test: Added an RDMA validation testcase that checks the expected RDMA enablement and Azure RDMA persistent naming setup. (#77)

To automatically verify that the role correctly enables RDMA in waagent, installs RDMA userland tools, and (on Azure/systemd) configures and maintains Azure persistent RDMA naming services.
How to run:
Execute {{ __hpc_azure_tests_dir }}/test-rdma.sh after the role completes.
Expected result:
Exit 0 with “Test Passed …” lines; non-zero with “Failed: …” explaining the missing/failed prerequisite.

Moved "Create Azure HPC resource directories" task at beginning to avoid path not found issue for other tasks.

Signed-off-by: Gaurav Goklani <ggoklani@redhat.com>
Co-authored-by: Gaurav Goklani <ggoklani@ggoklani-thinkpadt14gen4.punetw6.csb>

Author: ggoklani

files/rdma/test-rdma.sh

@@ -0,0 +1,99 @@
+#!/usr/bin/env bash
+# SPDX-License-Identifier: MIT
+#
+# RDMA Validation Script
+# Usage: test-rdma.sh
+#
+
+# This is test code, and some operations are expected to fail. Hence we can't
+# use set -e to automatically exit the script if something fails.
+set -u
+
+fail()
+{
+	echo Failed: "$1"
+	exit 1
+}
+
+require_file() {
+  local path="$1"
+  [[ -e "$path" ]] || fail "missing file: $path"
+}
+
+require_executable() {
+  local path="$1"
+  [[ -x "$path" ]] || fail "not executable: $path"
+}
+
+require_cmd() {
+  local cmd="$1"
+  command -v "$cmd" >/dev/null 2>&1 || fail "missing command in PATH: $cmd"
+}
+
+sys_vendor() {
+  if [[ -r /sys/class/dmi/id/sys_vendor ]]; then
+    cat /sys/class/dmi/id/sys_vendor
+  else
+    echo ""
+  fi
+}
+
+is_systemd() {
+  [[ "$(ps -p 1 -o comm= 2>/dev/null || true)" == "systemd" ]]
+}
+
+main() {
+	echo
+	echo "Testing waagent RDMA flag"
+	require_file /etc/waagent.conf
+	grep -Fxq "OS.EnableRDMA=y" /etc/waagent.conf || fail "expected 'OS.EnableRDMA=y' in /etc/waagent.conf"
+	echo Test Passed: "waagent RDMA flag is set"
+
+	echo
+	echo "Testing RDMA userland tools"
+	require_cmd ibv_devinfo
+	echo Test Passed: "RDMA tools are present (ibv_devinfo)"
+
+	# Azure persistent RDMA naming artifacts/services (Azure only)
+	if [ "$(sys_vendor)" != "Microsoft Corporation" ]; then
+		echo
+		echo "Testing Azure persistent RDMA naming (skip: not Azure)"
+		echo Test Passed: "not running on Azure; Azure persistent RDMA naming checks skipped"
+		return 0
+	fi
+
+	if ! is_systemd; then
+		echo
+		echo "Testing Azure persistent RDMA naming (skip: not systemd)"
+		echo Test Passed: "not running systemd; systemd unit checks skipped"
+		return 0
+	fi
+
+	echo
+	echo "Testing Azure persistent RDMA naming artifacts"
+	require_executable /usr/sbin/azure_persistent_rdma_naming.sh
+	require_executable /usr/sbin/azure_persistent_rdma_naming_monitor.sh
+	require_file /etc/systemd/system/azure_persistent_rdma_naming.service
+	require_file /etc/systemd/system/azure_persistent_rdma_naming_monitor.service
+	require_file /etc/udev/rules.d/99-azure-persistent-rdma-naming.rules
+	echo Test Passed: "Azure persistent RDMA naming artifacts exist"
+
+	echo
+	echo "Testing Azure persistent RDMA naming services"
+	require_cmd systemctl
+	systemctl is-enabled azure_persistent_rdma_naming.service >/dev/null 2>&1 || fail "azure_persistent_rdma_naming.service not enabled"
+	systemctl is-enabled azure_persistent_rdma_naming_monitor.service >/dev/null 2>&1 || fail "azure_persistent_rdma_naming_monitor.service not enabled"
+
+	# azure_persistent_rdma_naming.service is Type=oneshot, so it may not remain
+	# "active" after it runs. Treat "failed" as an error; other states are OK.
+	if [ "$(systemctl is-failed azure_persistent_rdma_naming.service 2>/dev/null || true)" = "failed" ]; then
+		fail "azure_persistent_rdma_naming.service is in failed state"
+	fi
+
+	# Monitor service should be continuously running.
+	systemctl is-active azure_persistent_rdma_naming_monitor.service >/dev/null 2>&1 || fail "azure_persistent_rdma_naming_monitor.service not active"
+	echo Test Passed: "Azure persistent RDMA naming services look healthy"
+}
+
+main "$@"
+

tasks/main.yml

@@ -313,6 +313,20 @@
           ternary('ansible.posix.rhel_rpm_ostree', omit) }}"
       register: __hpc_azure_packages_install
       until: __hpc_azure_packages_install is success
+      
+- name: Create Azure HPC resource directories
+  file:
+    path: "{{ item }}"
+    state: directory
+    owner: root
+    group: root
+    mode: '0755'
+  loop:
+    - "{{ __hpc_azure_resource_dir }}"
+    - "{{ __hpc_azure_resource_dir }}/bin"
+    - "{{ __hpc_azure_tools_dir }}"
+    - "{{ __hpc_azure_tests_dir }}"
+    - "{{ __hpc_azure_runtime_dir }}"
 
 - name: Install NVidia driver
   # Note that currently the role supports only Microsoft Azure
@@ -539,6 +553,14 @@
             state: started
             daemon_reload: "{{ __hpc_azure_persistent_rdma_naming_monitor_unit.changed | d(false) }}"
 
+    - name: Install RDMA validation script
+      copy:
+        src: rdma/test-rdma.sh
+        dest: "{{ __hpc_azure_tests_dir }}/test-rdma.sh"
+        owner: root
+        group: root
+        mode: "0755"
+
 - name: Install common OpenMPI packages
   when: hpc_install_system_openmpi or hpc_build_openmpi_w_nvidia_gpu_support
   package:
@@ -959,20 +981,6 @@
         mode: '0644'
       notify: Reload udev
 
-- name: Create Azure HPC resource directories
-  file:
-    path: "{{ item }}"
-    state: directory
-    owner: root
-    group: root
-    mode: '0755'
-  loop:
-    - "{{ __hpc_azure_resource_dir }}"
-    - "{{ __hpc_azure_resource_dir }}/bin"
-    - "{{ __hpc_azure_tools_dir }}"
-    - "{{ __hpc_azure_tests_dir }}"
-    - "{{ __hpc_azure_runtime_dir }}"
-
 - name: Install SKU Customisation scripts and services
   when: hpc_sku_customisation
   block: