[release-4.14] OCPBUGS-26003: E2E: PPC Test cases (#905)

* Add variables for mustgather and NTO image Adds to new environment variables NTO_IMAGE_REGISTRY and MUSGATHER_DIR for providing NTO image and mustgather directory respectively Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * gexec vendor updates Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * E2E: PPC test cases PPC Test cases executed through podman Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * Include E2E ppc to be executed as part of functional tests Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * Removed unused variables and minor review fixes Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * minor gofmt fixes Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * Remove hardcoded values of reserved and isolated Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * Remove unused struct and rename NtoImageRegistry as NtoImage Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * use cpuset to get length of cpus Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> * update the NTOImage variable Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> --------- Signed-off-by: Niranjan M.R <mrniranjan@redhat.com> Co-authored-by: Niranjan M.R <mrniranjan@redhat.com>
openshift · Jan 9, 2024 · 6d4553b · 6d4553b
1 parent c2dffde
commit 6d4553b
Show file tree

Hide file tree

Showing 12 changed files with 1,381 additions and 1 deletion.
diff --git a/Makefile b/Makefile
@@ -201,7 +201,7 @@ pao-functests: cluster-label-worker-cnf pao-functests-only
 pao-functests-only:
 	@echo "Cluster Version"
 	hack/show-cluster-version.sh
-	hack/run-test.sh -t "test/e2e/performanceprofile/functests/0_config test/e2e/performanceprofile/functests/1_performance test/e2e/performanceprofile/functests/6_mustgather_testing" -p "-v -r --fail-fast  --flake-attempts=2 --junit-report=report.xml" -m "Running Functional Tests"
+	hack/run-test.sh -t "test/e2e/performanceprofile/functests/0_config test/e2e/performanceprofile/functests/1_performance test/e2e/performanceprofile/functests/6_mustgather_testing test/e2e/performanceprofile/functests/10_performance_ppc" -p "-v -r --fail-fast  --flake-attempts=2 --junit-report=report.xml" -m "Running Functional Tests"
 
 .PHONY: pao-functests-updating-profile
 pao-functests-updating-profile: cluster-label-worker-cnf pao-functests-update-only

diff --git a/test/e2e/performanceprofile/functests/10_performance_ppc/10_ppc_suite_test.go b/test/e2e/performanceprofile/functests/10_performance_ppc/10_ppc_suite_test.go
@@ -0,0 +1,39 @@
+package __performance_ppc
+
+import (
+	"fmt"
+	"os/exec"
+	"testing"
+
+	. "github.com/onsi/ginkgo/v2"
+
+	. "github.com/onsi/gomega"
+	testutils "github.com/openshift/cluster-node-tuning-operator/test/e2e/performanceprofile/functests/utils"
+	testlog "github.com/openshift/cluster-node-tuning-operator/test/e2e/performanceprofile/functests/utils/log"
+)
+
+const (
+	DefaultPodmanBinaryPath = "/usr/bin/podman"
+)
+
+var _ = BeforeSuite(func() {
+	var err error
+	By("Check podman binary exists")
+	path, err := exec.LookPath(DefaultPodmanBinaryPath)
+	if err != nil {
+		Skip(fmt.Sprintf("%s doesn't exists", DefaultPodmanBinaryPath))
+	}
+	testlog.Infof("Podman binary executed from path %s", path)
+	By("Checking Environment Variables")
+	testlog.Infof("NTO Image used: %s", testutils.NTOImage)
+	if testutils.MustGatherDir == "" {
+		Skip("set env variable MUSTGATHER_DIR to ocp mustgather directory")
+	}
+	testlog.Infof("Mustgather Directory used: %s", testutils.MustGatherDir)
+
+})
+
+func TestPPC(t *testing.T) {
+	RegisterFailHandler(Fail)
+	RunSpecs(t, "PPC Suite")
+}
diff --git a/test/e2e/performanceprofile/functests/10_performance_ppc/ppc.go b/test/e2e/performanceprofile/functests/10_performance_ppc/ppc.go
@@ -0,0 +1,198 @@
+package __performance_ppc
+
+import (
+	"fmt"
+	"os/exec"
+	"regexp"
+	"strings"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+	"github.com/onsi/gomega/gexec"
+	performancev2 "github.com/openshift/cluster-node-tuning-operator/pkg/apis/performanceprofile/v2"
+	testutils "github.com/openshift/cluster-node-tuning-operator/test/e2e/performanceprofile/functests/utils"
+	testlog "github.com/openshift/cluster-node-tuning-operator/test/e2e/performanceprofile/functests/utils/log"
+	"k8s.io/kubernetes/pkg/kubelet/cm/cpuset"
+	"sigs.k8s.io/yaml"
+)
+
+type PPCTest struct {
+	PodmanMakeOptions func(args []string) []string
+	PodmanBinary      string
+}
+
+type PPCSession struct {
+	*gexec.Session
+}
+
+func (p *PPCTest) MakeOptions(args []string) []string {
+	return p.PodmanMakeOptions(args)
+}
+
+func (p *PPCTest) PodmanAsUserBase(args []string, noEvents, noCache bool) (*PPCSession, error) {
+	var command *exec.Cmd
+	podmanOptions := p.MakeOptions(args)
+	podmanBinary := p.PodmanBinary
+	fmt.Printf("Running: %s %s\n", podmanBinary, strings.Join(podmanOptions, " "))
+	command = exec.Command(podmanBinary, podmanOptions...)
+	session, err := gexec.Start(command, GinkgoWriter, GinkgoWriter)
+	if err != nil {
+		return nil, err
+	}
+
+	return &PPCSession{session}, nil
+}
+
+type PPCTestIntegration struct {
+	PPCTest
+}
+
+func (p *PPCTestIntegration) MakeOptions(args []string) []string {
+	return args
+}
+func PPCTestCreateUtil() *PPCTestIntegration {
+	p := &PPCTestIntegration{
+		PPCTest: PPCTest{
+			PodmanBinary: DefaultPodmanBinaryPath,
+		},
+	}
+	p.PodmanMakeOptions = p.MakeOptions
+	return p
+}
+
+var _ = Describe("[rfe_id: 38968] PerformanceProfile setup helper and platform awareness", func() {
+	mustgatherDir := testutils.MustGatherDir
+	ntoImage := testutils.NTOImage
+	Context("PPC Sanity Tests", func() {
+		ppcIntgTest := PPCTestCreateUtil()
+		It("[test_id:40940] Performance Profile regression tests", func() {
+			pp := &performancev2.PerformanceProfile{}
+			var reservedCpuCount = 2
+			defaultArgs := []string{
+				"run",
+				"--entrypoint",
+				"performance-profile-creator",
+				"-v",
+			}
+
+			cmdArgs := []string{
+				fmt.Sprintf("%s:%s:z", mustgatherDir, mustgatherDir),
+				ntoImage,
+				"--mcp-name=worker",
+				fmt.Sprintf("--reserved-cpu-count=%d", reservedCpuCount),
+				fmt.Sprintf("--rt-kernel=%t", true),
+				fmt.Sprintf("--power-consumption-mode=%s", "low-latency"),
+				fmt.Sprintf("--split-reserved-cpus-across-numa=%t", false),
+				fmt.Sprintf("--must-gather-dir-path=%s", mustgatherDir),
+			}
+			podmanArgs := []string{}
+			podmanArgs = append(defaultArgs, cmdArgs...)
+			session, err := ppcIntgTest.PodmanAsUserBase(podmanArgs, false, false)
+			Expect(err).ToNot(HaveOccurred(), "Podman command failed")
+			output := session.Wait(20).Out.Contents()
+			err = yaml.Unmarshal(output, pp)
+			Expect(err).ToNot(HaveOccurred(), "Unable to marshal the ppc output")
+			reservedCpus, err := cpuset.Parse(string(*pp.Spec.CPU.Reserved))
+			Expect(err).ToNot(HaveOccurred(), "Unable to parse cpus")
+			totalReservedCpus := reservedCpus.Size()
+			Expect(totalReservedCpus).To(Equal(reservedCpuCount))
+			Expect(*pp.Spec.RealTimeKernel.Enabled).To(BeTrue())
+			Expect(*pp.Spec.WorkloadHints.RealTime).To(BeTrue())
+			Expect(*pp.Spec.NUMA.TopologyPolicy).To(Equal("restricted"))
+			Eventually(session).Should(gexec.Exit(0))
+		})
+		It("[test_id:41405] Verify PPC script fails when the splitting of reserved cpus and single numa-node policy is specified", func() {
+			defaultArgs := []string{
+				"run",
+				"--entrypoint",
+				"performance-profile-creator",
+				"-v",
+			}
+			cmdArgs := []string{
+				fmt.Sprintf("%s:%s:z", mustgatherDir, mustgatherDir),
+				ntoImage,
+				"--mcp-name=worker",
+				fmt.Sprintf("--reserved-cpu-count=%d", 2),
+				fmt.Sprintf("--rt-kernel=%t", true),
+				fmt.Sprintf("--power-consumption-mode=%s", "low-latency"),
+				fmt.Sprintf("--split-reserved-cpus-across-numa=%t", true),
+				fmt.Sprintf("--topology-manager-policy=%s", "single-numa-node"),
+				fmt.Sprintf("--must-gather-dir-path=%s", mustgatherDir),
+			}
+			podmanArgs := []string{}
+			podmanArgs = append(defaultArgs, cmdArgs...)
+			session, err := ppcIntgTest.PodmanAsUserBase(podmanArgs, false, false)
+			Expect(err).ToNot(HaveOccurred(), "Podman command failed")
+			output := session.Wait(20).Err.Contents()
+			errString := "Error: failed to obtain data from flags not appropriate to split reserved CPUs in case of topology-manager-policy: single-numa-node"
+			ok, err := regexp.MatchString(errString, string(output))
+			if ok {
+				testlog.Info(errString)
+			}
+			Eventually(session).Should(gexec.Exit(1))
+		})
+
+		It("[test_id:41419] Verify PPC script fails when reserved cpu count is 2 and requires to split across numa nodes", func() {
+			defaultArgs := []string{
+				"run",
+				"--entrypoint",
+				"performance-profile-creator",
+				"-v",
+			}
+			cmdArgs := []string{
+				fmt.Sprintf("%s:%s:z", mustgatherDir, mustgatherDir),
+				ntoImage,
+				"--mcp-name=worker",
+				fmt.Sprintf("--reserved-cpu-count=%d", 2),
+				fmt.Sprintf("--rt-kernel=%t", true),
+				fmt.Sprintf("--power-consumption-mode=%s", "low-latency"),
+				fmt.Sprintf("--split-reserved-cpus-across-numa=%t", true),
+				fmt.Sprintf("--must-gather-dir-path=%s", mustgatherDir),
+			}
+			podmanArgs := []string{}
+			podmanArgs = append(defaultArgs, cmdArgs...)
+			session, err := ppcIntgTest.PodmanAsUserBase(podmanArgs, false, false)
+			Expect(err).ToNot(HaveOccurred(), "Podman command failed")
+			output := session.Wait(20).Err.Contents()
+			errString := "Error: failed to compute the reserved and isolated CPUs: can't allocate odd number of CPUs from a NUMA Node"
+			ok, err := regexp.MatchString(errString, string(output))
+			Expect(err).ToNot(HaveOccurred(), "did not fail with Expected:%s failure", errString)
+			if ok {
+				testlog.Info(errString)
+			}
+			Eventually(session).Should(gexec.Exit(1))
+		})
+
+		It("[test_id:41420] Verify PPC script fails when reserved cpu count is more than available cpus", func() {
+			defaultArgs := []string{
+				"run",
+				"--entrypoint",
+				"performance-profile-creator",
+				"-v",
+			}
+			cmdArgs := []string{
+				fmt.Sprintf("%s:%s:z", mustgatherDir, mustgatherDir),
+				ntoImage,
+				"--mcp-name=worker",
+				fmt.Sprintf("--reserved-cpu-count=%d", 100),
+				fmt.Sprintf("--rt-kernel=%t", true),
+				fmt.Sprintf("--power-consumption-mode=%s", "low-latency"),
+				fmt.Sprintf("--split-reserved-cpus-across-numa=%t", true),
+				fmt.Sprintf("--must-gather-dir-path=%s", mustgatherDir),
+			}
+			podmanArgs := []string{}
+			podmanArgs = append(defaultArgs, cmdArgs...)
+			session, err := ppcIntgTest.PodmanAsUserBase(podmanArgs, false, false)
+			Expect(err).ToNot(HaveOccurred(), "Podman command failed")
+			output := session.Wait(20).Err.Contents()
+			errString := "Error: failed to compute the reserved and isolated CPUs: please specify the reserved CPU count in the range [1,3]"
+			ok, err := regexp.MatchString(errString, string(output))
+			Expect(err).ToNot(HaveOccurred(), "did not fail with Expected:%s failure", errString)
+			if ok {
+				testlog.Info(errString)
+			}
+			Eventually(session).Should(gexec.Exit(1))
+		})
+
+	})
+})
diff --git a/test/e2e/performanceprofile/functests/utils/consts.go b/test/e2e/performanceprofile/functests/utils/consts.go
@@ -24,6 +24,12 @@ var NodesSelector string
 // ProfileNotFound is true when discovery mode is enabled and no valid profile was found
 var ProfileNotFound bool
 
+// NtoImage represents NTO Image location which is either quay.io or any other internal registry
+var NTOImage string
+
+// MustGatherDir represents Mustgather directory created using oc adm mustgather
+var MustGatherDir string
+
 func init() {
 	RoleWorkerCNF = os.Getenv("ROLE_WORKER_CNF")
 	if RoleWorkerCNF == "" {
@@ -41,6 +47,14 @@ func init() {
 		fmt.Sprintf("%s/%s", LabelRole, RoleWorkerCNF): "",
 	}
 
+	NTOImage = os.Getenv("NTO_IMAGE")
+
+	if NTOImage == "" {
+		NTOImage = "quay.io/openshift/origin-cluster-node-tuning-operator:latest"
+	}
+
+	MustGatherDir = os.Getenv("MUSTGATHER_DIR")
+
 	if discovery.Enabled() {
 		profile, err := discovery.GetDiscoveryPerformanceProfile(NodesSelector)
 		if err == discovery.ErrProfileNotFound {