add status handlers for dcgm and neuron (#169)

aws · May 15, 2024 · eb43162 · eb43162
1 parent b1f2465
commit eb43162
Show file tree

Hide file tree

Showing 6 changed files with 135 additions and 8 deletions.
diff --git a/controllers/dcgmexporter_controller.go b/controllers/dcgmexporter_controller.go
@@ -17,12 +17,11 @@ import (
 	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
-	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests/dcgmexporter"
-
 	"github.com/aws/amazon-cloudwatch-agent-operator/apis/v1alpha1"
 	"github.com/aws/amazon-cloudwatch-agent-operator/internal/config"
 	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests"
-	collectorStatus "github.com/aws/amazon-cloudwatch-agent-operator/internal/status/collector"
+	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests/dcgmexporter"
+	dcgmexporterStatus "github.com/aws/amazon-cloudwatch-agent-operator/internal/status/dcgmexporter"
 )
 
 // DcgmExporterReconciler reconciles a DcgmExporter object.
@@ -108,7 +107,7 @@ func (r *DcgmExporterReconciler) Reconcile(ctx context.Context, req ctrl.Request
 	}
 
 	err := reconcileDesiredObjects(ctx, r.Client, log, &params.DcgmExp, params.Scheme, desiredObjects...)
-	return collectorStatus.HandleReconcileStatus(ctx, log, params, err)
+	return dcgmexporterStatus.HandleReconcileStatus(ctx, log, params, err)
 }
 
 // BuildDcgmExporter returns the generation and collected errors of all manifests for a given instance.

diff --git a/controllers/neuronmonitor_controller.go b/controllers/neuronmonitor_controller.go
@@ -17,12 +17,11 @@ import (
 	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
-	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests/neuronmonitor"
-
 	"github.com/aws/amazon-cloudwatch-agent-operator/apis/v1alpha1"
 	"github.com/aws/amazon-cloudwatch-agent-operator/internal/config"
 	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests"
-	collectorStatus "github.com/aws/amazon-cloudwatch-agent-operator/internal/status/collector"
+	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests/neuronmonitor"
+	neuronmonitorStatus "github.com/aws/amazon-cloudwatch-agent-operator/internal/status/neuronmonitor"
 )
 
 // NeuronMonitorReconciler reconciles a NeuronMonitor object.
@@ -107,7 +106,7 @@ func (r *NeuronMonitorReconciler) Reconcile(ctx context.Context, req ctrl.Reques
 		return ctrl.Result{}, nil
 	}
 	err := reconcileDesiredObjects(ctx, r.Client, log, &params.NeuronExp, params.Scheme, desiredObjects...)
-	return collectorStatus.HandleReconcileStatus(ctx, log, params, err)
+	return neuronmonitorStatus.HandleReconcileStatus(ctx, log, params, err)
 }
 
 // BuildNeuronMonitor returns the generation and collected errors of all manifests for a given instance.

diff --git a/internal/status/dcgmexporter/dcgmexporter.go b/internal/status/dcgmexporter/dcgmexporter.go
@@ -0,0 +1,21 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+package dcgmexporter
+
+import (
+	"context"
+
+	"sigs.k8s.io/controller-runtime/pkg/client"
+
+	"github.com/aws/amazon-cloudwatch-agent-operator/apis/v1alpha1"
+	"github.com/aws/amazon-cloudwatch-agent-operator/internal/version"
+)
+
+func UpdateDcgmExporterStatus(ctx context.Context, cli client.Client, changed *v1alpha1.DcgmExporter) error {
+	if changed.Status.Version == "" {
+		changed.Status.Version = version.DcgmExporter()
+	}
+
+	return nil
+}
diff --git a/internal/status/dcgmexporter/handle.go b/internal/status/dcgmexporter/handle.go
@@ -0,0 +1,44 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+package dcgmexporter
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/go-logr/logr"
+	ctrl "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+
+	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests"
+)
+
+const (
+	eventTypeNormal  = "Normal"
+	eventTypeWarning = "Warning"
+
+	reasonError         = "Error"
+	reasonStatusFailure = "StatusFailure"
+	reasonInfo          = "Info"
+)
+
+func HandleReconcileStatus(ctx context.Context, log logr.Logger, params manifests.Params, err error) (ctrl.Result, error) {
+	log.V(2).Info("updating dcgmexporter status")
+	if err != nil {
+		params.Recorder.Event(&params.OtelCol, eventTypeWarning, reasonError, err.Error())
+		return ctrl.Result{}, err
+	}
+	changed := params.DcgmExp.DeepCopy()
+	statusErr := UpdateDcgmExporterStatus(ctx, params.Client, changed)
+	if statusErr != nil {
+		params.Recorder.Event(changed, eventTypeWarning, reasonStatusFailure, statusErr.Error())
+		return ctrl.Result{}, statusErr
+	}
+	statusPatch := client.MergeFrom(&params.DcgmExp)
+	if err := params.Client.Status().Patch(ctx, changed, statusPatch); err != nil {
+		return ctrl.Result{}, fmt.Errorf("failed to apply status changes to the DcgmExporter CR: %w", err)
+	}
+	params.Recorder.Event(changed, eventTypeNormal, reasonInfo, "applied status changes")
+	return ctrl.Result{}, nil
+}
diff --git a/internal/status/neuronmonitor/handle.go b/internal/status/neuronmonitor/handle.go
@@ -0,0 +1,44 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+package neuronmonitor
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/go-logr/logr"
+	ctrl "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+
+	"github.com/aws/amazon-cloudwatch-agent-operator/internal/manifests"
+)
+
+const (
+	eventTypeNormal  = "Normal"
+	eventTypeWarning = "Warning"
+
+	reasonError         = "Error"
+	reasonStatusFailure = "StatusFailure"
+	reasonInfo          = "Info"
+)
+
+func HandleReconcileStatus(ctx context.Context, log logr.Logger, params manifests.Params, err error) (ctrl.Result, error) {
+	log.V(2).Info("updating neuronmonitor status")
+	if err != nil {
+		params.Recorder.Event(&params.NeuronExp, eventTypeWarning, reasonError, err.Error())
+		return ctrl.Result{}, err
+	}
+	changed := params.NeuronExp.DeepCopy()
+	statusErr := UpdateNeuronMonitorStatus(ctx, params.Client, changed)
+	if statusErr != nil {
+		params.Recorder.Event(changed, eventTypeWarning, reasonStatusFailure, statusErr.Error())
+		return ctrl.Result{}, statusErr
+	}
+	statusPatch := client.MergeFrom(&params.NeuronExp)
+	if err := params.Client.Status().Patch(ctx, changed, statusPatch); err != nil {
+		return ctrl.Result{}, fmt.Errorf("failed to apply status changes to the NeuronMonitor CR: %w", err)
+	}
+	params.Recorder.Event(changed, eventTypeNormal, reasonInfo, "applied status changes")
+	return ctrl.Result{}, nil
+}
diff --git a/internal/status/neuronmonitor/neuronmonitor.go b/internal/status/neuronmonitor/neuronmonitor.go
@@ -0,0 +1,20 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+package neuronmonitor
+
+import (
+	"context"
+
+	"sigs.k8s.io/controller-runtime/pkg/client"
+
+	"github.com/aws/amazon-cloudwatch-agent-operator/apis/v1alpha1"
+	"github.com/aws/amazon-cloudwatch-agent-operator/internal/version"
+)
+
+func UpdateNeuronMonitorStatus(ctx context.Context, cli client.Client, changed *v1alpha1.NeuronMonitor) error {
+	if changed.Status.Version == "" {
+		changed.Status.Version = version.NeuronMonitor()
+	}
+	return nil
+}