feat: notify template owner on manual build failures

coder · BrunoQuaresma · Aug 13, 2024 · Aug 14, 2024 · Aug 14, 2024 · Aug 15, 2024
commit f673b57bea27d3df5ad472a96b122b9a0afe347e
diff --git a/coderd/database/migrations/000243_notifications_template_manual_build_failure.down.sql b/coderd/database/migrations/000243_notifications_template_manual_build_failure.down.sql
diff --git a/coderd/database/migrations/000243_notifications_template_manual_build_failure.up.sql b/coderd/database/migrations/000243_notifications_template_manual_build_failure.up.sql
@@ -0,0 +1,26 @@
+INSERT INTO
+    notification_templates (
+        id,
+        name,
+        title_template,
+        body_template,
+        "group",
+        actions
+    )
+VALUES (
+        '48a9d2b9-3655-430c-a31a-2442479e7519',
+        'Template Manual Build Failure',
+        E'Workspace with template "{{.Labels.name}}" failed to build',
+        E'Hi {{.UserName}}\n\nThe workspace **{{.Labels.workspaceName}}**, using the template **{{.Labels.name}}**, failed during a manual build({{.Labels.transition}}) initiated by the user **{{.Labels.initiator}}**.',
+        'Template Events',
+        '[
+        {
+            "label": "View build",
+            "url": "{{ base_url }}/@{{.Labels.workspaceUserName}}/{{.Labels.workspaceName}}/builds/{{.Labels.buildNumber}}"
+        },
+		{
+            "label": "View Template",
+            "url": "{{ base_url }}/templates/{{.Labels.name}}"
+        }
+    ]'::jsonb
+    );
diff --git a/coderd/notifications/events.go b/coderd/notifications/events.go
@@ -19,3 +19,8 @@ var (
 	TemplateUserAccountCreated = uuid.MustParse("4e19c0ac-94e1-4532-9515-d1801aa283b2")
 	TemplateUserAccountDeleted = uuid.MustParse("f44d9314-ad03-4bc8-95d0-5cad491da6b6")
 )
+
+// Template-related events.
+var (
+	TemplateTemplateManualBuildFailed = uuid.MustParse("48a9d2b9-3655-430c-a31a-2442479e7519")
+)
diff --git a/coderd/provisionerdserver/provisionerdserver.go b/coderd/provisionerdserver/provisionerdserver.go
@@ -1020,7 +1020,25 @@
 			return nil, err
 		}
 
-		s.notifyWorkspaceBuildFailed(ctx, workspace, build)
+		manualBuild := build.Reason.Valid() && build.Reason == database.BuildReasonInitiator
+		if !manualBuild {
+			s.notifyWorkspaceAutoBuildFailed(ctx, workspace, build)
+		} else {
+			template, err := s.Database.GetTemplateByID(ctx, workspace.TemplateID)
+			if err != nil {
+				return nil, xerrors.Errorf("get template to notify manual build failed: %w", err)
+			}
+
+			owner, err := s.Database.GetUserByID(ctx, workspace.OwnerID)
+			if err != nil {
+				return nil, xerrors.Errorf("get owner to notify manual build failed: %w", err)
+			}
+
+			// Only notify the template creator if the build was initiated by someone else
+			if build.InitiatorID != template.CreatedBy {
+				s.notifyTemplateManualBuildFailed(ctx, workspace, owner, template, build)
+			}
+		}
 
 		err = s.Pubsub.Publish(codersdk.WorkspaceNotifyChannel(build.WorkspaceID), []byte{})
 		if err != nil {
@@ -1095,13 +1113,8 @@
 	return &proto.Empty{}, nil
 }
 
-func (s *server) notifyWorkspaceBuildFailed(ctx context.Context, workspace database.Workspace, build database.WorkspaceBuild) {
-	var reason string
-	if build.Reason.Valid() && build.Reason == database.BuildReasonInitiator {
-		return // failed workspace build initiated by a user should not notify
-	}
-	reason = string(build.Reason)
-
+func (s *server) notifyWorkspaceAutoBuildFailed(ctx context.Context, workspace database.Workspace, build database.WorkspaceBuild) {
+	reason := string(build.Reason)
 	if _, err := s.NotificationsEnqueuer.Enqueue(ctx, workspace.OwnerID, notifications.TemplateWorkspaceAutobuildFailed,
 		map[string]string{
 			"name":   workspace.Name,
@@ -1110,7 +1123,24 @@
 		// Associate this notification with all the related entities.
 		workspace.ID, workspace.OwnerID, workspace.TemplateID, workspace.OrganizationID,
 	); err != nil {
-		s.Logger.Warn(ctx, "failed to notify of failed workspace autobuild", slog.Error(err))
+		s.Logger.Warn(ctx, "failed to notify of failed workspace autobuild", slog.F("workspace_id", workspace.ID), slog.Error(err))
+	}
+}
+
+func (s *server) notifyTemplateManualBuildFailed(ctx context.Context, workspace database.Workspace, owner database.User, template database.Template, build database.WorkspaceBuild) {
+	if _, err := s.NotificationsEnqueuer.Enqueue(ctx, template.CreatedBy, notifications.TemplateTemplateManualBuildFailed,
+		map[string]string{
+			"name":              template.Name,
+			"workspaceName":     workspace.Name,
+			"transition":        string(build.Transition),
+			"initiator":         build.InitiatorByUsername,
+			"workspaceUserName": owner.Username,
+			"buildNumber":       strconv.FormatInt(int64(build.BuildNumber), 10),
+		}, "provisionerdserver",
+		// Associate this notification with all the related entities.
+		workspace.ID, workspace.OwnerID, workspace.TemplateID, workspace.OrganizationID, template.CreatedBy,
+	); err != nil {
+		s.Logger.Warn(ctx, "failed to notify of failed template autobuild", slog.F("template_id", template.ID), slog.F("workspace_id", workspace.ID), slog.Error(err))
 	}
 }
 
@@ -1585,6 +1615,42 @@
 	}
 }
 
+func (s *server) notifyTemplateOwnerAboutManualBuildFailure(ctx context.Context, workspace database.Workspace, build database.WorkspaceBuild) {
+	var reason string
+	initiator := build.InitiatorByUsername
+	if build.Reason.Valid() {
+		switch build.Reason {
+		case database.BuildReasonInitiator:
+			if build.InitiatorID == workspace.OwnerID {
+				// Deletions initiated by self should not notify.
+				return
+			}
+
+			reason = "initiated by user"
+		case database.BuildReasonAutodelete:
+			reason = "autodeleted due to dormancy"
+			initiator = "autobuild"
+		default:
+			reason = string(build.Reason)
+		}
+	} else {
+		reason = string(build.Reason)
+		s.Logger.Warn(ctx, "invalid build reason when sending deletion notification",
+			slog.F("reason", reason), slog.F("workspace_id", workspace.ID), slog.F("build_id", build.ID))
+	}
+
+	if _, err := s.NotificationsEnqueuer.Enqueue(ctx, workspace.OwnerID, notifications.TemplateWorkspaceDeleted,
+		map[string]string{
+			"name":      workspace.Name,
+			"reason":    reason,
+			"initiator": initiator,
+		}, "provisionerdserver",
+		// Associate this notification with all the related entities.
+		workspace.ID, workspace.OwnerID, workspace.TemplateID, workspace.OrganizationID,
+	); err != nil {
+		s.Logger.Warn(ctx, "failed to notify of workspace deletion", slog.Error(err))
+	}
+}
 func (s *server) startTrace(ctx context.Context, name string, opts ...trace.SpanStartOption) (context.Context, trace.Span) {
 	return s.Tracer.Start(ctx, name, append(opts, trace.WithAttributes(
 		semconv.ServiceNameKey.String("coderd.provisionerd"),

diff --git a/coderd/provisionerdserver/provisionerdserver_test.go b/coderd/provisionerdserver/provisionerdserver_test.go
@@ -1568,7 +1568,7 @@ func TestInsertWorkspaceResource(t *testing.T) {
 func TestNotifications(t *testing.T) {
 	t.Parallel()
 
-	t.Run("Workspace deletion", func(t *testing.T) {
+	t.Run("WorkspaceDeletion", func(t *testing.T) {
 		t.Parallel()
 
 		tests := []struct {
@@ -1695,7 +1695,7 @@ func TestNotifications(t *testing.T) {
 		}
 	})
 
-	t.Run("Workspace build failed", func(t *testing.T) {
+	t.Run("WorkspaceAutoBuildFailed", func(t *testing.T) {
 		t.Parallel()
 
 		tests := []struct {
@@ -1705,12 +1705,12 @@ func TestNotifications(t *testing.T) {
 			shouldNotify bool
 		}{
 			{
-				name:         "initiated by owner",
+				name:         "InitiatedByOwner",
 				buildReason:  database.BuildReasonInitiator,
 				shouldNotify: false,
 			},
 			{
-				name:         "initiated by autostart",
+				name:         "InitiatedByAutoStart",
 				buildReason:  database.BuildReasonAutostart,
 				shouldNotify: true,
 			},
@@ -1804,6 +1804,168 @@ func TestNotifications(t *testing.T) {
 			})
 		}
 	})
+
+	t.Run("TemplateManualBuildFailed", func(t *testing.T) {
+		t.Parallel()
+
+		t.Run("NotifyTemplateOwner", func(t *testing.T) {
+			t.Parallel()
+
+			ctx := context.Background()
+			notifEnq := &testutil.FakeNotificationsEnqueuer{}
+
+			//	Otherwise `(*Server).FailJob` fails with:
+			// audit log - get build {"error": "sql: no rows in result set"}
+			ignoreLogErrors := true
+			srv, db, ps, pd := setup(t, ignoreLogErrors, &overrides{
+				notificationEnqueuer: notifEnq,
+			})
+
+			initiator := dbgen.User(t, db, database.User{})
+			templateOwner := dbgen.User(t, db, database.User{})
+
+			template := dbgen.Template(t, db, database.Template{
+				Name:           "template",
+				Provisioner:    database.ProvisionerTypeEcho,
+				OrganizationID: pd.OrganizationID,
+				CreatedBy:      templateOwner.ID,
+			})
+			template, err := db.GetTemplateByID(ctx, template.ID)
+			require.NoError(t, err)
+			file := dbgen.File(t, db, database.File{CreatedBy: initiator.ID})
+			workspace := dbgen.Workspace(t, db, database.Workspace{
+				TemplateID:     template.ID,
+				OwnerID:        initiator.ID,
+				OrganizationID: pd.OrganizationID,
+			})
+			version := dbgen.TemplateVersion(t, db, database.TemplateVersion{
+				OrganizationID: pd.OrganizationID,
+				TemplateID: uuid.NullUUID{
+					UUID:  template.ID,
+					Valid: true,
+				},
+				JobID: uuid.New(),
+			})
+			build := dbgen.WorkspaceBuild(t, db, database.WorkspaceBuild{
+				WorkspaceID:       workspace.ID,
+				TemplateVersionID: version.ID,
+				InitiatorID:       initiator.ID,
+				Transition:        database.WorkspaceTransitionDelete,
+				Reason:            database.BuildReasonInitiator,
+			})
+			job := dbgen.ProvisionerJob(t, db, ps, database.ProvisionerJob{
+				FileID: file.ID,
+				Type:   database.ProvisionerJobTypeWorkspaceBuild,
+				Input: must(json.Marshal(provisionerdserver.WorkspaceProvisionJob{
+					WorkspaceBuildID: build.ID,
+				})),
+				OrganizationID: pd.OrganizationID,
+			})
+			_, err = db.AcquireProvisionerJob(ctx, database.AcquireProvisionerJobParams{
+				OrganizationID: pd.OrganizationID,
+				WorkerID: uuid.NullUUID{
+					UUID:  pd.ID,
+					Valid: true,
+				},
+				Types: []database.ProvisionerType{database.ProvisionerTypeEcho},
+			})
+			require.NoError(t, err)
+
+			_, err = srv.FailJob(ctx, &proto.FailedJob{
+				JobId: job.ID.String(),
+				Type: &proto.FailedJob_WorkspaceBuild_{
+					WorkspaceBuild: &proto.FailedJob_WorkspaceBuild{
+						State: []byte{},
+					},
+				},
+			})
+			require.NoError(t, err)
+
+			require.Len(t, notifEnq.Sent, 1)
+			require.Equal(t, notifEnq.Sent[0].UserID, templateOwner.ID)
+			require.Contains(t, notifEnq.Sent[0].Targets, template.ID)
+			require.Contains(t, notifEnq.Sent[0].Targets, workspace.ID)
+			require.Contains(t, notifEnq.Sent[0].Targets, workspace.OrganizationID)
+			require.Contains(t, notifEnq.Sent[0].Targets, templateOwner.ID)
+			require.Contains(t, notifEnq.Sent[0].Targets, initiator.ID)
+		})
+
+		t.Run("DoNotNotifyTheInitiator", func(t *testing.T) {
+			t.Parallel()
+
+			ctx := context.Background()
+			notifEnq := &testutil.FakeNotificationsEnqueuer{}
+
+			//	Otherwise `(*Server).FailJob` fails with:
+			// audit log - get build {"error": "sql: no rows in result set"}
+			ignoreLogErrors := true
+			srv, db, ps, pd := setup(t, ignoreLogErrors, &overrides{
+				notificationEnqueuer: notifEnq,
+			})
+
+			initiator := dbgen.User(t, db, database.User{})
+			templateOwner := initiator
+
+			template := dbgen.Template(t, db, database.Template{
+				Name:           "template",
+				Provisioner:    database.ProvisionerTypeEcho,
+				OrganizationID: pd.OrganizationID,
+				CreatedBy:      templateOwner.ID,
+			})
+			template, err := db.GetTemplateByID(ctx, template.ID)
+			require.NoError(t, err)
+			file := dbgen.File(t, db, database.File{CreatedBy: initiator.ID})
+			workspace := dbgen.Workspace(t, db, database.Workspace{
+				TemplateID:     template.ID,
+				OwnerID:        initiator.ID,
+				OrganizationID: pd.OrganizationID,
+			})
+			version := dbgen.TemplateVersion(t, db, database.TemplateVersion{
+				OrganizationID: pd.OrganizationID,
+				TemplateID: uuid.NullUUID{
+					UUID:  template.ID,
+					Valid: true,
+				},
+				JobID: uuid.New(),
+			})
+			build := dbgen.WorkspaceBuild(t, db, database.WorkspaceBuild{
+				WorkspaceID:       workspace.ID,
+				TemplateVersionID: version.ID,
+				InitiatorID:       initiator.ID,
+				Transition:        database.WorkspaceTransitionDelete,
+				Reason:            database.BuildReasonInitiator,
+			})
+			job := dbgen.ProvisionerJob(t, db, ps, database.ProvisionerJob{
+				FileID: file.ID,
+				Type:   database.ProvisionerJobTypeWorkspaceBuild,
+				Input: must(json.Marshal(provisionerdserver.WorkspaceProvisionJob{
+					WorkspaceBuildID: build.ID,
+				})),
+				OrganizationID: pd.OrganizationID,
+			})
+			_, err = db.AcquireProvisionerJob(ctx, database.AcquireProvisionerJobParams{
+				OrganizationID: pd.OrganizationID,
+				WorkerID: uuid.NullUUID{
+					UUID:  pd.ID,
+					Valid: true,
+				},
+				Types: []database.ProvisionerType{database.ProvisionerTypeEcho},
+			})
+			require.NoError(t, err)
+
+			_, err = srv.FailJob(ctx, &proto.FailedJob{
+				JobId: job.ID.String(),
+				Type: &proto.FailedJob_WorkspaceBuild_{
+					WorkspaceBuild: &proto.FailedJob_WorkspaceBuild{
+						State: []byte{},
+					},
+				},
+			})
+			require.NoError(t, err)
+
+			require.Len(t, notifEnq.Sent, 0)
+		})
+	})
 }
 
 type overrides struct {