Target Allocator implementation (Part 1 - OTEL Operator Enhancements) (…

…open-telemetry#351)

Co-authored-by: Alexis Perez <perzmen@amazon.com>
Co-authored-by: Anthony Mirabella <a9@aneurysm9.com>

.github/workflows/publish-images.yaml

@@ -24,6 +24,7 @@ jobs:
       - name: Set env vars for the job
         run: |
           grep -v '\#' versions.txt | grep opentelemetry-collector | awk -F= '{print "OTELCOL_VERSION="$2}' >> $GITHUB_ENV
+          grep -v '\#' versions.txt | grep targetallocator | awk -F= '{print "TARGETALLOCATOR_VERSION="$2}' >> $GITHUB_ENV
           echo "VERSION_DATE=$(date -u +'%Y-%m-%dT%H:%M:%SZ')" >> $GITHUB_ENV
           echo "VERSION=$(git describe --tags | sed 's/^v//')" >> $GITHUB_ENV
 
@@ -72,5 +73,6 @@ jobs:
             VERSION=${{ env.VERSION }}
             VERSION_DATE=${{ env.VERSION_DATE }}
             OTELCOL_VERSION=${{ env.OTELCOL_VERSION }}
+            TARGETALLOCATOR_VERSION=${{ env.TARGETALLOCATOR_VERSION }}
           cache-from: type=local,src=/tmp/.buildx-cache
           cache-to: type=local,dest=/tmp/.buildx-cache

Dockerfile

@@ -21,9 +21,10 @@ ARG VERSION_PKG
 ARG VERSION
 ARG VERSION_DATE
 ARG OTELCOL_VERSION
+ARG TARGETALLOCATOR_VERSION
 
 # Build
-RUN CGO_ENABLED=0 GOOS=linux GOARCH=amd64 GO111MODULE=on go build -ldflags="-X ${VERSION_PKG}.version=${VERSION} -X ${VERSION_PKG}.buildDate=${VERSION_DATE} -X ${VERSION_PKG}.otelCol=${OTELCOL_VERSION}" -a -o manager main.go
+RUN CGO_ENABLED=0 GOOS=linux GOARCH=amd64 GO111MODULE=on go build -ldflags="-X ${VERSION_PKG}.version=${VERSION} -X ${VERSION_PKG}.buildDate=${VERSION_DATE} -X ${VERSION_PKG}.otelCol=${OTELCOL_VERSION} -X ${VERSION_PKG}.targetAllocator=${TARGETALLOCATOR_VERSION}" -a -o manager main.go
 
 # Use distroless as minimal base image to package the manager binary
 # Refer to https://github.com/GoogleContainerTools/distroless for more details

Makefile

@@ -4,7 +4,8 @@ VERSION_DATE ?= $(shell date -u +'%Y-%m-%dT%H:%M:%SZ')
 VERSION_PKG ?= "github.com/open-telemetry/opentelemetry-operator/internal/version"
 OTELCOL_VERSION ?= "$(shell grep -v '\#' versions.txt | grep opentelemetry-collector | awk -F= '{print $$2}')"
 OPERATOR_VERSION ?= "$(shell grep -v '\#' versions.txt | grep operator | awk -F= '{print $$2}')"
-LD_FLAGS ?= "-X ${VERSION_PKG}.version=${VERSION} -X ${VERSION_PKG}.buildDate=${VERSION_DATE} -X ${VERSION_PKG}.otelCol=${OTELCOL_VERSION}"
+TARGETALLOCATOR_VERSION ?= "$(shell grep -v '\#' versions.txt | grep targetallocator | awk -F= '{print $$2}')"
+LD_FLAGS ?= "-X ${VERSION_PKG}.version=${VERSION} -X ${VERSION_PKG}.buildDate=${VERSION_DATE} -X ${VERSION_PKG}.otelCol=${OTELCOL_VERSION} -X ${VERSION_PKG}.targetAllocator=${TARGETALLOCATOR_VERSION}"
 
 # Image URL to use all building/pushing image targets
 IMG_PREFIX ?= quay.io/${USER}
@@ -122,7 +123,7 @@ set-test-image-vars:
 
 # Build the container image, used only for local dev purposes
 container:
-	docker build -t ${IMG} --build-arg VERSION_PKG=${VERSION_PKG} --build-arg VERSION=${VERSION} --build-arg VERSION_DATE=${VERSION_DATE} --build-arg OTELCOL_VERSION=${OTELCOL_VERSION} .
+	docker build -t ${IMG} --build-arg VERSION_PKG=${VERSION_PKG} --build-arg VERSION=${VERSION} --build-arg VERSION_DATE=${VERSION_DATE} --build-arg OTELCOL_VERSION=${OTELCOL_VERSION} --build-arg TARGETALLOCATOR_VERSION=${TARGETALLOCATOR_VERSION} .
 
 # Push the container image, used only for local dev purposes
 container-push:

api/v1alpha1/opentelemetrycollector_types.go

@@ -41,6 +41,11 @@ type OpenTelemetryCollectorSpec struct {
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors=true
 	Image string `json:"image,omitempty"`
 
+	// TargetAllocator indicates a value which determines whether to spawn a target allocation resource or not.
+	// +optional
+	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors=true
+	TargetAllocator OpenTelemetryTargetAllocatorSpec `json:"targetAllocator,omitempty"`
+
 	// Mode represents how the collector should be deployed (deployment, daemonset, statefulset or sidecar)
 	// +optional
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors=true
@@ -116,6 +121,17 @@ type OpenTelemetryCollectorStatus struct {
 	Messages []string `json:"messages,omitempty"`
 }
 
+// OpenTelemetryTargetAllocatorSpec defines the configurations for the Prometheus target allocator.
+type OpenTelemetryTargetAllocatorSpec struct {
+	// Enabled indicates whether to use a target allocation mechanism for Prometheus targets or not.
+	// +optional
+	Enabled bool `json:"enabled,omitempty"`
+
+	// Image indicates the container image to use for the OpenTelemetry TargetAllocator.
+	// +optional
+	Image string `json:"image,omitempty"`
+}
+
 // +kubebuilder:object:root=true
 // +kubebuilder:resource:shortName=otelcol;otelcols
 // +kubebuilder:subresource:status

api/v1alpha1/opentelemetrycollector_webhook.go

@@ -21,6 +21,8 @@ import (
 	ctrl "sigs.k8s.io/controller-runtime"
 	logf "sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/webhook"
+
+	ta "github.com/open-telemetry/opentelemetry-operator/pkg/targetallocator/adapters"
 )
 
 // log is for logging in this package.
@@ -91,5 +93,18 @@ func (r *OpenTelemetryCollector) validateCRDSpec() error {
 		return fmt.Errorf("the OpenTelemetry Collector mode is set to %s, which does not support the attribute 'tolerations'", r.Spec.Mode)
 	}
 
+	// validate target allocation
+	if r.Spec.TargetAllocator.Enabled && r.Spec.Mode != ModeStatefulSet {
+		return fmt.Errorf("the OpenTelemetry Collector mode is set to %s, which does not support the target allocation deployment", r.Spec.Mode)
+	}
+
+	// validate Prometheus config for target allocation
+	if r.Spec.TargetAllocator.Enabled {
+		_, err := ta.ConfigToPromConfig(r.Spec.Config)
+		if err != nil {
+			return fmt.Errorf("the OpenTelemetry Spec Prometheus configuration is incorrect, %s", err)
+		}
+	}
+
 	return nil
 }

bundle/manifests/opentelemetry.io_opentelemetrycollectors.yaml

@@ -442,6 +442,19 @@ spec:
                 description: ServiceAccount indicates the name of an existing service
                   account to use with this instance.
                 type: string
+              targetAllocator:
+                description: TargetAllocator indicates a value which determines whether
+                  to spawn a target allocation resource or not.
+                properties:
+                  enabled:
+                    description: Enabled indicates whether to use a target allocation
+                      mechanism for Prometheus targets or not.
+                    type: boolean
+                  image:
+                    description: Image indicates the container image to use for the
+                      OpenTelemetry TargetAllocator.
+                    type: string
+                type: object
               tolerations:
                 description: Toleration to schedule OpenTelemetry Collector pods.
                   This is only relevant to daemonsets, statefulsets and deployments

config/crd/bases/opentelemetry.io_opentelemetrycollectors.yaml

@@ -430,6 +430,19 @@ spec:
                 description: ServiceAccount indicates the name of an existing service
                   account to use with this instance.
                 type: string
+              targetAllocator:
+                description: TargetAllocator indicates a value which determines whether
+                  to spawn a target allocation resource or not.
+                properties:
+                  enabled:
+                    description: Enabled indicates whether to use a target allocation
+                      mechanism for Prometheus targets or not.
+                    type: boolean
+                  image:
+                    description: Image indicates the container image to use for the
+                      OpenTelemetry TargetAllocator.
+                    type: string
+                type: object
               tolerations:
                 description: Toleration to schedule OpenTelemetry Collector pods.
                   This is only relevant to daemonsets, statefulsets and deployments

internal/config/main.go

@@ -29,8 +29,9 @@ import (
 )
 
 const (
-	defaultAutoDetectFrequency     = 5 * time.Second
-	defaultCollectorConfigMapEntry = "collector.yaml"
+	defaultAutoDetectFrequency           = 5 * time.Second
+	defaultCollectorConfigMapEntry       = "collector.yaml"
+	defaultTargetAllocatorConfigMapEntry = "targetallocator.yaml"
 )
 
 // Config holds the static configuration for this operator.
@@ -44,21 +45,24 @@ type Config struct {
 	onChange            []func() error
 
 	// config state
-	collectorImage          string
-	collectorConfigMapEntry string
-	platform                platform.Platform
-	version                 version.Version
+	collectorImage                string
+	collectorConfigMapEntry       string
+	targetAllocatorImage          string
+	targetAllocatorConfigMapEntry string
+	platform                      platform.Platform
+	version                       version.Version
 }
 
 // New constructs a new configuration based on the given options.
 func New(opts ...Option) Config {
 	// initialize with the default values
 	o := options{
-		autoDetectFrequency:     defaultAutoDetectFrequency,
-		collectorConfigMapEntry: defaultCollectorConfigMapEntry,
-		logger:                  logf.Log.WithName("config"),
-		platform:                platform.Unknown,
-		version:                 version.Get(),
+		autoDetectFrequency:           defaultAutoDetectFrequency,
+		collectorConfigMapEntry:       defaultCollectorConfigMapEntry,
+		targetAllocatorConfigMapEntry: defaultTargetAllocatorConfigMapEntry,
+		logger:                        logf.Log.WithName("config"),
+		platform:                      platform.Unknown,
+		version:                       version.Get(),
 	}
 	for _, opt := range opts {
 		opt(&o)
@@ -70,15 +74,21 @@ func New(opts ...Option) Config {
 		o.collectorImage = fmt.Sprintf("otel/opentelemetry-collector:%s", o.version.OpenTelemetryCollector)
 	}
 
+	if len(o.targetAllocatorImage) == 0 {
+		o.targetAllocatorImage = fmt.Sprintf("quay.io/opentelemetry/target-allocator:%s", o.version.TargetAllocator)
+	}
+
 	return Config{
-		autoDetect:              o.autoDetect,
-		autoDetectFrequency:     o.autoDetectFrequency,
-		collectorImage:          o.collectorImage,
-		collectorConfigMapEntry: o.collectorConfigMapEntry,
-		logger:                  o.logger,
-		onChange:                o.onChange,
-		platform:                o.platform,
-		version:                 o.version,
+		autoDetect:                    o.autoDetect,
+		autoDetectFrequency:           o.autoDetectFrequency,
+		collectorImage:                o.collectorImage,
+		collectorConfigMapEntry:       o.collectorConfigMapEntry,
+		targetAllocatorImage:          o.targetAllocatorImage,
+		targetAllocatorConfigMapEntry: o.targetAllocatorConfigMapEntry,
+		logger:                        o.logger,
+		onChange:                      o.onChange,
+		platform:                      o.platform,
+		version:                       o.version,
 	}
 }
 
@@ -155,6 +165,16 @@ func (c *Config) CollectorConfigMapEntry() string {
 	return c.collectorConfigMapEntry
 }
 
+// TargetAllocatorImage represents the flag to override the OpenTelemetry TargetAllocator container image.
+func (c *Config) TargetAllocatorImage() string {
+	return c.targetAllocatorImage
+}
+
+// TargetAllocatorConfigMapEntry represents the configuration file name for the TargetAllocator. Immutable.
+func (c *Config) TargetAllocatorConfigMapEntry() string {
+	return c.targetAllocatorConfigMapEntry
+}
+
 // Platform represents the type of the platform this operator is running.
 func (c *Config) Platform() platform.Platform {
 	return c.platform

internal/config/options.go

@@ -28,14 +28,16 @@ import (
 type Option func(c *options)
 
 type options struct {
-	autoDetect              autodetect.AutoDetect
-	autoDetectFrequency     time.Duration
-	collectorImage          string
-	collectorConfigMapEntry string
-	logger                  logr.Logger
-	onChange                []func() error
-	platform                platform.Platform
-	version                 version.Version
+	autoDetect                    autodetect.AutoDetect
+	autoDetectFrequency           time.Duration
+	targetAllocatorImage          string
+	collectorImage                string
+	collectorConfigMapEntry       string
+	targetAllocatorConfigMapEntry string
+	logger                        logr.Logger
+	onChange                      []func() error
+	platform                      platform.Platform
+	version                       version.Version
 }
 
 func WithAutoDetect(a autodetect.AutoDetect) Option {
@@ -48,6 +50,12 @@ func WithAutoDetectFrequency(t time.Duration) Option {
 		o.autoDetectFrequency = t
 	}
 }
+func WithTargetAllocatorImage(s string) Option {
+	return func(o *options) {
+		o.targetAllocatorImage = s
+	}
+}
+
 func WithCollectorImage(s string) Option {
 	return func(o *options) {
 		o.collectorImage = s
@@ -58,6 +66,11 @@ func WithCollectorConfigMapEntry(s string) Option {
 		o.collectorConfigMapEntry = s
 	}
 }
+func WithTargetAllocatorConfigMapEntry(s string) Option {
+	return func(o *options) {
+		o.targetAllocatorConfigMapEntry = s
+	}
+}
 func WithLogger(logger logr.Logger) Option {
 	return func(o *options) {
 		o.logger = logger

internal/version/main.go

@@ -21,9 +21,10 @@ import (
 )
 
 var (
-	version   string
-	buildDate string
-	otelCol   string
+	version         string
+	buildDate       string
+	otelCol         string
+	targetAllocator string
 )
 
 // Version holds this Operator's version as well as the version of some of the components it uses.
@@ -32,6 +33,7 @@ type Version struct {
 	BuildDate              string `json:"build-date"`
 	OpenTelemetryCollector string `json:"opentelemetry-collector-version"`
 	Go                     string `json:"go-version"`
+	TargetAllocator        string `json:"target-allocator-version"`
 }
 
 // Get returns the Version object with the relevant information.
@@ -41,16 +43,18 @@ func Get() Version {
 		BuildDate:              buildDate,
 		OpenTelemetryCollector: OpenTelemetryCollector(),
 		Go:                     runtime.Version(),
+		TargetAllocator:        TargetAllocator(),
 	}
 }
 
 func (v Version) String() string {
 	return fmt.Sprintf(
-		"Version(Operator='%v', BuildDate='%v', OpenTelemetryCollector='%v', Go='%v')",
+		"Version(Operator='%v', BuildDate='%v', OpenTelemetryCollector='%v', Go='%v', TargetAllocator='%v')",
 		v.Operator,
 		v.BuildDate,
 		v.OpenTelemetryCollector,
 		v.Go,
+		v.TargetAllocator,
 	)
 }
 
@@ -64,3 +68,14 @@ func OpenTelemetryCollector() string {
 	// fallback value, useful for tests
 	return "0.0.0"
 }
+
+// TargetAllocator returns the default TargetAllocator to use when no versions are specified via CLI or configuration.
+func TargetAllocator() string {
+	if len(targetAllocator) > 0 {
+		// this should always be set, as it's specified during the build
+		return targetAllocator
+	}
+
+	// fallback value, useful for tests
+	return "0.0.0"
+}

internal/version/main_test.go

@@ -34,3 +34,18 @@ func TestVersionFromBuild(t *testing.T) {
 	assert.Equal(t, otelCol, OpenTelemetryCollector())
 	assert.Contains(t, Get().String(), otelCol)
 }
+
+func TestTargetAllocatorFallbackVersion(t *testing.T) {
+	assert.Equal(t, "0.0.0", TargetAllocator())
+}
+
+func TestTargetAllocatorVersionFromBuild(t *testing.T) {
+	// prepare
+	targetAllocator = "0.0.2" // set during the build
+	defer func() {
+		targetAllocator = ""
+	}()
+
+	assert.Equal(t, targetAllocator, TargetAllocator())
+	assert.Contains(t, Get().String(), targetAllocator)
+}

main.go

@@ -76,6 +76,7 @@ func main() {
 	logger.Info("Starting the OpenTelemetry Operator",
 		"opentelemetry-operator", v.Operator,
 		"opentelemetry-collector", v.OpenTelemetryCollector,
+		"opentelemetry-targetallocator", v.TargetAllocator,
 		"build-date", v.BuildDate,
 		"go-version", v.Go,
 		"go-arch", runtime.GOARCH,