Merge pull request #421 from vprashar2929/bump-model-svr

chore: Bump up model-server to v0.7.11-2
sustainable-computing-io · Sep 26, 2024 · e9da953 · e9da953
2 parents a895fb9 + d460128
commit e9da953
Show file tree

Hide file tree

Showing 6 changed files with 76 additions and 10 deletions.
diff --git a/pkg/components/estimator/estimator.go b/pkg/components/estimator/estimator.go
@@ -28,12 +28,12 @@ import (
 
 const (
 	// NOTE: update tests/images.yaml when changing this image
-	StableImage          = "quay.io/sustainable_computing_io/kepler_model_server:v0.7.11"
+	StableImage          = "quay.io/sustainable_computing_io/kepler_model_server:v0.7.11-2"
 	waitForSocketCommand = "until [ -e /tmp/estimator.sock ]; do sleep 1; done && %s"
 )
 
 var (
-	shellCommand = []string{"/bin/sh", "-c"}
+	shellCommand = []string{"/usr/bin/bash", "-c"}
 )
 
 // NeedsEstimatorSidecar returns true if any of estimator config has sidecar enabled
@@ -76,8 +76,8 @@ func Container(image string) corev1.Container {
 		ImagePullPolicy: corev1.PullIfNotPresent,
 		Name:            "estimator",
 		VolumeMounts:    mounts,
-		Command:         []string{"python3.10"},
-		Args:            []string{"-u", "src/estimate/estimator.py"},
+		Command:         []string{"estimator"},
+		Args:            []string{"-l", "info"},
 	}
 }
 

diff --git a/pkg/components/estimator/estimator_test.go b/pkg/components/estimator/estimator_test.go
@@ -109,7 +109,7 @@ func TestModifiedContainer(t *testing.T) {
 	exporterContainer := &corev1.Container{
 		Command: []string{"kepler", "-v=1"},
 	}
-	expectedCommand := []string{"/bin/sh", "-c"}
+	expectedCommand := []string{"/usr/bin/bash", "-c"}
 	expectedArgs := []string{"until [ -e /tmp/estimator.sock ]; do sleep 1; done && kepler -v=1"}
 	expectedVolumeMounts := []string{"cfm", "mnt", "tmp"}
 	keplerVolumes := []corev1.Volume{k8s.VolumeFromEmptyDir("kepler-volume")}

diff --git a/pkg/components/exporter/exporter.go b/pkg/components/exporter/exporter.go
@@ -102,6 +102,11 @@ func NewDaemonSet(detail components.Detail, k *v1alpha1.KeplerInternal) *appsv1.
 		k8s.VolumeFromConfigMap("cfm", k.Name),
 	} // exporter default Volumes
 
+	ms := k.Spec.ModelServer
+	if ms != nil && ms.Enabled {
+		addModelServerCommand(&exporterContainer, k.ModelServerDeploymentName(), k.Namespace(), ms)
+	}
+
 	if estimator.NeedsEstimatorSidecar(k.Spec.Estimator) {
 		// add sidecar container and update kepler-exporter container
 		// add shared volumes
@@ -656,7 +661,17 @@ func newExporterContainer(kiName, dsName string, deployment v1alpha1.InternalExp
 func addEstimatorSidecar(estimatorImage string, exporterContainer *corev1.Container, volumes []corev1.Volume) ([]corev1.Container, []corev1.Volume) {
 	sidecarContainer := estimator.Container(estimatorImage)
 	volumes = append(volumes, estimator.Volumes()...)
+	// NOTE: if the exporter has arguments (e.g., for an estimator with model server deployment),
+	// update the exporter's container command with the arguments. This allows AddEstimatorDependency
+	// to append the necessary socket wait command to the arguments
+	if exporterContainer.Args != nil {
+		exporterContainer.Command = exporterContainer.Args
+	}
 	exporterContainer = estimator.AddEstimatorDependency(exporterContainer)
 	containers := []corev1.Container{*exporterContainer, sidecarContainer}
 	return containers, volumes
 }
+
+func addModelServerCommand(exporterContainer *corev1.Container, deployName, deployNamespace string, ms *v1alpha1.InternalModelServerSpec) {
+	modelserver.AddModelServerDependency(exporterContainer, deployName, deployNamespace, ms)
+}
diff --git a/pkg/components/modelserver/modelserver.go b/pkg/components/modelserver/modelserver.go
@@ -18,6 +18,7 @@ package modelserver
 
 import (
 	"fmt"
+	"strings"
 
 	"github.com/sustainable.computing.io/kepler-operator/api/v1alpha1"
 	"github.com/sustainable.computing.io/kepler-operator/pkg/components"
@@ -37,8 +38,9 @@ const (
 )
 
 const (
-	defaultModelServer = "http://%s.%s.svc.cluster.local:%d"
-	StableImage        = "quay.io/sustainable_computing_io/kepler_model_server:v0.7.11"
+	defaultModelServer        = "http://%s.%s.svc.cluster.local:%d"
+	StableImage               = "quay.io/sustainable_computing_io/kepler_model_server:v0.7.11-2"
+	waitForModelServerCommand = "until [[ \"$(curl -s -o /dev/null -w %%{http_code} %s/best-models)\" -eq 200 ]]; do sleep 1; done"
 )
 
 var (
@@ -53,6 +55,10 @@ var (
 	})
 )
 
+var (
+	shellCommand = []string{"/usr/bin/bash", "-c"}
+)
+
 func NewDeployment(deployName string, ms *v1alpha1.InternalModelServerSpec, namespace string) *appsv1.Deployment {
 	pvcName := deployName + PVCNameSuffix
 	configMapName := deployName + ConfigMapSuffix
@@ -86,8 +92,8 @@ func NewDeployment(deployName string, ms *v1alpha1.InternalModelServerSpec, name
 			Name:          "http",
 		}},
 		VolumeMounts: mounts,
-		Command:      []string{"python3.10"},
-		Args:         []string{"-u", "src/server/model_server.py"},
+		Command:      []string{"model-server"},
+		Args:         []string{"-l", "info"},
 	}}
 
 	return &appsv1.Deployment{
@@ -218,3 +224,20 @@ func serverUrl(deployName, deployNamespace string, ms v1alpha1.InternalModelServ
 	serviceName := deployName + ServiceSuffix
 	return fmt.Sprintf(defaultModelServer, serviceName, deployNamespace, port)
 }
+
+func formatModelServerCommand(deployName, deployNamespace string, ms v1alpha1.InternalModelServerSpec) string {
+	return fmt.Sprintf(waitForModelServerCommand, serverUrl(deployName, deployNamespace, ms))
+}
+
+func addModelServerWaitCmd(exporterContainer *corev1.Container, deployName, deployNamespace string, ms v1alpha1.InternalModelServerSpec) *corev1.Container {
+	cmd := exporterContainer.Command
+	exporterContainer.Command = shellCommand
+	exporterContainer.Args = []string{fmt.Sprintf("%s && %s", formatModelServerCommand(deployName, deployNamespace, ms), strings.Join(cmd, " "))}
+	return exporterContainer
+
+}
+
+func AddModelServerDependency(exporterContainer *corev1.Container, deployName, deployNamespace string, ms *v1alpha1.InternalModelServerSpec) *corev1.Container {
+	exporterContainer = addModelServerWaitCmd(exporterContainer, deployName, deployNamespace, *ms)
+	return exporterContainer
+}
diff --git a/pkg/components/modelserver/modelserver_test.go b/pkg/components/modelserver/modelserver_test.go
@@ -22,6 +22,7 @@ import (
 	"github.com/stretchr/testify/assert"
 	"github.com/sustainable.computing.io/kepler-operator/api/v1alpha1"
 	"github.com/sustainable.computing.io/kepler-operator/pkg/components"
+	corev1 "k8s.io/api/core/v1"
 )
 
 func TestConfigMap(t *testing.T) {
@@ -142,3 +143,30 @@ func TestServerAPIContainer(t *testing.T) {
 	}
 
 }
+
+func TestServerContainer(t *testing.T) {
+	exporterContainer := &corev1.Container{
+		Command: []string{"kepler", "-v=1"},
+	}
+	ms := v1alpha1.InternalModelServerSpec{
+		Port: 8080,
+	}
+	deployName := "kepler-internal"
+	deployNamespace := "default"
+	expectedCommand := []string{"/usr/bin/bash", "-c"}
+	expectedArgs := []string{"until [[ \"$(curl -s -o /dev/null -w %{http_code} http://kepler-internal-svc.default.svc.cluster.local:8080/best-models)\" -eq 200 ]]; do sleep 1; done && kepler -v=1"}
+
+	t.Run("server container", func(t *testing.T) {
+		t.Parallel()
+		_ = v1alpha1.KeplerInternal{
+			Spec: v1alpha1.KeplerInternalSpec{
+				ModelServer: &ms,
+			},
+		}
+		exporterContainer := AddModelServerDependency(exporterContainer, deployName, deployNamespace, &ms)
+		actualCommand := exporterContainer.Command
+		actualArgs := exporterContainer.Args
+		assert.Equal(t, expectedCommand, actualCommand)
+		assert.Equal(t, expectedArgs, actualArgs)
+	})
+}
diff --git a/tests/images.yaml b/tests/images.yaml
@@ -1,3 +1,3 @@
 images:
 - component: 'model-server'
-  image: 'quay.io/sustainable_computing_io/kepler_model_server:v0.7.11'
+  image: 'quay.io/sustainable_computing_io/kepler_model_server:v0.7.11-2'