Add hack script to build hive install-log-regexes from ARO definition

hack/genhiveconfig/genhiveconfig.go

@@ -0,0 +1,84 @@
+package main
+
+// Copyright (c) Microsoft Corporation.
+// Licensed under the Apache License 2.0.
+
+import (
+	"context"
+	"os"
+
+	"github.com/ghodss/yaml"
+	corev1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
+	failure "github.com/Azure/ARO-RP/pkg/hive/failure"
+	utillog "github.com/Azure/ARO-RP/pkg/util/log"
+)
+
+const (
+	hiveNamespaceName  = "hive"
+	configMapName      = "additional-install-log-regexes"
+	configMapPath      = "hack/hive-config/hive-additional-install-log-regexes.yaml"
+	regexDataEntryName = "regexes"
+)
+
+type installLogRegex struct {
+	Name                  string   `json:"name"`
+	SearchRegexStrings    []string `json:"searchRegexStrings"`
+	InstallFailingReason  string   `json:"installFailingReason"`
+	InstallFailingMessage string   `json:"installFailingMessage"`
+}
+
+func run(ctx context.Context) error {
+	ilrs := []installLogRegex{}
+
+	for _, reason := range failure.Reasons {
+		ilrs = append(ilrs, failureReasonToInstallLogRegex(reason))
+	}
+
+	ilrsRaw, err := yaml.Marshal(ilrs)
+	if err != nil {
+		return err
+	}
+
+	configmap := &corev1.ConfigMap{
+		TypeMeta: metav1.TypeMeta{
+			APIVersion: corev1.SchemeGroupVersion.String(),
+			Kind:       "ConfigMap",
+		},
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: hiveNamespaceName,
+			Name:      configMapName,
+		},
+		Data: map[string]string{
+			regexDataEntryName: string(ilrsRaw),
+		},
+	}
+
+	configmapRaw, err := yaml.Marshal(configmap)
+	if err != nil {
+		return err
+	}
+	return os.WriteFile(configMapPath, configmapRaw, 0666)
+}
+
+func failureReasonToInstallLogRegex(reason failure.InstallFailingReason) installLogRegex {
+	ilr := installLogRegex{
+		Name:                  reason.Name,
+		InstallFailingReason:  reason.Reason,
+		InstallFailingMessage: reason.Message,
+		SearchRegexStrings:    []string{},
+	}
+	for _, regex := range reason.SearchRegexes {
+		ilr.SearchRegexStrings = append(ilr.SearchRegexStrings, regex.String())
+	}
+	return ilr
+}
+
+func main() {
+	log := utillog.GetLogger()
+
+	if err := run(context.Background()); err != nil {
+		log.Fatal(err)
+	}
+}

hack/genhiveconfig/genhiveconfig_test.go

@@ -0,0 +1,37 @@
+package main
+
+// Copyright (c) Microsoft Corporation.
+// Licensed under the Apache License 2.0.
+
+import (
+	"reflect"
+	"regexp"
+	"testing"
+
+	"github.com/Azure/ARO-RP/pkg/hive/failure"
+)
+
+func TestFailureReasonToInstallLogRegex(t *testing.T) {
+	input := failure.InstallFailingReason{
+		Name:    "TestReason",
+		Reason:  "AzureTestReason",
+		Message: "This is a sentence.",
+		SearchRegexes: []*regexp.Regexp{
+			regexp.MustCompile(".*"),
+			regexp.MustCompile("^$"),
+		},
+	}
+
+	want := installLogRegex{
+		Name:                  "TestReason",
+		InstallFailingReason:  "AzureTestReason",
+		InstallFailingMessage: "This is a sentence.",
+		SearchRegexStrings:    []string{".*", "^$"},
+	}
+
+	got := failureReasonToInstallLogRegex(input)
+
+	if !reflect.DeepEqual(got, want) {
+		t.Errorf("got %v, want %v", got, want)
+	}
+}

hack/hive-config/hive-additional-install-log-regexes.yaml

@@ -0,0 +1,14 @@
+apiVersion: v1
+data:
+  regexes: |
+    - installFailingMessage: The template deployment failed. Please see details for more
+        information.
+      installFailingReason: AzureInvalidTemplateDeployment
+      name: AzureInvalidTemplateDeployment
+      searchRegexStrings:
+      - '"code":\w?"InvalidTemplateDeployment"'
+kind: ConfigMap
+metadata:
+  creationTimestamp: null
+  name: additional-install-log-regexes
+  namespace: hive

pkg/hive/failure/reasons.go

@@ -0,0 +1,38 @@
+package failure
+
+// Copyright (c) Microsoft Corporation.
+// Licensed under the Apache License 2.0.
+
+import "regexp"
+
+type InstallFailingReason struct {
+	Name          string
+	Reason        string
+	Message       string
+	SearchRegexes []*regexp.Regexp
+}
+
+var Reasons = []InstallFailingReason{
+	// Order within this array determines precedence. Earlier entries will take
+	// priority over later ones.
+	AzureRequestDisallowedByPolicy,
+	AzureInvalidTemplateDeployment,
+}
+
+var AzureRequestDisallowedByPolicy = InstallFailingReason{
+	Name:    "AzureRequestDisallowedByPolicy",
+	Reason:  "AzureRequestDisallowedByPolicy",
+	Message: "Cluster Deployment was disallowed by policy.  Please see install log for more information.",
+	SearchRegexes: []*regexp.Regexp{
+		regexp.MustCompile(`"code":\w?"InvalidTemplateDeployment".*"code":\w?"RequestDisallowedByPolicy"`),
+	},
+}
+
+var AzureInvalidTemplateDeployment = InstallFailingReason{
+	Name:    "AzureInvalidTemplateDeployment",
+	Reason:  "AzureInvalidTemplateDeployment",
+	Message: "The template deployment failed. Please see install log for more information.",
+	SearchRegexes: []*regexp.Regexp{
+		regexp.MustCompile(`"code":\w?"InvalidTemplateDeployment"`),
+	},
+}

pkg/hive/failure/reasons_test.go

@@ -0,0 +1,95 @@
+package failure
+
+// Copyright (c) Microsoft Corporation.
+// Licensed under the Apache License 2.0.
+
+import (
+	"reflect"
+	"regexp"
+	"testing"
+)
+
+func TestInstallFailingReasonRegexes(t *testing.T) {
+	for _, tt := range []struct {
+		name       string
+		installLog string
+		want       InstallFailingReason
+	}{
+		{
+			name: "InvalidTemplateDeployment - no known errors",
+			installLog: `
+level=info msg=running in local development mode
+level=info msg=creating development InstanceMetadata
+level=info msg=InstanceMetadata: running on AzurePublicCloud
+level=info msg=running step [Action github.com/openshift/ARO-Installer/pkg/installer.(*manager).Manifests.func1]
+level=info msg=running step [Action github.com/openshift/ARO-Installer/pkg/installer.(*manager).Manifests.func2]
+level=info msg=resolving graph
+level=info msg=running step [Action github.com/openshift/ARO-Installer/pkg/installer.(*manager).Manifests.func3]
+level=info msg=checking if graph exists
+level=info msg=save graph
+Generates the Ignition Config asset
+
+level=info msg=running in local development mode
+level=info msg=creating development InstanceMetadata
+level=info msg=InstanceMetadata: running on AzurePublicCloud
+level=info msg=running step [AuthorizationRetryingAction github.com/openshift/ARO-Installer/pkg/installer.(*manager).deployResourceTemplate-fm]
+level=info msg=load persisted graph
+level=info msg=deploying resources template
+level=error msg=step [AuthorizationRetryingAction github.com/openshift/ARO-Installer/pkg/installer.(*manager).deployResourceTemplate-fm] encountered error: 400: DeploymentFailed: : Deployment failed. Details: : : {"code":"InvalidTemplateDeployment","message":"The template deployment failed with multiple errors. Please see details for more information.","details":[]}
+level=error msg=400: DeploymentFailed: : Deployment failed. Details: : : {"code":"InvalidTemplateDeployment","message":"The template deployment failed with multiple errors. Please see details for more information.","details":[]}`,
+			want: AzureInvalidTemplateDeployment,
+		},
+		{
+			name: "InvalidTemplateDeployment - RequestDisallowedByPolicy",
+			installLog: `
+level=info msg=running in local development mode
+level=info msg=creating development InstanceMetadata
+level=info msg=InstanceMetadata: running on AzurePublicCloud
+level=info msg=running step [Action github.com/openshift/ARO-Installer/pkg/installer.(*manager).Manifests.func1]
+level=info msg=running step [Action github.com/openshift/ARO-Installer/pkg/installer.(*manager).Manifests.func2]
+level=info msg=resolving graph
+level=info msg=running step [Action github.com/openshift/ARO-Installer/pkg/installer.(*manager).Manifests.func3]
+level=info msg=checking if graph exists
+level=info msg=save graph
+Generates the Ignition Config asset
+
+level=info msg=running in local development mode
+level=info msg=creating development InstanceMetadata
+level=info msg=InstanceMetadata: running on AzurePublicCloud
+level=info msg=running step [AuthorizationRetryingAction github.com/openshift/ARO-Installer/pkg/installer.(*manager).deployResourceTemplate-fm]
+level=info msg=load persisted graph
+level=info msg=deploying resources template
+level=error msg=step [AuthorizationRetryingAction github.com/openshift/ARO-Installer/pkg/installer.(*manager).deployResourceTemplate-fm] encountered error: 400: DeploymentFailed: : Deployment failed. Details: : : {"code":"InvalidTemplateDeployment","message":"The template deployment failed with multiple errors. Please see details for more information.","details":[{"additionalInfo":[],"code":"RequestDisallowedByPolicy","message":"Resource 'test-bootstrap' was disallowed by policy. Policy identifiers: ''.","target":"test-bootstrap"}]}
+level=error msg=400: DeploymentFailed: : Deployment failed. Details: : : {"code":"InvalidTemplateDeployment","message":"The template deployment failed with multiple errors. Please see details for more information.","details":[{"additionalInfo":[],"code":"RequestDisallowedByPolicy","message":"Resource 'test-bootstrap' was disallowed by policy. Policy identifiers: ''.","target":"test-bootstrap"}]}`,
+			want: AzureRequestDisallowedByPolicy,
+		},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			// This test uses a "mock" version of Hive's real implementation for matching install logs against regex patterns.
+			// https://github.com/bennerv/hive/blob/fec14dcf0-plus-base-image-update/pkg/controller/clusterprovision/installlogmonitor.go#L83
+			// The purpose of this test is to test the regular expressions themselves, not the implementation.
+			got := mockHiveIdentifyReason(tt.installLog)
+
+			if !reflect.DeepEqual(got, tt.want) {
+				t.Errorf("got %v, want %v", got, tt.want)
+			}
+		})
+	}
+}
+
+func mockHiveIdentifyReason(installLog string) InstallFailingReason {
+	for _, reason := range Reasons {
+		for _, regex := range reason.SearchRegexes {
+			if regex.MatchString(installLog) {
+				return reason
+			}
+		}
+	}
+
+	return InstallFailingReason{
+		Name:          "UnknownError",
+		Reason:        "UnknownError",
+		Message:       installLog,
+		SearchRegexes: []*regexp.Regexp{},
+	}
+}