['driver_base_url'] = "#{node['cluster']['artifacts_s3_url']}/dependencies/nvidia_driver"
 default['cluster']['nvidia']['dcgm_base_url'] = "#{node['cluster']['artifacts_s3_url']}/dependencies/nvidia_dcgm"
 default['cluster']['nvidia']['fabricmanager']['base_url'] = "#{node['cluster']['artifacts_s3_url']}/dependencies/nvidia_fabric"

.kiro/specs/diagnostics-tool/tasks.md

@@ -0,0 +1,37 @@
+Metadata-Version: 2.4
+Name: pcluster-diag
+Version: 1.0.0
+Summary: Diagnostics tool for AWS ParallelCluster nodes.
+Author: Amazon Web Services
+License: Apache-2.0
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: click~=8.0
+Requires-Dist: pyyaml~=6.0
+Provides-Extra: dev
+Requires-Dist: pytest~=7.0; extra == "dev"
+Requires-Dist: pytest-cov~=5.0; extra == "dev"
+Requires-Dist: tox~=4.0; extra == "dev"
+
+# pcluster-diag
+
+Diagnostics tool for AWS ParallelCluster nodes.
+
+It runs a context-aware set of checks against a running cluster and produces a structured
+report as JSON plus console output.
+
+The tool is invoked via the `pcluster-diag` command and runs as root from whatever cluster node.
+
+## Updating the tool
+
+You can update it without waiting for a ParallelCluster release by running the commands below as root.
+
+```bash
+# Resolve the virtualenv pcluster-diag is installed in (from the PATH symlink)
+VENV_PATH="$(dirname "$(dirname "$(readlink -f /usr/local/bin/pcluster-diag)")")"
+
+# From the cookbook repo subfolder on GitHub, taking the source from the develop branch
+# (replace @develop with another branch or tag to pin a different version)
+"${VENV_PATH}/bin/pip" install --upgrade \
+  "git+https://github.com/aws/aws-parallelcluster-cookbook.git@develop#subdirectory=cookbooks/aws-parallelcluster-platform/files/pcluster-diag"
+```

CHANGELOG.md

@@ -0,0 +1,34 @@
+README.md
+pyproject.toml
+pcluster_diag/__init__.py
+pcluster_diag/cli.py
+pcluster_diag.egg-info/PKG-INFO
+pcluster_diag.egg-info/SOURCES.txt
+pcluster_diag.egg-info/dependency_links.txt
+pcluster_diag.egg-info/entry_points.txt
+pcluster_diag.egg-info/requires.txt
+pcluster_diag.egg-info/top_level.txt
+pcluster_diag/checks/__init__.py
+pcluster_diag/checks/cfn_hup.py
+pcluster_diag/checks/sample.py
+pcluster_diag/commands/__init__.py
+pcluster_diag/commands/run.py
+pcluster_diag/core/__init__.py
+pcluster_diag/core/constants.py
+pcluster_diag/core/context_builder.py
+pcluster_diag/core/exception_handler.py
+pcluster_diag/core/registry.py
+pcluster_diag/core/runner.py
+pcluster_diag/models/__init__.py
+pcluster_diag/models/check.py
+pcluster_diag/models/context.py
+pcluster_diag/models/exceptions.py
+pcluster_diag/models/report.py
+pcluster_diag/models/result.py
+pcluster_diag/util/__init__.py
+pcluster_diag/util/io_utils.py
+pcluster_diag/util/logging_utils.py
+pcluster_diag/util/serialization.py
+pcluster_diag/util/services.py
+pcluster_diag/util/shell.py
+tests/test_cli.py

cookbooks/aws-parallelcluster-platform/attributes/platform.rb

@@ -0,0 +1 @@
+

cookbooks/aws-parallelcluster-platform/files/pcluster-diag/pcluster_diag.egg-info/PKG-INFO

@@ -0,0 +1,2 @@
+[console_scripts]
+pcluster-diag = pcluster_diag.cli:main