Merge branch 'master' into lujieduan-remove-centos7-gpu

dev-docs/dev.md

@@ -158,14 +158,14 @@ $ tree $CONFIG_OUT
 ```
 
 *   Sample generated
-    [golden fluent bit main conf](https://github.com/GoogleCloudPlatform/ops-agent/blob/master/confgenerator/testdata/valid/linux/default_config/golden_fluent_bit_main.conf)
+    [golden fluent bit main conf](https://github.com/GoogleCloudPlatform/ops-agent/blob/master/confgenerator/testdata/goldens/builtin/golden/linux/fluent_bit_main.conf)
     at `$CONFIG_OUT/fluent_bit_main.conf`.
 *   Sample generated
-    [golden fluent bit parser conf](https://github.com/GoogleCloudPlatform/ops-agent/blob/master/confgenerator/testdata/valid/linux/default_config/golden_fluent_bit_parser.conf)
+    [golden fluent bit parser conf](https://github.com/GoogleCloudPlatform/ops-agent/blob/master/confgenerator/testdata/goldens/builtin/golden/linux/fluent_bit_parser.conf)
     at `$CONFIG_OUT/fluent_bit_parser.conf`.
 *   Sample generated
-    [golden otel conf](https://github.com/GoogleCloudPlatform/ops-agent/blob/master/confgenerator/testdata/valid/linux/default_config/golden_otel.conf)
-    at `$CONFIG_OUT/otel.conf`.
+    [golden otel yaml](https://github.com/GoogleCloudPlatform/ops-agent/blob/master/confgenerator/testdata/goldens/builtin/golden/linux/otel.yaml)
+    at `$CONFIG_OUT/otel.yaml`.
 
 ## Build and test manually on GCE VMs
 

integration_test/gce/gce_testing.go

@@ -1101,6 +1101,25 @@ func addFrameworkMetadata(imageSpec string, inputMetadata map[string]string) (ma
 		if _, ok := metadataCopy["startup-script"]; ok {
 			return nil, errors.New("the 'startup-script' metadata key is reserved for future use. Instead, wait for the instance to be ready and then run things with RunRemotely() or RunScriptRemotely()")
 		}
+		// TODO(b/380470389): we actually *can't* do RunRemotely() on DLVM images due to a bug.
+		// The workaround for the bug is to deploy a fix in-VM via startup scripts.
+		if strings.Contains(imageSpec, "common-gpu-debian-11-py310") {
+			metadataCopy["startup-script"] = fmt.Sprintf(`
+#!/bin/bash
+# Give time for the guest agent and jupyter stuff to finish modifying
+# /etc/passwd and test_user home directory
+sleep 120
+HOMEDIR=/home/%[1]s
+SSHFILE=$HOMEDIR/.ssh/authorized_keys
+if [ ! -f "$SSHFILE" ]; then
+  sudo mkdir -p "$HOMEDIR/.ssh"
+  sudo touch "$SSHFILE"
+fi
+sudo chown -R %[1]s:%[1]s "$HOMEDIR"
+sudo chmod 600 "$SSHFILE"`,
+				sshUserName,
+			)
+		}
 	}
 	return metadataCopy, nil
 }

transformation_test/transformation_test.go

@@ -444,9 +444,24 @@ func (transformationConfig transformationTest) runOTelTestInner(t *testing.T, na
 	}
 
 	var errors []any
+	var exitErr error
 
 	// Read from stderr until EOF and put any errors in `errors`.
 	eg.Go(func() error {
+		// Wait for the process to exit.
+		defer eg.Go(func() error {
+			if err := cmd.Wait(); err != nil {
+				if _, ok := err.(*exec.ExitError); ok {
+					exitErr = err
+					t.Logf("process terminated with error: %v", err)
+				} else {
+					return fmt.Errorf("process failed: %w", err)
+				}
+			}
+			cancel()
+			return nil
+		})
+
 		consumingCount := 0
 		r := bufio.NewReader(stderr)
 		d := json.NewDecoder(r)
@@ -498,6 +513,7 @@ func (transformationConfig transformationTest) runOTelTestInner(t *testing.T, na
 			}
 		}
 	})
+
 	// Read and sanitize requests.
 	eg.Go(func() error {
 		for r := range requestCh {
@@ -506,21 +522,6 @@ func (transformationConfig transformationTest) runOTelTestInner(t *testing.T, na
 		return nil
 	})
 
-	var exitErr error
-	// Wait for the process to exit.
-	eg.Go(func() error {
-		if err := cmd.Wait(); err != nil {
-			if _, ok := err.(*exec.ExitError); ok {
-				exitErr = err
-				t.Logf("process terminated with error: %v", err)
-			} else {
-				return fmt.Errorf("process failed: %w", err)
-			}
-		}
-		cancel()
-		return nil
-	})
-
 	if err := eg.Wait(); err != nil {
 		t.Errorf("errgroup failed: %v", err)
 	}