Backport PowerVM+shfmt+gperf fixes on 3.21 (4.7) (#2069)

Co-authored-by: Afsan Hossain <[email protected]>
Co-authored-by: Mauro Ezequiel Moltrasio <[email protected]>

Author: 3 people

.github/workflows/lint.yml

@@ -31,7 +31,7 @@ jobs:
 
       - name: Install shfmt
         run: |
-          GOBIN="${HOME}/bin/" go install mvdan.cc/sh/v3/cmd/shfmt@latest
+          GOBIN="${HOME}/bin/" go install mvdan.cc/sh/v3/cmd/shfmt@v3.10.0
 
       - name: Install actionlint
         run: |

ansible/requirements.yml

@@ -4,5 +4,6 @@ collections:
 - community.general
 - community.docker
 - containers.podman
-- ibm.cloudcollection
 - kubernetes.core
+- name: ibm.cloudcollection
+  version: 1.51.0

collector/lib/ProfilerHandler.cpp

@@ -13,7 +13,10 @@
 
 namespace collector {
 
-const std::string ProfilerHandler::kCPUProfileFilename = "/module/cpu_profile";
+// The following route is used by the CPU profiler to dump information
+// into. When running in read only file systems, /var/profiles needs to
+// be writable, or the CPU profiler will not work.
+const std::string ProfilerHandler::kCPUProfileFilename = "/var/profiles/cpu_profile";
 const std::string ProfilerHandler::kBaseRoute = "/profile";
 const std::string ProfilerHandler::kCPURoute = kBaseRoute + "/cpu";
 const std::string ProfilerHandler::kHeapRoute = kBaseRoute + "/heap";

docs/troubleshooting.md

@@ -453,6 +453,15 @@ $ curl collector:8080/profile/heap
 The resulting profile could be processed with `pprof` to get a human-readable
 output with debugging symbols.
 
+Collector also exposes a CPU profiler under `/profile/cpu`, which operates in
+a very similar manner to the heap profiler.
+
+---
+**_NOTE_**: If the CPU profiler fails to start, make sure /var/profiles exists
+in the collector container and is writable.
+
+---
+
 ## Benchmark CI step
 
 Whenever in doubt about performance implications of your changes, there is an

integration-tests/integration_test.go

@@ -7,6 +7,7 @@ import (
 	"github.com/stretchr/testify/suite"
 
 	"github.com/stackrox/collector/integration-tests/pkg/collector"
+	"github.com/stackrox/collector/integration-tests/pkg/common"
 	"github.com/stackrox/collector/integration-tests/pkg/config"
 	"github.com/stackrox/collector/integration-tests/pkg/types"
 	"github.com/stackrox/collector/integration-tests/suites"
@@ -516,6 +517,9 @@ func TestPerfEvent(t *testing.T) {
 }
 
 func TestGperftools(t *testing.T) {
+	if ok, arch := common.ArchSupported("x86_64"); !ok {
+		t.Skip("[WARNING]: skip GperftoolsTestSuite on ", arch)
+	}
 	suite.Run(t, new(suites.GperftoolsTestSuite))
 }
 

integration-tests/pkg/collector/collector_docker.go

@@ -42,6 +42,7 @@ func NewDockerCollectorManager(e executor.Executor, name string) *DockerCollecto
 		"/host/usr/lib:ro":          "/usr/lib",
 		"/host/sys/kernel/debug:ro": "/sys/kernel/debug",
 		"/etc/stackrox:ro":          "/tmp/collector-test",
+		"/var/profiles":             "", // gperftools dump directory
 	}
 
 	return &DockerCollectorManager{

integration-tests/suites/gperftools.go

@@ -1,6 +1,7 @@
 package suites
 
 import (
+	"fmt"
 	"net/http"
 	"strings"
 	"time"
@@ -36,30 +37,29 @@ func (s *GperftoolsTestSuite) TearDownSuite() {
 // NOTE: The test will only be performed on supported architectures (only
 // x86_64 at the moment).
 func (s *GperftoolsTestSuite) TestFetchHeapProfile() {
-	if ok, arch := common.ArchSupported("x86_64"); !ok {
-		s.T().Skip("[WARNING]: skip GperftoolsTestSuite on ", arch)
-	}
+	s.runProfilerTest("heap")
+}
 
-	var (
-		response *http.Response
-		err      error
-	)
+func (s *GperftoolsTestSuite) TestFetchCpuProfile() {
+	s.runProfilerTest("cpu")
+}
 
-	heap_api_url := "http://localhost:8080/profile/heap"
+func (s *GperftoolsTestSuite) runProfilerTest(resource string) {
+	heap_api_url := fmt.Sprintf("http://localhost:8080/profile/%s", resource)
 	data_type := "application/x-www-form-urlencoded"
 
-	response, err = http.Post(heap_api_url, data_type, strings.NewReader("on"))
+	response, err := http.Post(heap_api_url, data_type, strings.NewReader("on"))
 	s.Require().NoError(err)
-	s.Assert().Equal(response.StatusCode, 200, "Failed to start heap profiling")
+	s.Assert().Equalf(response.StatusCode, 200, "Failed to start %s profiling", resource)
 
 	// Wait a bit to collect something in the heap profile
 	common.Sleep(1 * time.Second)
 
 	response, err = http.Post(heap_api_url, data_type, strings.NewReader("off"))
 	s.Require().NoError(err)
-	s.Assert().Equal(response.StatusCode, 200, "Failed to stop heap profiling")
+	s.Assert().Equalf(response.StatusCode, 200, "Failed to stop %s profiling", resource)
 
 	response, err = http.Get(heap_api_url)
 	s.Require().NoError(err)
-	s.Assert().Equal(response.StatusCode, 200, "Failed to fetch heap profile")
+	s.Assert().Equalf(response.StatusCode, 200, "Failed to fetch %s profile", resource)
 }