SRE-3703 ci: Fault injection testing stage on VM/bare metal

Jenkinsfile

@@ -47,42 +47,6 @@ void job_step_update(def value=currentBuild.currentResult) {
     jobStatusUpdate(job_status_internal, env.STAGE_NAME, value)
 }
 
-Map nlt_test() {
-    // groovylint-disable-next-line NoJavaUtilDate
-    Date startDate = new Date()
-    try {
-        unstash('nltr')
-    } catch (e) {
-        print 'Unstash failed, results from NLT stage will not be included'
-    }
-    sh label: 'Fault injection testing using NLT',
-       script: './ci/docker_nlt.sh --class-name fault-injection fi'
-    List filesList = []
-    filesList.addAll(findFiles(glob: '*.memcheck.xml'))
-    int vgfail = 0
-    int vgerr = 0
-    if (filesList) {
-        String rcs = sh label: 'Check for Valgrind errors',
-               script: "grep -E '<error( |>)' ${filesList.join(' ')} || true",
-               returnStdout: true
-        if (rcs) {
-            vgfail = 1
-        }
-        String suite = sanitizedStageName()
-        junitSimpleReport suite: suite,
-                          file: suite + '_valgrind_results.xml',
-                          fails: vgfail,
-                          errors: vgerr,
-                          name: 'Valgrind_Memcheck',
-                          class: 'Valgrind',
-                          message: 'Valgrind Memcheck error detected',
-                          testdata: rcs
-    }
-    int runTime = durationSeconds(startDate)
-    Map runData = ['nlttest_time': runTime]
-    return runData
-}
-
 // For master, this is just some wildly high number
 String next_version() {
     return '1000'
@@ -392,8 +356,11 @@ pipeline {
                defaultValue: 'ci_vm9',
                description: 'Label to use for 9 VM functional tests')
         string(name: 'CI_NLT_1_LABEL',
-               defaultValue: 'ci_nlt_1',
+               defaultValue: 'ci_nlt_vm1',
                description: 'Label to use for NLT tests')
+        string(name: 'CI_FI_1_LABEL',
+               defaultValue: 'ci_fi_vm1',
+               description: 'Label to use for Fault Injection (FI) tests')
         string(name: 'FUNCTIONAL_HARDWARE_MEDIUM_LABEL',
                defaultValue: 'ci_nvme5',
                description: 'Label to use for the Functional Hardware Medium (MD on SSD) stages')
@@ -750,11 +717,22 @@ pipeline {
                         job_step_update(
                             unitTest(timeout_time: 60,
                                      inst_repos: daosRepos(),
-                                     test_script: 'ci/unit/test_nlt.sh',
+                                     test_script: 'ci/unit/test_nlt.sh' +
+                                                  ' --system-ram-reserved 4' +
+                                                  ' --max-log-size 1950MiB' +
+                                                  ' --dfuse-dir /localhome/jenkins/' +
+                                                  ' --log-usage-save nltir.xml' +
+                                                  ' --log-usage-export nltr.json' +
+                                                  ' --class-name nlt all',
+                                     with_valgrind: 'memcheck',
+                                     valgrind_pattern: '*memcheck.xml',
+                                     always_script: 'ci/unit/test_nlt_post.sh',
+                                     testResults: 'nlt-junit.xml',
                                      unstash_opt: true,
                                      unstash_tests: false,
                                      inst_rpms: unitPackages(target: 'el9'),
-                                     image_version: 'el9.7'))
+                                     image_version: 'el9.7',
+                                     prov_env_vars: 'VM_CPUS=14'))
                         // recordCoverage(tools: [[parser: 'COBERTURA', pattern:'nltir.xml']],
                         //                 skipPublishingChecks: true,
                         //                 id: 'tlc', name: 'Fault Injection Interim Report')
@@ -764,8 +742,9 @@ pipeline {
                         always {
                             unitTestPost artifacts: ['nlt_logs/'],
                                          testResults: 'nlt-junit.xml',
-                                         always_script: 'ci/unit/test_nlt_post.sh',
-                                         valgrind_stash: 'nlt-memcheck'
+                                         valgrind_stash: 'nlt-memcheck',
+                                         valgrind_pattern: '*memcheck.xml',
+                                         NLT: true
                             recordIssues enabledForFailure: true,
                                          failOnError: false,
                                          ignoreQualityGate: true,
@@ -989,52 +968,35 @@ pipeline {
                         expression { !skipStage() }
                     }
                     agent {
-                        dockerfile {
-                            filename 'utils/docker/Dockerfile.el.9'
-                            label 'docker_runner_fi'
-                            additionalBuildArgs dockerBuildArgs(repo_type: 'stable',
-                                                                parallel_build: true,
-                                                                deps_build: true) + 
-                                                ' --build-arg POINT_RELEASE=.7 ' +
-                                                " --build-arg PYTHON_VERSION=${env.PYTHON_VERSION}"
-                            args '--tmpfs /mnt/daos_0'
-                        }
+                        label params.CI_FI_1_LABEL
                     }
                     steps {
                         job_step_update(
-                            sconsBuild(parallel_build: true,
-                                       scons_args: 'PREFIX=/opt/daos TARGET_TYPE=release BUILD_TYPE=debug',
-                                       build_deps: 'no'))
-                        job_step_update(nlt_test())
-                        // recordCoverage(tools: [[parser: 'COBERTURA', pattern:'nltr.xml']],
-                        //                skipPublishingChecks: true,
-                        //                id: 'fir', name: 'Fault Injection Report')
+                            unitTest(timeout_time: 240,
+                                     inst_repos: daosRepos(),
+                                     test_script: 'ci/unit/test_nlt.sh --memcheck no' +
+                                                  ' --system-ram-reserved 4 --server-debug WARN' +
+                                                  ' --log-usage-import nltr.json' +
+                                                  ' --log-usage-save nltr.xml' +
+                                                  ' --class-name fault-injection fi',
+                                     with_valgrind: '',
+                                     always_script: 'ci/unit/test_nlt_post.sh',
+                                     testResults: 'nlt-junit.xml',
+                                     unstash_opt: true,
+                                     unstash_tests: false,
+                                     inst_rpms: unitPackages(target: 'el9') + ' daos-client-tests',
+                                     image_version: 'el9.7',
+                                     prov_env_vars: 'VM_CPUS=14'))
                     }
                     post {
                         always {
+                            unitTestPost artifacts: ['nlt_logs/'],
+                                         testResults: 'nlt-junit.xml',
+                                         with_valgrind: '',
+                                         FI: true
                             discoverGitReferenceBuild referenceJob: 'daos-stack/daos/master',
                                                       scm: 'daos-stack/daos',
                                                       requiredResult: hudson.model.Result.UNSTABLE
-                            recordIssues enabledForFailure: true,
-                                         /* ignore warning/errors from PMDK logging system */
-                                         filters: [excludeFile('pmdk/.+')],
-                                         failOnError: false,
-                                         ignoreQualityGate: true,
-                                         qualityGates: [[threshold: 1, type: 'TOTAL_ERROR'],
-                                                        [threshold: 1, type: 'TOTAL_HIGH'],
-                                                        [threshold: 1, type: 'NEW_NORMAL', unstable: true],
-                                                        [threshold: 1, type: 'NEW_LOW', unstable: true]],
-                                         tools: [issues(pattern: 'nlt-errors.json',
-                                                        name: 'Fault injection issues',
-                                                        id: 'Fault_Injection'),
-                                                 issues(pattern: 'nlt-client-leaks.json',
-                                                        name: 'Fault injection leaks',
-                                                        id: 'NLT_client')],
-                                         scm: 'daos-stack/daos'
-                            junit testResults: 'nlt-junit.xml'
-                            stash name: 'fault-inject-valgrind',
-                                  includes: '*.memcheck.xml',
-                                  allowEmpty: true
                             archiveArtifacts artifacts: 'nlt_logs/fault-injection/',
                                              allowEmptyArchive: true
                             job_status_update()
@@ -1261,8 +1223,7 @@ pipeline {
     post {
         always {
             valgrindReportPublish valgrind_stashes: ['nlt-memcheck',
-                                                     'unit-memcheck',
-                                                     'fault-inject-valgrind']
+                                                     'unit-memcheck']
             job_status_update('final_status')
             jobStatusWrite(job_status_internal)
         }

ci/unit/test_nlt.sh

@@ -13,7 +13,7 @@ mydir="$(cd "$(dirname "${BASH_SOURCE[0]}")" >/dev/null 2>&1 && pwd)"
 # Copy over the install tree and some of the build tree.
 rsync -rlpt -z -e "ssh $SSH_KEY_ARGS" .build_vars* opt-daos.tar utils requirements-utest.txt jenkins@"$NODE":build/
 
-# shellcheck disable=SC2029
-ssh -tt "$SSH_KEY_ARGS" jenkins@"$NODE" "DAOS_HTTPS_PROXY=\"${DAOS_HTTPS_PROXY:-}\" \
-                                         DAOS_NO_PROXY=\"${DAOS_NO_PROXY:-}\" \
-                                         $(cat "$mydir/test_nlt_node.sh")"
+ssh -T "$SSH_KEY_ARGS" jenkins@"$NODE" \
+    "DAOS_HTTPS_PROXY=\"${DAOS_HTTPS_PROXY:-}\" \
+     DAOS_NO_PROXY=\"${DAOS_NO_PROXY:-}\" \
+     bash -s -- $*" < "$mydir/test_nlt_node.sh"

ci/unit/test_nlt_node.sh

@@ -6,7 +6,6 @@
 set -uex
 
 sudo bash -c 'echo 1 > /proc/sys/kernel/sysrq'
-sudo mkdir -p /mnt/daos
 # using mmap()'ed ULT stacks requires to bump system default
 if [ "$(sudo sysctl -n vm.max_map_count)" -lt "1000000" ] ; then
     sudo sysctl vm.max_map_count=1000000
@@ -46,7 +45,11 @@ pip install /opt/daos/lib/daos/python/
 sudo prlimit --nofile=1024:262144 --pid $$
 prlimit -n
 
-HTTPS_PROXY="${DAOS_HTTPS_PROXY:-}" \
+mkdir -p nlt_logs
+sudo mount -t tmpfs tmpfs nlt_logs
+sudo chown jenkins:jenkins nlt_logs
+
+TMPDIR="$(pwd)/nlt_logs" \
+    HTTPS_PROXY="${DAOS_HTTPS_PROXY:-}" \
     NO_PROXY="${DAOS_NO_PROXY:-}" \
-    ./utils/node_local_test.py --max-log-size 1950MiB \
-    --dfuse-dir /localhome/jenkins/ --log-usage-save nltir.xml --log-usage-export nltr.json all
+    exec ./utils/node_local_test.py "$@"

ci/unit/test_nlt_post.sh

@@ -13,14 +13,18 @@ mkdir nlt_logs
 # Copy any log files.  Use rsync filters here to allow us to specify
 # all files we want to copy, as it's much more flexible than using
 # standard wildcards.
-rsync -v -dprt -e "ssh $SSH_KEY_ARGS" jenkins@"$NODE":/tmp/ \
+
+# Assuming that node_local_test.py is run with --class-name,
+# the logs will be in build/nlt_logs/ on the node.
+rsync -v -rlpt -e "ssh $SSH_KEY_ARGS" jenkins@"$NODE":build/nlt_logs/ \
       --filter="include dnt*.log" --filter="include dnt*.log.bz2" \
-      --filter="include dnt_fi_*_logs" \
+      --filter="include dnt_fi_*_logs" --filter="include */" \
       --filter="exclude *" nlt_logs/
 
 rsync -v -dpt -z -e "ssh $SSH_KEY_ARGS" jenkins@"$NODE":build/ \
       --filter="include nlt*.json" --filter="include dnt*.xml" \
       --filter="include nltir.xml" --filter="include nltr.json" \
       --filter="include nlt-junit.xml" --filter="exclude *" ./
+
 mkdir -p vm_test
 mv nlt-errors.json vm_test/

src/tests/ftest/cart/util/cart_logtest.py

@@ -225,7 +225,6 @@ def __init__(self, log_iter, quiet=False):
         self.fi_triggered = False
         self.fi_location = None
         self.skip_suffixes = []
-        self.skip_substrings = []
         self._tracers = []
         self.ftest_mode = False
 
@@ -444,10 +443,6 @@ def _check_pid_from_log_file(self, pid, abort_on_warning, leak_wf, show_memleaks
                         show = False
                     if show and any(map(line.get_msg().endswith, self.skip_suffixes)):
                         show = False
-                    if show:
-                        line_msg = line.get_msg().casefold()
-                        if any(sub in line_msg for sub in self.skip_substrings):
-                            show = False
                     if show:
                         # Allow WARNING or ERROR messages, but anything higher like assert should
                         # trigger a failure.

utils/nlt_server.yaml

@@ -14,7 +14,6 @@ engines:
       - DAOS_MD_CAP=1024
       - DAOS_STRICT_SHUTDOWN=1
       - DAOS_TARGET_OVERSUBSCRIBE=1
-      - ABT_STACK_OVERFLOW_CHECK=mprotect
     storage:
       -
         class: ram

utils/node_local_test.py

@@ -5025,13 +5025,6 @@ def sizeof_fmt(num, suffix='B'):
     if ignore_busy:
         lto.skip_suffixes.append(" DER_BUSY(-1012): 'Device or resource busy'")
 
-    lto.skip_substrings.extend([
-        'sluggish ec boundary report from rank',
-        'sluggish stable epoch reporting',
-        'progress callback was not called for too long',
-        'rpc failed; rc:',
-    ])
-
     try:
         lto.check_log_file(abort_on_warning=True,
                            show_memleaks=show_memleaks,
@@ -5924,7 +5917,7 @@ def _prep(self):
         # pylint: disable-next=no-member
         num_cores = len(os.sched_getaffinity(0))
 
-        if num_cores < 20:
+        if num_cores < 14:
             max_child = 1
         else:
             max_child = int(num_cores / 4 * 3)
@@ -6751,12 +6744,14 @@ def run(wf, args):
     run_fi = False
 
     if args.perf_check or fi_test or fi_test_dfuse:
-        fs = subprocess.run([os.path.join(conf['PREFIX'], 'bin', 'fault_status')], check=False)
+        fi_env = os.environ.copy()
+        fi_env['PATH'] = f'{conf["PREFIX"]}/bin:{fi_env["PATH"]}'
+        fs = subprocess.run(['fault_status'], check=False, env=fi_env)
         print(fs)
         if fs.returncode == 0:
             run_fi = True
         else:
-            print("Unable to detect fault injection feature, skipping testing")
+            print("Unable to detect fault injection feature - skipping FI testing")
 
     if run_fi:
         args.server_debug = 'INFO'