Test to verify numa balancing by verifying numa_pte_updates and numa_hint_faults

memory/autonuma.py

@@ -0,0 +1,115 @@
+#!/usr/bin/env python
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+#
+# See LICENSE for more details.
+#
+# Copyright: 2025 IBM
+# Author: Pavithra Prakash <pavrampu@linux.vnet.ibm.com>
+#
+
+
+import os
+from avocado import Test
+from avocado.utils import build
+from avocado.utils import archive
+from avocado.utils import process
+from avocado.utils.software_manager.manager import SoftwareManager
+from avocado.utils import distro
+
+
+class AutoNuma(Test):
+
+    def count_numa_pte_updates(self):
+        """
+        This function retrieves the current value of numa_pte_updates
+        from /proc/vmstat and returns it as an integer.
+        """
+        cmd = 'cat /proc/vmstat  | grep numa_pte_updates'
+        output = process.run(cmd, shell=True).stdout_text
+        return int(output.split()[-1])
+
+    def count_numa_hint_faults(self):
+        """
+        This function fetches the current value of numa_hint_faults from
+        /proc/vmstat and returns it as an integer.
+        """
+        cmd = 'cat /proc/vmstat  | grep numa_hint_faults | head -1'
+        output = process.run(cmd, shell=True).stdout_text
+        return int(output.split()[-1])
+
+    def setUp(self):
+        """
+        This function sets up the test environment by downloading, extracting, and
+        building the ebizzy workload from its source.
+        """
+        smm = SoftwareManager()
+
+        detected_distro = distro.detect()
+        deps = ['gcc', 'make', 'time']
+        if detected_distro.name == "rhel":
+            deps.extend(['numactl-devel'])
+        for packages in deps:
+            if not smm.check_installed(packages) and not smm.install(packages):
+                self.cancel('%s is needed for the test to be run' % packages)
+        self.url = self.params.get('ebizzy_url',
+                                   default='https://sourceforge.net/projects/ebizzy/files/ebizzy/0.3/ebizzy-0.3.tar.gz')
+        tarball = self.fetch_asset("ebizzy-0.3.tar.gz", locations=[self.url], expire='7d')
+        archive.extract(tarball, self.workdir)
+        version = os.path.basename(tarball.split('.tar.')[0])
+        self.sourcedir = os.path.join(self.workdir, version)
+        os.chdir(self.sourcedir)
+        process.run('[ -x configure ] && ./configure', shell=True)
+        build.make(self.sourcedir)
+
+    def test(self):
+        """
+        This test performs the following steps:
+
+        - Records the initial values of numa_pte_updates and numa_hint_faults.
+        - Disables NUMA balancing and runs ebizzy workload.
+        - Records the new values of numa_pte_updates and numa_hint_faults.
+        - Checks if there's any change in the values if yes, the test fails.
+        - Enables NUMA balancing and runs ebizzy workload again.
+        - Checks if numa_pte_updates and numa_hint_faults have increased by at least 10 after
+          enabling NUMA balancing. If not, the test fails.
+        """
+        init_pte_updates = self.count_numa_pte_updates()
+        init_hint_faults = self.count_numa_hint_faults()
+        process.run('echo "0" > /proc/sys/kernel/numa_balancing', shell=True)
+        process.run('./ebizzy', shell=True)
+        pte_updates_0 = self.count_numa_pte_updates()
+        hint_faults_0 = self.count_numa_hint_faults()
+        if(pte_updates_0 - init_pte_updates) or (hint_faults_0 - init_hint_faults):
+            self.fail("Numa balancing disable test failed")
+        process.run('echo "1" > /proc/sys/kernel/numa_balancing', shell=True)
+        process.run('./ebizzy', shell=True)
+        pte_updates_1 = self.count_numa_pte_updates()
+        hint_faults_1 = self.count_numa_hint_faults()
+        if(pte_updates_1 - init_pte_updates) < 10:
+            self.fail("numa_pte_updates has not incremented even after "
+                      "running workload with numa balancing enabled.")
+        if(hint_faults_1 - init_hint_faults) < 10:
+            self.fail("numa_hint_faults has not incremented even after"
+                      " running workload with numa balancing enabled.")
+
+    def test_autonuma_benchmark(self):
+        """
+        This test case runs downloading, extracting, and running the autonuma-benchmark tests,
+        Test results need to be verified manually.
+        """
+        url_autonuma = self.params.get('url_autonuma',
+                                       default='https://github.com/pholasek/autonuma-benchmark/archive/refs/heads/master.zip')
+        tarball = self.fetch_asset("master.zip", locations=[url_autonuma], expire='7d')
+        archive.extract(tarball, self.workdir)
+        self.sourcedir = os.path.join(self.workdir, "autonuma-benchmark-master")
+        os.chdir(self.sourcedir)
+        process.system_output('./start_bench.sh -A', shell=True, ignore_status=True)
+        self.log.warn("Have run the tests please review the results in debug.log")

memory/autonuma.py.data/autonuma.yaml

@@ -0,0 +1,2 @@
+ebizzy_url: 'https://sourceforge.net/projects/ebizzy/files/ebizzy/0.3/ebizzy-0.3.tar.gz'
+url_autonuma: 'https://github.com/pholasek/autonuma-benchmark/archive/refs/heads/master.zip'